Skip to main content
Welcome guest. | Register | Login | Post

Alternative points system

29 replies [Last post]
libervisco's picture
Offline
Joined: 2006-05-04

The current points system is described here.

However I just thought of an alternative and potentially much simpler way of determining the best player of the game fest - the ratio between matches they played and matches they won.

This would potentially remove all the complexity associated with the game type, mode of play, team play etc.. All we would have to determine is what counts as a win and what counts as a match played.

As a match would count each match an individual player participated in, no matter the game type or mode, including playing as part of a team.

As a win would count each match an individual player won, including matches (s)he played as part of a team. Essentially if (s)he played as part of a team and that team won, this win will count as his and every other team member's win (this does encourage team members and leaders to really act like a team so each would deserve the win they get).

The fact that this is not a mere count of wins, but rather a ratio between wins and matches played resolves the issue of someone winning simply because he could afford more time to play (say played a multiple game components). With a ratio things like that are a non-issue.

What do you think?

Offline
Joined: 2007-09-10
How do you determine the

How do you determine the players who go into the final phase in this case?

I thought the number of games played (in one "gameslot") would be the same for all during the qualification phase. For example 5 matches guaranteed for everyone, and then those that have collected the most points go into the final phase.
Of course this system is designed for each gameslot separately.

Since you are planning an overall winner and additional "betting matches", this win/played ratio system makes sense.
But those it mean that the points system will be completely removed or is it an additional system only for the overall rating?

And as for the overall ratings, I think it was also mentioned somewhere else, I'm all for special awards like best teamplay, best sharpshooter, pimpest play, etc
But those should be awarded through a vote by all participants.

["pimpest plays" is a reference to those amazing starcraft replays showing of the skills of korean gosu gamers for those who know this. Eye
I don't know if something similar will be seen in this gamefest (maybe amazing frags or such), but I hope so. Smiling ]

libervisco's picture
Offline
Joined: 2006-05-04
KIAaze wrote: But those it
KIAaze wrote:

But those it mean that the points system will be completely removed or is it an additional system only for the overall rating?

I thought it to be a replacement rather than an addition, considering that the old system was meant to award points to every individual, regardless of whether (s)he plays in a team or not which makes it rather inefficient as an overall "universal" points system.

However, we can still have separate points system for each game slot. Now that it isn't tied to being used as universal we can make it more fitting. So instead of awarding points to individuals in Tremulous we can just award it to teams, for the purpose of qualifications. See my last post in the tremulous thread. If we go for splitting all teams into groups wins can have 3 points, draws 2 points, losses 1 point and forfeits 0, or something like that. It would have effect only on the qualifications.

So in a sense, the answer to your question is that it is the latter, it is an additional system, but making the old system different in that it is completely dependable on the given game slot's qualifications stage.

KIAaze wrote:

And as for the overall ratings, I think it was also mentioned somewhere else, I'm all for special awards like best teamplay, best sharpshooter, pimpest play, etc
But those should be awarded through a vote by all participants.

Indeed, this can't be determined by any of the points systems really, so it is left to voting. I suppose we can establish some polls where people can vote, but we first have to know who to nominate for such polls, which is again something the community of players should do. If we don't have spectators then the only ones who can nominate are players themselves who seen other players participate (basically meaning that they've either played a match with them or against them). Teams could have their own nominees to put forth and individuals can nominate either of the opponents they played with.

It's something to consider. Smiling

Offline
Joined: 2007-09-10
Points to keep in mind.. and don't throw complexity out too soon

If we do a winning percentage calculation, we have to consider having minimum "at bats" to prevent a player from winning the first game played (or first whatever played) and then sitting out through the rest of the fest.

We can have different scoring awards, but we do have to finalize something that will be used throughout the majority of a particular tournament. I think having multiple tournaments allows us to experiment with point sytems throughout the fest.

We may also want to try a variation tournament with one of the more popular games that is played early on. Bringing that game back towards the end of the fest would be interesting. Many would get a second opportunity to do things differently.

I think we should also encourage participation in various games whenever possible. Lack of interest may be because of a lack of awareness. Competition among players that are all weak can be very interesting for those participating (depends on the game) even if it would be boring to watch.

To deal with those that have less time, we can have a system where the best (eg) 4 scores (tournament results, etc) are taken. This way the person playing 4 such matches/games/tournaments/whatever theoretically has as much of a chance to finish first as anyone else. Naturally, if you can play more, you have more lottery draws.

Also, some tournament/matches will be easier to win than others. We may want to consider a system where we take into account quality of opponents (this may be doable easily or not, don't assume it will be complex). Maybe as a simplification, tournament points used to compute an overall winner may be scaled/normalized proportionately with the number of participants. Eg, winning in a tournament of 16 may carry 16 points, coming in second may award 15, etc.

Complexity in calculation is not an issue because computers do the work virtually instantaneously. Where we should avoid complexity is in things we cannot automate. Maybe counting frags or wins can be done automatically and maby not. In one case we can deal with a complex system in the other we shouldn't. Before eliminating complexity unnecessarily, consider how much you would prefer to be having this discussion over snail mail. Email and a good computer platform are complex, but that is alright if the end user doesn't really see it.

Offline
Joined: 2007-09-10
http://planetnexuiz.com/ladd
libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote: If we do a
Jose wrote:

If we do a winning percentage calculation, we have to consider having minimum "at bats" to prevent a player from winning the first game played (or first whatever played) and then sitting out through the rest of the fest.

I thought of that, and even wanted to reply to my original post with an answer to that anticipated question. I then decided I don't need to post any special explanations since it's so simple. Everyone will play a sufficient number of games because they will have to. Otherwise they'll be forfeited and yes we certainly wont count one win on one game as worthy of including in any counts. Bottom line, if a player signed up he's expected to go through the end, wherever it may be in both qualifications and final tournament.

Jose wrote:

I think having multiple tournaments allows us to experiment with point sytems throughout the fest.

We have multiple tournaments, but only one system will be employed across all to determine the best player of all that participated in any of the qualifications/tournaments AKA "game components" or "game slots".

Points systems that are specific to each game can merely be used to come up with a qualifications chart at the end of the qualifications stage.

Jose wrote:

We may also want to try a variation tournament with one of the more popular games that is played early on. Bringing that game back towards the end of the fest would be interesting. Many would get a second opportunity to do things differently.

I'm not sure what you mean here? To repeat a tournament, but with different settings? If yes, well.. I guess that would depend on the interest, but something like that can't be guaranteed and it could very well have adverse effects. People generally expect one qualifications/tournament set per each game component and then to see who comes on top of each and who comes on top as the best player. Some people choosing to repeat with a variation and others not could create a bit of adversity on part of those who don't want to go through this because those who do will get a chance to get better wins ratio..

Not sure..

Jose wrote:

I think we should also encourage participation in various games whenever possible. Lack of interest may be because of a lack of awareness. Competition among players that are all weak can be very interesting for those participating (depends on the game) even if it would be boring to watch.

I agree it may be lack of awareness, but we did and we will once again point out that other games besides the current three are basically "ready and waiting" for people to come, have a deal regarding the rules and initiate the tournament for these too.. But so far.. we just can't with confidence be announcing the start of these games - not just yet. We'll encourage and see what happens.

Anyway, some interesting ideas further on, but (and I do speak for myself, if anyone thinks otherwise feel free to say) I'm not looking to change much right now, but rather consolidate on what we have and come up with the final guidelines for the overall fest and each game included.

Offline
Joined: 2007-09-10
Points for breakfast, points for lunch, points for any occasion

I was about to post and then realized this is another long post, so here is an outline:
-- clarify earlier comments, include pros and cons. Also relate to the actual rules if possible.
-- -- at bats
-- -- different point systems
-- -- tournament encore/ variation tournament
-- mention some other related items
-- -- encouraging less popular games
-- -- adding more types of recognitions beyond one winner per tournament or fest
-- -- recording as many stats as possible for later processing
-- -- about some potential applications for taking advantage of recorded stats in the future

The minimum at bats comment was intended as symbolic for the potential introduction of a loophole if the minimum criteria had already been set prior to this posting on using winning perc average. [In English: I haven't have time to keep up with the wiki pages on the rules and was just trying to caution to have a min at bats that was reasonable. It really wasn't worth the entirety of two cents. It sort of just came out on reflex.]

One reason to possibly have different point rules apply to different tournaments would be that some systems may thrive with many participants but be trashy if there are only a few participants. Allowing for the possibility of a different system may make it easier to include more games, the less popular ones, in the fest. [I have no specific examples to give right now.] If we go a step further and actually try to have some variation, it gives us the opportunity to experiment a bit to help make future tournaments better. It would be recognizing that a little bit of rough edges might be best long term, and it might be wise to hedge a bit (not put all eggs in one basket).

I did recently look at the point system being drawn into the rules. I do think it is decent, generally, but I don't know enough about all the specific games so as to take a more definite stand.

The variation tournament may not be a good idea. I was thinking of something like having the audience stand up and ask for an encore (or get one to their surprise). Perhaps if it fell completely outside of the calculations for Fest Ultimate Player. Think of an all-star game. Also, it might even contribute but in a very small way. It may not affect some points calculations/qualifications but may it may affect others. I am being vague on purpose because I don't know how final some decisions are and because I am only offering some ideas and not the complete solution package.

Encouraging other games is something that can be done through (eg) a point system. Already the fest ultimate player looks to be based on the results of the various game-mode tournaments in some way. The question I ask is how to structure the weight of the various tournament contributions to perhaps encourage participation in less popular tournaments while recognizing that the popular tournaments should have greater weight. I think currently there are points given for things like a victory. Well, if a small tournament has less games played, that would automatically be a way to scale downward the contributions from that tournament. Simultaneously, it would encourage participation since it offers a source of extra points. If I read the draft rules correctly, I think this is currently what we have to work with. So if we make the number of games in a tournament correlate with the number of participants, we would achieve some sort of balance.

One reason to encourage participation would be for the benefit of those that do like those games and want a greater field. Tournament play may even help win some people over to a game they may not have given much attention to otherwise. This may be good or may not be that big a deal.

I want to say something in favor of having numerous recognitions besides tournament winners and a fest winner. With this variation, you can recognize more things without sacrificing others. This may be prefered by the players (especially if there are a lot of them) as it gives more opportunity to take home something besides the memory of having had a super great time (TM). It may also be a way to experiment to see what ends up being a more popular or well-regarded point system [it's actually possible we would anticipate incorrectly.. gasp]. For example, if one point system gave equal weight to all tournaments for choosing the fest winner while another system gave more weight to some tournaments over others, then we would be able to recognize players with different strengths. In the first case, the best all around player might win. In the second case, a good all around player that wasn't the best (if we weigh all tournaments equally) but who did really well in the more popular tournaments might win. Which is the better thing to have? Well, we could pick one to be the official fest winner but still recognize the other as having won some other title. We would have to be careful about having recognition dilution by having too many recognitions. [An example of the second type of point system just mentioned would be if the points are based on victories, some tournaments have more chances for victories than others, and the victories are worth the same in all tournaments. In this case, tournaments with more games get a greater weight in determining the fest winner.]

Another thing is that, if we make good measurements, recognition can be done after the fact (ie, once the tournament is over). If we don't make good recordings, then "wysiwyg" by the time the tournament ends.

As an example, consider baseball and stats that are dreamt up years after they were recorded. Maybe no one thought that ERA was a good measure of a pitcher's worth until 20 years after the information needed to computer ERA started to be recorded. In that case, there were 20 years that can be re-evaluated instead of starting from scratch when ERA is recognized and leaving an extra hole wrt the past. [The ERA example is hypothetical. I don't know it's history. To calculate it in a way that is distinguished from simple runs scored, you need things like the concept of errors. You would need errors to have been identified at the time they happened. You also need to measure who pitched each ball (as opposed simply to who started the inning or was in there the longest) and which pitches resulted in hits, walks, etc.]

In this vein, I propose that we record (a la screencast or through an in game mechanism) as many games as possible (and archive them). If that is not a good practical option or is not even possible for some games, we should still try to record as many details about the game as can be practically verified. Perhaps each player (or an observer) would do a voluntary write-up and have it checked off by the players (or by the observer). We should try to include interesting events, maybe take a screenshot of the final tallies, or do something along these lines.

FWIW, I am working on (as I find time) a file format (eg, in XML) that can hold a lot of information about the fest. This way we can do a write up and archive it. We can accumulate data from multiple tournaments (even from game tournaments unrealted to this one). Players and tournament sponsors/organizers can thus begin to build a record. After the fact, applications can process the data to come up with any statistic imaginable (that can be abstracted from the data that was preserved).

Some questions we might be curious about later on: Who has had the most "perfect games" in a Freedomware Tournament during finals play? What team player has the highest percent win when on a team (ie, who do I want on my team)? In which tournament/fest did I get the most points? How many come from behind victories have I had? Who has the best average against that particular team when playing during November?

Does any standardized format exist for holding game tournament data (or something that could be considered a file format and is public, royalty free, etc)?

Also FWIW, I spent only a few hours trying to pen roughly the fields/stats we might want to collect for the fest. OTOH, I have some experience writing code to extract just this sort of stat from a file to generate query results. In one case, I wrote a query parser in javascript to manipulate an array holding compact raw data (softball stats) in order to generate a table of top results for your query. I did not eliminate all bugs or finish it (I had some ambitious queries as a goal), but I got some very interesting results. One of my goals in using javascript (which would not have been my first choice language otherwise) was to have a system you could be seen online AND preserved for offline use forever. Once the data was compacted into the few arrays (perl code generated these javascript arrays from season game data previously captured), the javascript would use it to initialize itself and then would use it as a (data) base to produce whatever query you fed it. Most websites don't provide this sort of capability because they want hits and to be able to keep track of sessions. Otherwise, it makes great sense for the end user to be able to have this in a tidy package, a webpage you simply save and can access any time without any more conversations with any server, and you would get the same data even if the server changed things or went off line. Of course, you would want to have the official version accessible online. People would be tempted to edit the saved page in order to produce impressive results to show to friends.

Offline
Joined: 2007-09-10
and by the way...

Something here might be useful elsewhere besides this upcoming gamefest. Don't interpret the comment as a persistent attempt to change the rules. It's like throwing a lot of things at the wall seeing what sticks. Nothing might, but then something might. And something that doesn't stick may end up being useful regardless (someone's trash is another's treasure).

libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote:One reason to
Jose wrote:

One reason to possibly have different point rules apply to different tournaments would be that some systems may thrive with many participants but be trashy if there are only a few participants. Allowing for the possibility of a different system may make it easier to include more games, the less popular ones, in the fest
(...)

You're right. The troubles in trying to fit the former universal points system to all games was what led to this alternative ratio-based system. Now that this one is dealing with the determination of the universal winner we can focus on game specific systems and if necessary make them better reflect each game.

The reason I said such systems would be used for qualifications only is because in the second stage, the tournament, it isn't really needed since the tournament is all about elimination until the winner remains. We could keep recording points even for tournament matches just to see who ends up winning per points and compare it to who ends up winning per elimination. I'd bet the result would be same or very close. Smiling

Also, each game slot can have its own tournament brackets. That's why we're discussing separate guidelines for each game we currently know can be included. This also means that there is no strain on other diverse games from any universal system cause each can be handled in a slightly different way. Smiling

Jose wrote:

Think of an all-star game.

That little bit caught my attention. Maybe we could in the end arrange for some sort of an all-star game whose results would be outside of the main system, just for posterity and further fun. Who would we consider to be "stars" we'll see.. Perhaps we don't have to draw that one up right now. We can see who comes up on top based on both the points system and overall reputation among players and audience watching the demos. Eye

Jose wrote:

Well, if a small tournament has less games played, that would automatically be a way to scale downward the contributions from that tournament. Simultaneously, it would encourage participation since it offers a source of extra points. If I read the draft rules correctly, I think this is currently what we have to work with.

I think you're close. The current drafts don't yet entirely reflect the new idea of ratio based points which makes it irrelevant how many games a player has played and which games those were, as long as a certain minimal amount is met, which will be ensured in each game no matter how less popular it is. So someone who played only 10 games can in essence have the same chances to be the overall winner as someone who played 20 games. The one with only 10 games can win if, for example, he won 7 of these 10 games (70%) while the one with 20 games won only 12 (60%). It might not seem fair at first, but I think it is because each new game that a player plays carries just as much chance to contribute to his winning potential as it has to detribute.

So more games or less games played is neither an advantage nor a disadvantage. Another chance to win is at the same time equally another chance to lose.

Jose wrote:

I want to say something in favor of having numerous recognitions besides tournament winners and a fest winner. With this variation, you can recognize more things without sacrificing others.

I agree having additional recognitions / titles may be a good thing and we may yet have something like that, even if it would come out later in the process. You seem to suggest a way to determine these titles based on some custom points systems, but I don't currently envision an exact system that would really do this effectively. It's hard to measure things like "skill" just as much as it is hard to measure "humor". Smiling These are the things only humans can judge about.

So a way to have such special recognitions, if mathematical measurements (with some points system) aren't sufficient is to have players or audience watching demos actually nominate and vote people. That much I can see happening. We'll see.

Jose wrote:

In this vein, I propose that we record (a la screencast or through an in game mechanism) as many games as possible (and archive them).

That's the idea. Smiling

Jose wrote:

Some questions we might be curious about later on:

All of these questions can be answered from either a given points system or human voting or the combination of the two.

Jose wrote:

Does any standardized format exist for holding game tournament data (or something that could be considered a file format and is public, royalty free, etc)?

Not at the moment. We'll use the simplest existing tools possible. It's going to be on the website or in forums.

Offline
Joined: 2007-09-10
recognizing many types of good players

>> So someone who played only 10 games can in essence have the same chances to be the overall winner as someone who played 20 games.

OK. Do we want a person that competed and did well in 2 tournaments to beat out someone that did almost as well but competed in 5 tournaments?

Consider three players. The first player competes in three of the smaller tournaments going 19-3. The second player competes in the two larger tournaments going 23-5. The third player competes in almost all the tournaments (including team play) and goes 40-10. In this case, the players scoring fewer total points got the higher averages, but all three averages where very good. Do we want to give the 19-3 player the title of "Ultimate Player"? I would think that the 40-10 might deserve it. The main reason is that this player competed in a lot of tournaments and didn't just get lucky in one or two. Perhaps the player competed in games they weren't that good in and that dropped the average down. Should that player be penalized for taking the risks and playing all the games (and doing quite well despite everything)? Doesn't a title that represents the whole fest make more sense being won by someone that didn't just pick and choose the one or two games/modes they were really good at?

Things get more difficult to judge if the player playing in everything didn't do that well, eg, 27-23.

Conversely, we might find that someone is not that great but played in a lot of minor tournaments that were attended by non serious players or players that forfeited. In this way, they would have padded their average.

I think we should list the qualities we want for each/any title. Then we look at what point system might be suitable. We may still end up with only one or two point systems.

Now to go back to the point system that I saw in the rule book. It might be a good system for picking a fest winner ***IF*** we want the fest winner to be all-around and have competed in a lot of different tournaments. Sure, competing in a lot of tournaments doesn't guarantee you will win, but it helps. And it seems fair to me to have the more popular tournaments (with more competing players) contribute more to the final score of fest champ. The point system described in the rule book would achieve this IF, for example, we make sure that popular tournaments have more matches than less popular tournaments. [you may want to reread this paragraph]

I see how using averages to define the top player makes sense. I am not saying we can't improve upon the point system in the rulebook. To cut to the chase, let me suggest that we mix both systems. We do averages but require a large amount of "at bats". In this way we rely primarily on averages but can make sure that all-around success is factored in.

The "at bats" may be "total points" (total victories or total games played achieve roughly the same effect if we use the point system in the rulebook). We may define "at bats" as "compete in at least 3 games (or 30% or...) in 70% of the tournaments." A definition/requirement like this will reward those that have high averages (since that is being used as the deciding factor) but ensure that we filter out those with high(er) averages that did not participate very much throughout the entire fest. 3+ games in 70% of the tournaments may be fair, or it may be too stringent. Again, we should perhaps look at what qualities you want to reward in a "fest champ". There may even be a title that rewards all-around quality play and another completely separate title that looks almost purely at averages (with some easy to meet minimum "at bats"). And we can have other titles, too.

Here is an example of the 3+ in 70% minimum requirement rule in play (there is nothing magical about 3 or 70). The following player's average would qualify for consideration: Competed in season play and lost early in three tournaments. Played 4,6,and 8 total games in these tournaments. Also competed and went far in three other tournaments: 7, 12, and 13 total games a piece. Played only 1 game in a seventh tournament and skipped out in the remaining tournament [so I assume there were 8 tournaments that were used for qualifying the players.. we may not even use all tournaments to figure out the hypothetical 70%, eg, we may not count really small tournaments]. This player played 3 games or more in 6 of the 8 tournaments, ie, the player played 3+ games in at least 70% of the tournaments (in 75% of them to be exact). This player would have met the hypothetical "at bat" requirements for Ultimate Fest Champ.

If it was missed, a main point of this reply so far is that having multiple titles may be the best way to recognize good players without having to discriminate in some arbitrary fashion. We can still discriminate in some arbitrary fashion for a particular title (eg, for Ultimate Fest Player), but we would not pick one set of discrimination and toss all other rankings/point systems in the garbage. If this was major league baseball, consider that there is a winner for most "runs batted in" as well as most "hits" and highest "average". Maybe we want to designate the one with the highest average to be "THE" player of the year, but we can also recognize that *arguably* there is a better way to measure player of the year (RBI, hits, etc), so we give away a few titles AT THE SAME TIME AS we pick one to be the most important one.

We only have one fest (for now), so we should milk it. Fests are expensive. Coming up with titles is cheap. It is impractical to have a whole new fest every time we want to recognize a different set of qualities in a player (ie, a new title). Instead we have one fest and abstract all the various reasonable titles from it that we want. [To go back to the baseball analogy...] There is no need to replay an entire baseball season to come up with an RBI winner this time around. No. We have one season and abstract the various titles from that one season.

At least that is my opinion. It recognizes that there is more than one way to skin a cat, and that if we had to pick one best way, we might goof and not pick the best one. Imagine if we thought that the best baseball player was the one that waved the most to fans. This (arguable) error in judgement would not be that bad if we also recognize other titles (like home run champ). OTOH, it would spoil the fun if we ignored every other contribution made by other players in order to focus on just that one player that waved and bowed A LOT. So I think we should hedge against possible misjudgement by having runner up titles, ie, "less" important but still recognized titles.

Of course, if people have a very well defined view of what is the Ultimate Player and they don't want to bother with "arguable" alternative interpretations of what constitutes the best Ultimate Player, then forget about this. Other people may not share my views on multiple titles and may simply want one top player to receive all the recognition, prize money, etc.

Offline
Joined: 2007-09-10
Recording lots of data.. just in case

>> All of these questions can be answered from either a given points system or human voting or the combination of the two.

I don't follow. For example, "How many come from behind victories have I had?" is not something you could answer without knowing when the points were scored throughout the match. I am not saying that it would be easy to record the information necessary to answer this, but you would need more info than simply the total points scored or even how many points came from each match (even assuming the date/time is available for each match). If we record a movie of the match, then we would have enough raw data to answer this and many other questions that simple point totals could not answer. Providing a video of the match may be too much data but it certainly would be enough in most cases.

So my point was that we should try to save as much data as is reasonable since we can't anticipate all the data we may want in the future (and disk space does keep getting cheaper and cheaper so a vid may be a very good idea if possible).

Maybe this example will help (I'm assuming you didn't follow me.. if you did, this clarification may still be useful to others). If you record a good amount of film for a sports match (maybe from several angles), you may be able to return to the film in the future and get the answers to questions that were not asked or anticipated at the time of the match. For example, you would be able to answer, "what was the hair color of all the players?" There are many such questions that some researcher, movie writer/producer, social scientist, or fan may like to know in the future but which a person covering the match superficially (without camara or very good detailed record) as it was occuring would not have jotted down. Perhaps all that was recorded were the names of the teams, the dates, and the final scores. Thus, I am saying that we should record "extra" information because we can't anticipate what will be useful or desireable to know later on.

Where a screencast of some sort is not possible, we might want to adopt some sort of standard. It would depend on the game, but the idea would be to try and capture as much information about the matches as is reasonably doable.

libervisco's picture
Offline
Joined: 2006-05-04
Yes, you're right. I forgot

Yes, you're right. I forgot to add that to the ways in which the questions posed could be answered, but definitely recording as much as we can would be a good idea. All three games which are set to start immediately have demo recording capabilities so we can just tell all players to turn the recording feature on, just in case and also have a spectator (I could do that) also record.

libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote:Consider three
Jose wrote:

Consider three players. The first player competes in three of the smaller tournaments going 19-3. The second player competes in the two larger tournaments going 23-5. The third player competes in almost all the tournaments (including team play) and goes 40-10. In this case, the players scoring fewer total points got the higher averages, but all three averages where very good. Do we want to give the 19-3 player the title of "Ultimate Player"? I would think that the 40-10 might deserve it.

Hmm in theory it seemed fair, but common sense and gut tells me you're indeed right. Smiling

So what comes to mind as a way to resolve this is to also take into account the number of games played so as for the winner to be the one who played most games and has best ratio, not just one of the two.

Jose wrote:

I think we should list the qualities we want for each/any title. Then we look at what point system might be suitable. We may still end up with only one or two point systems.

Ah ok.. I am trying not to get confused here.. (I never anticipated this level of complexity when I first initiated the idea of a tournament to be honest). So you're talking about some additional point systems for various titles... How many point systems would we end up having then? One official for qualifications stage per each game slot + one for each title of each game slot.. Now we're already looking at... SIX of them! Hmm.. or I am already confused.

Jose wrote:

The point system described in the rule book would achieve this IF, for example, we make sure that popular tournaments have more matches than less popular tournaments.

They likely will.. considering that they may have more than bare minimum players to work with whereas less popular ones would only reach minimums..

Jose wrote:

The "at bats" may be "total points" (total victories or total games played achieve roughly the same effect if we use the point system in the rulebook). We may define "at bats" as "compete in at least 3 games (or 30% or...) in 70% of the tournaments."

I am officially confused. Now I am not sure what exactly do you mean with this "at bats" term. Initially I thought you meant backup players waiting "at bats" to replace the official players if needed...

Jose wrote:

Imagine if we thought that the best baseball player was the one that waved the most to fans.

Nothing even comparable to that has ever been conceived in this fest so far. How about measuring how the player actually played, not waved and bowed to fans... I know that in essence that's what you're saying too, but this example feels like sort of a red herring.. it was never the case and has nothing to do with this.

Jose wrote:

Other people may not share my views on multiple titles and may simply want one top player to receive all the recognition, prize money, etc.

That was not to happen should the pool of prizes be big enough. All along we wanted to recognize winners of specific game slots as well. Your proposal of having some extra titles was by me understood as adding some special titles for things that we CAN recognize without too much extra burden. You apparently laid out some system above which unfortunately failed to fully grasp, but what keeps ringing in my head continues to be *simple* ways like nominations and voting, not whole mathematical equations.. or what have you.. I am not very good at maths and I might not be a good organizer either (as I am realizing now). So I can't criticize you for trying your best, albeit in a way that barely reaches me and pretty much doesn't reach anyone else here (nobody else replies but me anyway).

In all honesty I never expected it to become this complex, almost "bureaucratic". If I did I would've thought twice before doing anything and perhaps rather moved on to some other worthy project.. That said, if this is what it really takes then I am afraid I wont be the one "leading" and organizing the 2008 fest. I can provide hosting, marketing, donations, whatever.. just not organizing it. Man, maybe you'd do a better job.

But we'll pull through. I'll strive to keep it as simple as possible, but address the most important issues brought forward. I also recognize that you greatly helped put some of these issues on the map and help find the best way to address it, even though I feel that you went even further.. where I might not afford to go.

Thank you.

Offline
Joined: 2007-09-10
I don't want to drag this

I don't want to drag this on if you can believe that. Let me explain what I meant by "at bats" as it seems you didn't understand it. When Major League Baseball (MBA.. the monopolistic professional baseball league in the US w/ Canadian participation) recognizes a season top performer in some category based on averages, they usually require that the player achieve some minimum number of appearances batting. It's to prevent players that sit out most of the season after having a good stretch of 10 or 20 games (out of over 150 for the full season) where they accumulated an unordinarily high average that no normal human could maintain for a full season of competitive play. I think you understand that. I used the term "at bats" to mean that some sort of minimum number of appearances should be required.

One obvious example that would apply in our situation is requiring players/teams to have a minimum number of matches to qualify for the Top Fest Player award **if** that award is otherwised based solely on winning percentage as I believe you were considering. As an example, if a player enters a small tournament and comes out without a loss (eg, 6-0-0), then that player already has the highest average possible. Would it be fair to give that player an all-around fest award vs other players that had records of 12-2-1, 23-3-1, etc? I don't think so. Likely if the 6-0-0 player had entered more tournaments to accumulate sufficient "at bats", s/he would have lost some more games. In fact, I have seen many times players start off on a tremendous winning streak only to have the law of averages catch up with them so as to end up around 50%. This is a normal scenario. In fact, the above examples (12-2-1, 23-3-1) may have started 10-0-0 only to lose late in the season or tournament (perhaps against each other).

Besides this point that a large number of samples is required to filter out lucky streaks (and to test a player under many scenarios and on bad days as well as on good days), there is the added fact that a "fest" award should probably require a player prove their skill in several different game/modes. One way to require this sort of "at bat" minimum number of appearances is to require that the fest award requires that a player accumulate at least 5 games in at least 3 tournaments.

If we ignore averages and go strictly by points, we end up with the problems you were discussing where a player with 23-20-4 might end up with more points than a player going 18-0-1 even if the first player if only slightly better than average while the other player is really skilled.

The best is to make sure that we have elements that reward both averages and stamina.

FWIW, let me revisit the example I gave that had 30% 70% and what not. It's no more complex in concept than what I mentioned here. In different words it means the following. Let's say that we decide there will be 6 tournaments and these will have about the same size (assume this to keep the example simple). I was suggesting that perhaps we should require a player to accumulate 3 matches or more in at least 4 tournaments. This would be an example of a minimum number of "at bats" that would allow a player to have his winning percentage qualify for consideration of Fest Ultimate Player. Once you meet this requirement, your ranking is based on the actual winning percentage as you were suggesting. This system would filter out the person that participated in only one tournament going 6-0-0. In fact, it would also filter out a good player that went 20-0-0 but in only 3 tournaments. We may want this or we may not. The other thing we could do is to have several requirements any of which can be met. For example, if we added the optional/alternative requirement that 18 matches be accumulated total, then the 6-0-0 would still fail each of the two tests, but the 20-0-0 would now qualify despite having failed the first test.

I hope I make sense. This really is not too complicated as the prior paragraph showed. Once we know how the fest will look, we simply pick the numbers that must be met to qualify. Once qualified, it would solely be the winning percentage that would determine the winner, the runner-up, etc.

I was looking for a bit more complexity than you, I think, because I have experience in this sort of area and it seemed natural to me. I know I haven't been nearly as clear as I could be (and I'll clarify more if you want).

I'm not offended that these suggestions would seem overkill. Whoever puts the tournament together is responsible for setting goals, priorities, etc. It's even possible to have a great tournament without a rulebook. The organizers essentially define the scope based on the time and energy they are willing to put in.. which is more than what those not organizing are putting in so there is no shame at all. In this particular case, a lot more thought and work was put in that I originally expected. Why would I, doing little but nagging (in a sense), want to look a gift horse in the mouth?

Have to run now, but I should have more time later on.

libervisco's picture
Offline
Joined: 2006-05-04
Thank you for a

Thank you for a clarification and understanding. I believe I now better understand what you were trying to say and I agree with the at-bats idea, that is, having a minimum amount of matches played to qualify for the ratio-based ranking.

That said, there are still some issues here. If we require that this minimum amount of matches be accumulated through at least 3 tournaments (game slots) this might in fact end up disqualifying the majority of all who are playing in this game fest, because people are mostly signing up for one or two games at most. Tremulous, for example, is to be the biggest part of the game fest, but almost all players involved with it will play no other game but Tremulous, meaning that they would fail to qualify if we require 3 game slots minimum.

So logically, that kind of requirement wouldn't get us anywhere, but on the other hand, not having this requirement means that the ultimate winner of the fest would be one playing only one or two of the game slots.. I'm not sure if this is necessarily a problem or not though.. With enough matches played and many wins, no matter in which game, the given player may naturally be recognized as The One who gave most effort and was most brilliant in what he gamed..

There are basically two ways we can go about this, neither of which is a perfect solution:

-- Don't have an ultimate winner at all, but rather only game slot winners (problem here is that we may not have enough prizes to properly and equally reward all of them. Some prizes, like a gamers keyboard or a computer can't be shared.. so to winner of which game slot should it be given? That's a bit tricky.)

-- Come up with a high minimum at-bats number for each game slot based on the number of games expected to be played in that game slot. So if the overall number of games expected in Nexuiz Sole play is 50, the minimum could be as high as 25 or even 30. This at least makes sure that those qualified for the universal prize, even if all they played was one game, really is a very active and also a very good player.

In a way, the universal winner of the fest will be a sort of an universal win for one game over others.. Will Tremulous be the one churning out the best player based on this system or will it be OpenArena, Nexuiz or some other.. It may be interesting to see.. In a way, it would be like asking "what does it take to be the best of the best in this particular game" and then the game which requires the player the most effort to be the best is also the one that gets proclaimed the best player of the game fest.

Jose wrote:

I was looking for a bit more complexity than you, I think, because I have experience in this sort of area and it seemed natural to me.

That does seem to be the case, yes. Smiling I'm glad you understand.

Offline
Joined: 2007-09-10
A major difference between

A major difference between our approaches is that I am thinking about design without (nearly as) much care at this moment for when or who will implement things. You are very focused on executing, on getting this specific project done. This is yet another reason why I couldn't possibly get insulted and why I actually feel a little guilty. I'm sort of playing in the clouds or discussing things as if the deadlines were months away and numerous people might get involved to hash things out.

I assumed the tournaments would all be similar in size to each other just to help get the point across. There was some wishful thinking on my part (think pie in the sky) in using that example and not assuming right off that only a few tournaments would be popular. OK. Enough of that. You are probably very correct that the tournaments will be lopsided if we even have very many beyond the main 2 or 3.

I agree with your analysis. If this tournament is limited as you say, then what you propose is an interesting solution that is probably among the more workable ones. This is one reason why I wanted the goals and requirements for the titles/awards to be defined (maybe they had been and I just didn't know). ..so I do think it is reasonable to define the Fest Winner as you have defined it.

[I am/was hoping that people would want to take the plunge and try out many games. I was looking for incentives to "encourage" people to spread out. It's one way to give those that do want to play the other games a decent opportunity for a large field. My preference would be to have the Fest Winner be one that participates in many distinct tournaments. But we have to plan as if people will not all of a sudden show interest not formerly expressed.]

One major benefit of the system you propose is that even if many people play in many tournaments, it still provides an adequate definition of a fest champ; ie, if things go "wrong" and we have many vibrant tournaments, we still end up with a Fest Champ that is reasonable. However, we haven't solved the issue completely. Let's consider this plausible situation. We hold 4 tournaments. The champs of those tournaments are as follows: 45-8-3; 22-3-3; 12-0-1; 6-0-0. So who is the Ultimate Player? On the one hand it is unfair to give equal weight to a tournament with few players (for the purposes of determining the overall champ). We can add more games to the small tournament by doing many rounds but we will still have only a few players in that tournament. On the other hand, does this mean that we will likely only be considering the top 1 or 2 game/modes for calculating the Fest Winner?

If things come to this, I prefer we try to have only a few (more than one, please) tournaments that will be used to define the champ. We disregard the less popular game/modes to focus on where the action is occurring. We make that clear. If it is a small number of tournaments that get the attention and the games are similar to each other, we may get most people that participate in one to perhaps try the others. Regardless, we should be able to have at least 2 or 3 "major" tournaments.

In this case, we host titles for the smaller tournaments, but people will understand that these other tournaments will only have their one small title awarded. These less popular games would be something like exhibition tournaments with the hope that next time around more would participate (kind of how some contests might be introduced quasi-officially into the Olympics one year to help give them exposure for future Olympics).

Since we are doing things piece-meal, we can set deadlines to close registration for each tournament one at a time (that's the current plan I think). At the time registration gets closed, we decide if that tournament will be valid for determining Fest Champ. At that moment (or when we come up with the schedule) we decide exactly what the requirements need to be (eg, play in 30 games). We may even add handicaps. For example, a less popular tournament may result in a "handicap" of 90% meaning that a record of 8-2-0 (80%) would qualify as 72% (ie, 80% times 90%; .8 * .9 = .72).

As far as prizes, I don't know what the plans are except what I gather indirectly from your comments. I tend to like spreading the loot so that people don't think about poisoning each others drinks when the competitor is distracted; however, expecting we will limit prizes to a small number, I think we should make sure that not all eggs are in one basket. A high prize makes things competitive for those that last towards the end and it may result in fun for spectators, but we don't want to over-encourage cheating or to have people be bitter when they lose the tie-breaker to the tie-breaker by one quarter frag. Over competition can be unhealthy, and this may discourage people from participating in the future if they don't think they have much of a chance to get anything (except of course for the great memories).

Offline
Joined: 2007-09-10
I just realized I need to

I just realized I need to augment the description of the handicap scheme mentioned at the end. We can use the first tournament as the base and give it a handicap of 1 (100%). Thereafter, there are various things we can do to come up with the handicap value. I won't discuss this now. What I do want to point out is that if a future tournament ends up being more popular than the base one, that is not a problem. We can assign a "handicap" greater than 100%. Note that this means that it is possible to end up with a qualifying adjusted percent value greater than 1 (100%). This would be the case, for example, if you were to win all of your games (go perfect) in a tournament with a multiplier/handicap greater than 1.

libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote: This is yet
Jose wrote:

This is yet another reason why I couldn't possibly get insulted and why I actually feel a little guilty.

No need to feel guilty. You did help with your input to address some things which are probably better addressed than not.

Jose wrote:

I am/was hoping that people would want to take the plunge and try out many games. I was looking for incentives to "encourage" people to spread out.

Ideally we would have people play multiple games, and so we can try to encourage that. In fact I think even with a system that recognizes most people will play only one game there still remains an encouraging window of opportunity for those who will play multiple games because they'll be able to play more matches and hence have more chances of being qualified high. We can mention this in our guidelines.

However, as you said further on, we can't expect this. This is just a first game fest and the most efficient way to attract people is to attract specific game communities to participate, which means these communities will likely want to participate in a tournament specific only to that game the community is about.

Jose wrote:

Since we are doing things piece-meal, we can set deadlines to close registration for each tournament one at a time (that's the current plan I think).

Well, the current plan is for games that have enough players to start roughly at the same time, so having the same deadline, and the other games with insufficient amount of players to start once (IF) enough players join them.

Jose wrote:

On the one hand it is unfair to give equal weight to a tournament with few players

There is a certain reasonable minimum of players needed to have a tournament of a certain game. Now I suppose what you mean is few in comparison to tournaments which have much more than just a minimum..

Jose wrote:

On the other hand, does this mean that we will likely only be considering the top 1 or 2 game/modes for calculating the Fest Winner?

Based on the last idea I proposed (coming up with a high minimum at-bats number for each game slot based on the number of games expected to be played in that game slot), the minimum number of games to play is reciprocal to the overall number of games in one game slot / tournament. I think a problem of few players only comes up when we deliberately induce more games than these few players would play.

This reveals one way to resolve this problem is simply not to induce more games. Just as the minimum number of games needed to be played for a player to qualify for the Fest Winner is reciprocal to the number of matches overall, it too should be reciprocal to the number of players participating. We mess with that reciprocity and create a problem you describe.

But anyway, as I went into exploring this I thought maybe we're just complicating things too much... The thing is, of all the tournaments there will be one player/team that wins. Why have any other qualifying mechanism for the Universal Winner other than that really? Heh looks like I know how to make things more complex than they have to be. Sticking out tongue

Let's just have all the winners qualify and that's it. They won their tournaments after all? Logic dictates they are the ones who are best eligible to proceed, right?

So basically it would go like this.

Let's say we have these 5 tournaments:

- Nexuiz Solo
- Nexuiz Team
- OpenArena Solo
- OpenArena Team
- Tremulous

Obviously there will be 2 persons as winners of the solo tournaments (as there are only two solo tournaments). The other winners are participants of the winning team. We just count all of them up as eligible for Fest Winner and to make this determination do one of the following:

- Simply apply our ratio system and have the math do it (among all the tournament winners, the one with biggest ratio and most games played wins)

- Do a final cross-games tournament putting them all through each of the five tournaments and see who comes out on top.

The latter adds more complexity and would probably not work because many players who came to play, say Tremulous, probably wont be in it for OpenArena after they already won Tremulous.. I just don't think they'll be in for it.

So we're left with our ratio system. And the thing is, since now we count only tournament winners, the reciprocity and fairness is already taken care off. Those who won less popular tournaments will obviously have less games behind them, impairing their ratio no matter how good it is, cause it was basically easier for them to win against less opponents. Those from more popular tournaments will have needed to win against many more opponents to win and that reflects in the number of games played. The only was one from a smaller tournament to end up a Fest Winner is for it to have a perfect ratio of 100% wins and the winner of a more popular tournament had a lousier score and just barely won. Smiling

I think this is really the best way to go.

libervisco's picture
Offline
Joined: 2006-05-04
Just one more thing. You

Just one more thing. You might ask about all these players in winning teams since by the current ratio system all team-wins count as individual team-member wins making each of team-member's ratios exactly the same.

So the way we address is to pick the best player from each team based on game specific data such as number of frags for Nexuiz and OpenArena and best voted team player in Tremulous (which is a much more tight team play). Then these best team players take the whole team's ratio as their own.

Ultimately we are left with 5 people (if we have 5 tournaments) eligible to win, each with a ratio between the number of games played and a number of games won, of which those with best ratio and most games played wins the whole game fest. Smiling

Offline
Joined: 2007-09-10
Handicapping or adjusting the top scores from tournament winners

The following may be considered too complex, but please don't stop reading. You may not think it is too complex.

Before getting to any math, I'll quickly lay out the reason why not adjusting scores is unfair. If we want a fest winner to be determined as libervisco suggested here http://www.nuxified.org/topic/alternative_points_system#comment-10746 , we need to compare scores from the winners of the different tournaments. The simplest thing to do is to have the fest winner be the tournament winner that got the best winning percentage along the way to winning their particular tournament. Alternatively, we can adjust the scores of the tournament winners in an attempt to make things fair for those that play a lot of games compared to those playing few games. A second consideration is that a winner that competed against a small number of opponents is more likely to find less real competition than a winner that competed against many opponents.

Players don't need to worry about the details below. All they care about after a tournament has started is that if it is a big tournament and they win it, their score will be adjusted positively compared to someone winning a small tournament.

[An alternative to adjusting scores using a multiplier known at the time the tournament starts is to do some sort of voting. Voting is useful because it may capture many elements that no formula can capture and thus be more fair. We may for example want to have a period of a week towards the end of the fest where players can watch films of the various winners on their own to cast a vote for the fest winner.]

OK, so here we go. [Warning, this may or may not hurt]

The value to be used to adjust the winning percentages (I don't know how forfeits or draws or dropouts will factor in) will be composed of two parts that are multiplied by each other (so each component is a multiplier adjustment itself).

The first part takes into account the number of participants in the tournament. The theory being that if there are too few participants, the winning percentages of the players will not be very wholesome.

The second component considers the number of matches played. The theory being that too few matches can lead to anomalies and streaks that are not generally sustainable in a longer contest.

If tournament has
-- 2-6 people/teams, multiply by 0.8
-- 7-15, keep the same
-- 16 or more, multiply by 1.1

If tournament has
-- less than 10 matches, multiply by 0.8
-- 11-20, keep the same
-- 21 or more, multiply by 1.1

I assume tournament means season+tournament.

The minute we know the number of participants and the number of games in the tournament under consideration, we will know the net multiplier.

Example 1: If the winner of a tournament of 5 individuals (or teams) plays 30 games, his/her winning percentage will be multiplied by 0.88 (0.8 * 1.1), so if the winner went 28-2, earning a .9333 (28 / 30), then the final adjusted score would be 0.821 (.9333 * 0.88).

Example 2: If the winner of a tournament of 20 individuals (or teams) plays 50 games, his/her winning percentage will be multiplied by 1.21 (1.1 * 1.1), so if the winner went 40-10, earning a .8000 (40 / 50), then the final adjusted score would be 0.968 (.8000 * 1.21).

We see that the second player would win over the first for the fest award despite having a lower "raw" winning percentage. This isn't too unfair since it's very possible that the first player had 4 other competitors that weren't very strong while the second player likely faced some sort of real competition.

Anyway, this system is definitely not a completely fair system by any means (eg, the type of game play might lead to many weak players ganging up on the strong player, leading to losses for the strong player), but it is something to consider. Even if we opt for voting, at least we have something objective as a reference point. Voting has many positives (it could account for the "ganging up" example just mentioned), but it can also be abused. Personalities and not just quality game play might play a role in deciding the outcome. Also, votes may be cast for the winner of the tournament you liked the most. This could lead to the winner of the most popular tournament winning the whole thing mostly because s/he was able to get the most votes, having won the *most popular* tournament.

And another way to avoid adjustments is simply to only consider tournaments that are large and to make sure that they all have sufficient games. The equivalent adjustment in this case would involve multiplying these scores all by the same 1.21 (so effectively nothing gets adjusted) while completely dropping the scores from smaller tournaments.

Take the above FWIW...

..from http://www.selfgrowth.com/articles/knowlton2.html
>> If we take one, two, three or ten setbacks as evidence of failure, how would we progress? Thomas Edison is known to have tried to create the electric light bulb more than 1000 different ways. One of his financial backers is said to have asked, "Tom, why don’t you quit? Can’t you see this idea of yours is a failure?" Edison’s response was, "Every time it did not work I got feedback on how to make it better. I have now eliminated 1000 ways it does not work and I get closer and closer to success." Every failure moved him toward to his goal.

Consider this attempt at using adjusters/handicaps as yet another attempt to build the game fest scoring system. Can't stop shooting at the target.

Offline
Joined: 2007-09-10
I had to laugh reading your

I had to laugh reading your post because it seems we were both trying to find a specific simple "formula" and came up with a potential solution or other at about the same time. I was almost worried sick to death that I was pressing the subject too much by continuing to post. Sickness to death is healthy .. in limited doses.

>> In fact I think even with a system that recognizes most people will play only one game there still remains an encouraging window of opportunity for those who will play multiple games because they'll be able to play more matches and hence have more chances of being qualified high. We can mention this in our guidelines.

Definitely. Once we have fixed rules, we can go through them gathering bits and pieces that highlight and support the various goals of the fest.

>> - Simply apply our ratio system and have the math do it (among all the tournament winners, the one with biggest ratio and most games played wins)

I missed something. If the player with the biggest ratio does not have the most games played, who wins? Would you mind recapping what you are describing through one or more examples (using sample numbers and computations). I think such an example would also clarify other questions I have.

Offline
Joined: 2007-09-10
>> You might ask about all

>> You might ask about all these players in winning teams since by the current ratio system all team-wins count as individual team-member wins making each of team-member's ratios exactly the same...

Yes, this system of putting the winners of each tournament against each other based solely on the results (winning perc?) of the particular tournament (not ideal but practical) makes some things fall out rather easily.. like this method of converting a team to a player.

The top player of a team can be decided on by a stat or even by voting (or a combination of both). Voting would allow members that played crucial roles that were partially defensive to win out over a player whose role was mostly offensive and easier to "stat". (Subjective) teammates voting might work the best here. Hard stats might be used by each of the candidates in order to bolster their case to their teammates.

libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote: I missed
Jose wrote:

I missed something. If the player with the biggest ratio does not have the most games played, who wins?

Good question. It may seem a bit weird, but what it essentially amounts to is a ratio between wins and games played and a ratio between that ratio and the games played. Smiling

So..

Joe Fantastic stats:

- Played: 18 matches
- Won: 12

His ratio: 12/18 = 0.666666667 = 66.666666667% (wow this one was evil Sticking out tongue )

And now since we decided mere ratio is not going to be enough because someone who played only say 8 games and won whole 7 of them would have a ratio of 0.875 and win over the one who had 18 games, which didn't seem fair.

So we basically do a second ratio calculation, this time between the ratio itself and once again the number of matches played: 18 / 0.666666667 = 26.999999987

Now take the result of the one who played only 8 games yet got a higher ratio. He is still defeated because 9 / 0.875 = 10.285714286 which is so much less than 26. In other words, just a percentage of games won from games played is not enough to win if the number of games played is too low - exactly what we wanted to achieve. So who wins? The one who played most games and won most games, counting both at the same time. Smiling

Offline
Joined: 2007-09-10
There is a mismatch between

There is a mismatch between that formula and the results you (we) are after.

Let's say you win 1 game out of 100. Then your final score would be 10,000 -- WINNER! ..by a mile! ..unfortunately.

I have more thoughts on what I think you are trying to achive, but I won't comment until you adjust the formula. It's easier to analyze something tangible, and I don't want to put words into your mouth.

***

As for what I mentioned in http://www.nuxified.org/topic/alternative_points_system#comment-10769 , the gist is simply to do a penalty but have it be a minor penalty and have it depend on two items only (number of games played and number of competitors in tournament).

That result works out differently than what you mentioned here in that it only adjusts scores in a "quantum" step (and there are few steps). Anyone that falls into any of the 3 divisions is grouped together. If you go 300 out of 1000 you will get no bonus over someone going 30 out of 100. You each get the same bonus as anyone that played in more than 20 games.

The reason I only did 3 groups was to keep it as simple as possible (TM) and because I anticipate that the difference in number of matches played between the participants in different tournaments would not be too great (ie, it wouldn't be 100 vs 1000 as in the crazy example above).

The details can easily be changed (eg, add more groups or adjust the range values), but do you like the overall structure? I can try and describe it again if you want, but try and guide me by asking specific questions if possible.. if you are interested [No need to feign interest for my sake. Let's just get this over and done with.]

Another important thing I considered was voting. I mentioned it in two places I think. Basically voting can simplify things, and it can account for things that even a comlex formula might miss or not be able to measure. On the other hand, we would need to consider a few things if we want to avoid voting becoming a popularity competition instead of based on gamesmanship.

For reference, here is where I specifically mentioned voting:

>> [An alternative to adjusting scores using a multiplier known at the time the tournament starts is to do some sort of voting. Voting is useful because it may capture many elements that no formula can capture and thus be more fair. We may for example want to have a period of a week towards the end of the fest where players can watch films of the various winners on their own to cast a vote for the fest winner.]

>> Anyway, this system is definitely not a completely fair system by any means (eg, the type of game play might lead to many weak players ganging up on the strong player, leading to losses for the strong player), but it is something to consider. Even if we opt for voting, at least we have something objective as a reference point. Voting has many positives (it could account for the "ganging up" example just mentioned), but it can also be abused. Personalities and not just quality game play might play a role in deciding the outcome. Also, votes may be cast for the winner of the tournament you liked the most. This could lead to the winner of the most popular tournament winning the whole thing mostly because s/he was able to get the most votes, having won the *most popular* tournament.

In this last part I am saying that if you win the tournament that the most number of people played, then a vote might very well make you the ultimate fest winner because it's possible that the most number of people sympathize with or appreciate the winner of the tournament in which they themselves participated. They may not even be interested in looking over the other tournament winners.

Offline
Joined: 2007-09-10
More readable explanation with quasi-realistic example.

That other post is long. What the formula boils down to is this:

-- A tournament winner's winning percentage (ignoring ties and forfeits, for now anyway) is adjusted in order to determine the fest winner. The one with the highest adjusted score is crowned fest ultimate player.
-- A winning percentage of X is adjusted to mnX [ie, m times n times X]
-- m is determined using the following chart:
0.8 : if the tournament has 2-6 participants
1.0 : if the tournament has 7-15 participants
1.1 : if the tournament has 15+ participants
-- n is determined using the following chart:
0.8 : if the tournament has 2-10 matches
1.0 : if the tournament has 11-20 matches
1.1 : if the tournament has 21+ matches

An example:

Player 1:
won with a record of 13-4
in a tournament with 10 players
Winning percentage is 13 / 17 = 0.765
m = 1.0
n = 1.0

The adjusted score is 0.765 * 1.0 * 1.0
= ** 0.765 **

Player 2:
23-10
22 players
X = 23 / 33 = 0.697
m = 1.1
n = 1.1

Adj score: 0.697 * 1.1 * 1.1
= 0.843

Player 3:
9-1
6 players
X = 9 / 10 = 0.900
m = 0.8
n = 0.8

Adj score: 0.900 * 0.8 * 0.8
= 0.576

Player 4:
10-3
6 players
X = 10 / 13 = 0.769
m = 0.8
n = 1.0

Adj score: 0.769 * 0.8 * 1.0
= 0.615

Player 5:
25-12
28 players
X = 25 / 37 = 0.676
m = 1.1
n = 1.1
Adj score: 0.676 * 1.1 * 1.1
= 0.818

So the ultimate winner is Player 2 with 0.843 adjusted score. The rankings would be

P2 0.843 23-10 0.697 22 players
P5 0.818 25-12 0.676 28 players
P1 0.765 13-04 0.765 10 players
P4 0.615 10-03 0.769 06 players
P3 0.576 09-01 0.900 06 players

Does this make sense?
Does this "formula" sound reasonable (at least based on the example here)?

We may want to specify what we want the borderline cases to look like and from that deduce the parameters for the formula. For example, as a completely hypothetical question: do we want a player going 10-0 in a 6 player tournament to finish behind or ahead of a player going 13-5 in a 18 player tournament?

[Note that 10-0 w/6 pl finishes just *behind* 13-11 w/21 pl using the formula described; you may not like this boundary case; I think it's slightly on the bad side but not too much; maybe I would raise the 0.8 above, or better have an extra cluster near the bottom because we *do* want to penalize a record of 7-0 heavily (too short of "at bats" in my opinion). Also, I think all tournaments are likely to play more than 10 games if we count season plus tournament. So most will probably fall into the middle or top clusters as concerns games played (the bottom serves almost as a filter.. that's why it was so harsh on the 10-0 example). Taking that into account, we are then likely to be really penalizing only for participating in a tournament with too few participants. This may not be a great idea, or we may want to adjust the values.. or it may be a great idea. In any case, that is what the above "formula" does.]

I tried to pick ranges that might be reasonable, but feel free to consider the boundary cases above and then tweak the parameters (the boundary markers). Also feel free to add more clusters beyond the 3 in the "formula" given above.

Or feel free to ignore this "fest winner" point computation system.

The choice is yours with Linux (R)!

libervisco's picture
Offline
Joined: 2006-05-04
Hah I hate maths! It's

Hah I hate maths! It's devious. Sticking out tongue

Anyway, thanks for pointing that miss out and suggesting a new possible solution.. It does seem reasonable, but there is that little issue with 10-0 player. Not sure yet really.. I don't want to jump to conclusion at this point. I'll have to go over it and see what comes out as the best way.. I think it may be a (probably slight) modification of the system you last proposed. It at least allows for less mathematic deviousness. Eye

Thanks

Offline
Joined: 2007-09-10
Nearing the end?

[There is a cookie at the end if you make it through the post]

It is devious because you can have the right idea and a little slip anywhere can lead to crazy results. I happen to like it because I have liked it for a long time, giving me an opportunity to be fairly good by now at some types of math. If you are fluent with the details, then it is useful and powerful in that it is predictive and allows for definite answers to what would otherwise always be intuition only. The attraction to this power is what overcomes the apparent abstract/blandness of it (to me at least).

I think you took too many ratios. No, actually the only real problem is that you didn't have a precise grip of what you had in mind. When you shot, you also probably didn't check your guess/intuition carefully since the example you picked seemed to work.

No prob (unless you are the final inspector before a shuttle launch).

We do all make mistakes. Triple checking is part of the deal if you want to do math so that you give the appearance of knowing it well. [I constantly do reality check tests and do find myself having made mistakes periodically if not routinely.. and this despite having had a lot of practice!]

***

I don't want to try and fill in pieces on your behalf because the best I think I can manage is to multiply the 18 and 12/18 (instead of doing division), but doing this just takes the total number of games out of the equation leaving you with only the number of wins, which was sort of what we had originally but which this "alternative point system" thread was trying to reconsider.

Good intentions don't always work out the first time. On the positive side, once you get a formula/algorithm down, a computer can take care of the rest for all time, super quick, with little effort on our part, and without errors essentially.

For future reference, I do like math (in general), so don't hesistate to call.

[It's never too late to improve math skills/abilities marginally.. these "margins" add up over time. Here is a tip. Don't limit yourself to a standard approach; just experiment and use pictures or fingers or whatever to gain insight. Eventually, you have to delve into memorization to be accurate, but it is **much** easier to memorize something that makes sense and fits well in the puzzle. If you forget details of a procedure, the rest of the puzzle provides clues or alternative means.]

***

As for the 10-0, we don't have to decide on the details until we have a better idea of how the tournaments might be stocked (so that we have realistic numbers to work with). One quick solution would be to adjust the range so that 2-8 games get the 0.8 penalty (instead of 2-10). This solves the problem for 10-0 and 9-0. Remember that this scenario also involved a tournament with few players. If you compete against a handful of others and win all 8 to 10 games, should you really win the fest award against someone that won a tournament having 22 people? I mean with more games and more competition you are expected to lose more. Also remember that we are matching the winners of the tournaments and are trying to keep the point system simple. In real life 13-11 would not have won a tournament. So this point system does appear to break down if you wanted to rank everyone, but our only real goal is to, not even rank but, find the ultimate fest champ among the individual tournament winners. In this case, expect to match the 10-0 with something like 20-6. This is the sort of case we need to really consider for tuning the values (not the 13-11 case where the assumptions don't apply and the simple model breaks down).

And do keep in mind an alternative interpretation, that although 20-6 is worse, it does represent that you were the best out of (eg) 22 players vs. being the best out of only (eg) 6.

Recap of the main ways that parameters can be tuned while keeping the overall idea:
-- change the cutoff values (eg, decrease the "tournament players" low cutoff value from 6 to 4; eg, decrease the "matches" low cutoff value from 10 to 8 )
-- change the adjusting multipliers (eg, make 0.8 into 0.9)
-- add more ranges/clusters beyond the current three; this increases the resolution [this is the best option for fairness but it adds a little more complexity (not much though)]
-- drop one of the adjusting multipliers (either the total matches multiplier (n) or the total participants multiplier (m))

***

One more thing to mention is that I am not ready to chuck the system that I read about in the wiki (x points for win, y points for loss, z points for tie, etc). Wouldn't both systems be usable?

In this discussion I have been looking to determine the fest champ, and I defaulted to using the "winning percentages" within the calculations. We can still determine the tournament brackets based on the other point system. We can even replace "winning" percentages with the existing more elaborate point system but divided by the total number of games or by some other quantity.

So we get (as example):
P1: 94 points in 25 games
P2: 48 points in 12 games

From this we get 94/25 and 48/12 instead of "winning" percentages like say 20/25 and 8/12 (using hypothetical, semi-realistic numbers hopefully).

[In baseball, slugging percentages would be a comparable measurement (as contrasted with batting averages). Despite the ugly looking formula http://en.wikipedia.org/wiki/Slugging_percentage , slugging percentage is just the total number of bases (eg, home runs count as 4 bases) divided by the official at bats; whereas, batting average is the total number of hits (eg, home runs count as 1 hit) divided by the official at bats.]

Then we finish off with the procedure described in this conversation but apply the m and n adjustments to the 94/25 and the 48/12 (and not to the 20/25 and 8/12). You will see that while the numbers are now larger, the relative rankings are mostly the same. As a bonus, we have a way to factor ties and forfeits (and anything else) into the calculations.

I think this allows the existing system (and whatever we liked about that.. eg, it takes into account ties and forfeits) to work with a system that is predominantly based on averages (ie, "normalized" by the number of games played) and using the adjustments (to help simulate "at bats"). We also use the system discussed where we take the winners of each tournament, using only the record compiled from that tournament, for determining the overall fest winner (this also helps simulate "at bats" as the winners were prequalified by the tournaments they won). And we have a simple way of converting team scores into a player score (the top team player gets the team score).

It seems that all the ingredients discussed can be harmonized.

Overall, I like it, maybe even like it a lot (considering all the constraints). I would only add that we may want to still incorporate voting.

Maybe have a part of the raw score contain a contribution of points from the results of some election. For example, we would add an additional voting component to the 94 and to the 48.

To this we can also add the gambling results.

So overall this seems complex but each of the several pieces of the computation has a reason for being (to add some element of fairness or fun) and is relatively simple for fulfilling its role.

Summarizing this point system:

-- The ultimate winner is the one with the top score where the score is determined by (a) taking their points from the qualifying tournament and dividing by the number of matches played in that tournament and (b) adjusting this result by one or more multipliers. Points are for wins, ties, losses, forfeits, gamblings, voting, others, etc. Matches are from season/qualifying play and from tournament play. Currently, adjustments are the two values that account for the size of the tournament based on the total number of participants in that tournament and the total number of matches played by the tournament winner.

This works for me (give or take a tweak).

Smiling

If after more discussion you like it, I can try and codify it, and you can go ahead and clean it up and fill in missing pieces.

Smiling .
Smiling

libervisco's picture
Offline
Joined: 2006-05-04
In short, I like it. So we

In short, I like it. Smiling

So we have a points system adjusted to each game tournament (component) and just take those points into account for the determination of the universal winner, with the addition of a few small "checks" to make sure it's as fair as possible, that being the ratio between points and games played offset by the appropriate multiplier.

Yep, sounds like you may have just about nailed it. Smiling

All variables are then taken into account. The game specific point systems count the wins, forfeits and draws. The additional checks cover the number of players (size of given tournament) and number of games actually played by the winner and balance it off..

I'd say on top of that, some sort of a voting for confirmation may just be a cherry on the pie, but probably not mandatory, just like human inspection merely making sure it panned out the way it was supposed to and expected.

Cheers

Offline
Joined: 2007-09-10
Just when you thought it was safe to go back in the water

Don't shoot me.

You may not want to read this post.

... [Don't say I didn't warn you] ...

After reading your comment, I realized that the assumption I made, that the points system for each tournament would all be the same, won't necessarily hold, so something should be changed to fix this problem. Let's say one tournament uses win=1 and loss=0 while another uses win=4, ties=3,..., then everything stated above would fail horribly.

We can salvage the situation perhaps this way: adjust scores with another adjustment value so that all the point systems are on equal footing. This value would be different for scores that use different systems but the same for scores that use the same system.

Example 1:
A - Point system w=4,t=3,l=2,ff=0 has adjustment value 0.5
B - Point system w=1,l=0 has adjustment value 2.

In action this means that a person using system "A" would likely end up with about 4 times a higher raw point score than if system "B" had been in place. Thus we level the score playing field by multiplying the scores with the high point totals, "A" scores, by 0.5, while multiplying "B" by 2. [There is a reason for this pair of numbers instead of something like 1 and 4. In this case, it wouldn't make a difference, but I am setting up for later examples.]

To keep things simple, I suggest we come up with these values by intuition. We may be able to (or want to) finalize this adjustment value at the same time as a tournament's point system is finalized.

A second example: if we have one tournament were the team points are the sum of the frags of all team members, and the average number of total team frags per match appears to be something like 100, then we might want to multiply this by 0.01 to level things out with the w=1.l=0 system (which gets multiplied by 2).

Example 2:
A - In consideration for fest ultimate player, Player X gets his team points because Player X was the best player in the team that won a particular tournament. The team had 30 matches, winning 24 and losing 6 (this was a 1-on-1 team tournament). The average per match point total for all teams was about 100. This team ended up with 5000 points.
B - Player Y won a tournament with 24 points (one point for each of the 24 wins and no points for the 6 losses).

It seems these two probably should get similar rankings in the chase for the fest ultimate player award since they basically went 24-6. Yet one of these will clearly have a much higher score if we don't cut it down to size.

A good adjustment value for "A" is probably 0.01 and for "B" 2.

So the first player gets the 5000 adjusted downward to 50 while the second player has the 24 adjusted upward to 48.

Now, where did I get these numbers from? The idea was to find a scaling value to give 1, and we considered what the average points might be for the tournament. We don't look at who scored the most. We look at the averages as this represents the strength of the tournament point system better (in general) than any individual participant point total would.

In the case where we use the same point system, this basically boils down to dividing by the number of matches. The numbers would be high in this case if we used w=4... vs using w=1, but so long as we used one and only one of these two systems, we could even things out by dividing by the number of matches.

Now, we want to deal with raw point totals that are based on different systems. The proposal is to scale based on the average points scorable for such a point system.

This isn't a perfect method by any means [in statistics, I think something called z score (or t score) fulfills this role in a better way, but I think that is asking for too much since these values are more complex than a single simple average]. One problem is *anticipating* the average when you are talking about things like frags.

Anyway, are there any comments? I will have to rewrite this in a shorter form later on, but is the need for a solution to the problem of having different point systems something that is recognized, and does solving the problem using one more adjustment multiplier, itself based on averages, seem relatively simple enough (to solve the problem)?

PS#1: The case of w=1 l=0 has an average point of 0.5. This means that on average, you get half a point. This is because on average (if you aren't too good or too bad) you will win one game out of every two so the number of points (wins) per match is 0.5. We adjust this by multiplying by 2 as I did in Example 1 (ie, divide by 0.5) ... On the other hand, the average for the w=4,t=3,l=2,ff=0 case is not 2 (which if it was, would be adjusted by multiplying by 0.5 as I did in Example 1). I pretended it was so that the example would look quasi reasonable without appearing too intimidating. The actual average would depend on how likely participants would forfeit. It may be around 2 but would be higher if forfeits are rare or would be lower if they are common. Remember, this 2 is the average points per match, and it would be adjusted by multiplying by 0.5 since I am using the value 1 as the target for normalizing/scaling, ie, divide by the average points per match to get the adjustment, ie, get the reciprocal of the average points per match and this is the adjustment/scale/normalizing value.

PS#2: I likely won't be able to reply for a couple days (hopefully not more than that.. maybe less).

libervisco's picture
Offline
Joined: 2006-05-04
I think what this calls for

I think what this calls for is to simply standardize the points systems with respect to the universal winner determination. This kind of closes a full circle to the old universal points system where for any tournament the same values applied for wins, losses etc. But there is an exception nevertheless; we are looking at it inward out rather than outward in: we don't force the specific games to the universal system, but rather just try to balance things out.

To get to the point, we try to give the same amount of points in each game for wins, losses, forfeits and this way remove the discrepancy you described without adding another layer of multipliers.

This is still not the old points system though because we still allow for a different set of rules about exact situations which count as wins, losses, forfeits etc.. It's just standardazing the numbers.

Anyway, I'll have to look over and think thro