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

Organizing Armagetron Advanced tournament

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

While we don't have many players signed up for Armagetron Advanced yet, we have been getting some new sign ups lately from the Armagetron community and it is possible we will have more. At least 10 players is necessary. I know this is a rather small number, but I believe even with 10 competitors we can make a small tournament. Smiling

So.. this thread is dedicated to discussing the rules and guidelines by which it is to be played and other organizational details related to Armagetron Advanced tournament.

I will say of all the games here I have the least personal experience with this one, so I will greatly rely on the community and your feedback in making the final decisions.

So, these are the available modes as mentioned by epsy in the AA forums:

Last man standing - "Everyone is put into an arena, tries to kill each other...none respawns and once there's only 1 player remaining, the server starts a new round."

Fortress - "2 teams are trying to conquer an enemy zone by staying in it."

Capture The Flag - "As (in) other games, except flag and base are on different positions, and you should go back into your base to respawn teammate."

Sumo - Like fortress, but not in teams

Thanks epsy for posting it there! Smiling

So that's where we start, which mode is preferred?

Reading on the AA forums thread, sumo so far seems to be getting most votes because it is the simplest to organize and also newbie-friendly in terms of coordination. From what I understand, sumo is basically a "Free For All" style game, like classic DM in FPS games. What counts is how many times any of the enemy players are cornered and destroyed. Correct me if I'm wrong of course.

Thank you

Offline
Joined: 2007-11-08
In sumo, the idea is to

In sumo, the idea is to stay inside your zone while pushing the other player/s out of his/their zone/s.

libervisco's picture
Offline
Joined: 2006-05-04
I see, thanks for the

I see, thanks for the clarification. Smiling

Offline
Joined: 2007-11-09
Well, the gmaetype which i

Well, the gmaetype which i see most servers do use is last man standing...it's the in the default config files.
If you heard so munch about sumo, it's probably because most of the posters on the AA forums are Fortress/Sumo players(me included)...well, it's almost like AA's players are [cough]divided[/cough] between fortress-sumo and LMS(i'm not even mentioning CTF)...splitting the AA tournament into 2(Sumo vs LMS) or defining the played gametypes would help you getting more players for arma i think

what i've also seen here is a Clan/Team signup thread...the main Team game in armagetron is Fortress, and we already have some Fortress tournaments...maybe merge Durka's AFL into this Festival ?

Also, i'm asking...what about CTF Teams ? I roughly heard of that « CTF Ladle » but i heard as well there was only 2 editions...oh...and durka was managing it as well, wasn't he ? Laughing out loud

Offline
Joined: 2007-11-09
i think sumo would be

i think sumo
would be easier to
pick up an play
than fort for people
last man standing even easier
but you'd get campers
so sumo has my vote

free-zombie's picture
Offline
Joined: 2006-03-08
I've played armagetron for

I've played armagetron for a while, and I think we should definately have some form of LMS, it being the classic mode, and the default. It's also the most straightforward to understand, I think.

I think fortress is strongest in the community that beta-tested AA 0.2.8 --- very active players, active on the forums, we (I'm a lot less active now than I was then) got a brilliant team-play game which was new and exciting. If we want team play, we should go for fortress.

Sumo developed from Fortress; it's a good game, and I have no problem with including it, but not alone.

CTF.... VERY recent development, not very common/popular (I think). We should probably ignore it.

libervisco's picture
Offline
Joined: 2006-05-04
Judging from what you

Judging from what you posted it seems logical to have two tournaments. One LMS for sole players and another, if we get enough players for that, team play on fortress mode.

So basically, we go for LMS as the default tournament and if there are enough players expand to another tournament with fortress.

About Durka's AFL, I'd like more information about that to know whether it'd be a good idea to merge under the game fest umbrella or not. I think, basically, it shouldn't be a problem if everyone agrees, to consider it as part of this fest and have it just proceed as it normally would, but as part of the gamefest it would get further recognition and final winner would also be eligible for the universal gamefest prize based on his or her performance.

Offline
Joined: 2007-11-08
AFL info
libervisco wrote:

About Durka's AFL, I'd like more information about that...

click

Offline
Joined: 2007-11-11
Hi there, i play AA for

Hi there, i play AA for over 3-4 years now, i can call myself one of the veteran players of the game.
To do a good tournement, my suggestion is doing a Last man standing tournement in an arena, not SUMO. To clarify,
sumo is a game type in which you are in a circle, and you lose the game if your out of that circle for a while or if you are 'coredumped'. This is not the way for making a global tournement since many players are not used to this type of game. besides, sumo is not 'noob friendly'... how to explain it... in sumo and fortress you dont got normal 'turns' (you need to wait like 1 second between each turn)

My suggestion is putting every single player in 1 arena (i think that my own clanserver, the ~|DS|~DarkSyndicate arena, can become available if i ask for it) then do 'best of 3' 1vs1 matches until 10 points. Make it a ladder tournement.

So basicly, the so called '(high-)rubber game' (what i explained above) is the 'free for all' gametype you asked about, or what you want to call it (the DM mode in a FPS)

feel free to ask any questions, or visit our clansite at dsclan.dl.am

-Q

libervisco's picture
Offline
Joined: 2006-05-04
High rubber (AKA healthy I

High rubber (AKA healthy I believe) LMS tournament is the way to go. Just need to iron out the organizational details..

Qbrain wrote:

My suggestion is putting every single player in 1 arena (i think that my own clanserver, the ~|DS|~DarkSyndicate arena, can become available if i ask for it) then do 'best of 3' 1vs1 matches until 10 points. Make it a ladder tournement

I suppose that could be a good alternative to what was proposed here, but as mentioned in PM I'd welcome some clarifications, especially about what these "10 points" represent exactly, in-game or our standard tournament points?

Hmm maybe you mean just these best 3 players playing until someone hits 10 points mark first? Smiling

Thank you

Offline
Joined: 2007-11-02
noob friendliness
Qbrain wrote:

besides, sumo is not 'noob friendly'... how to explain it... in sumo and fortress you dont got normal 'turns' (you need to wait like 1 second between each turn)

Sumo is more noob friendly than any game mode where using doublebinds gives you a significant advantage. Having two keys that do the same thing can't be considered intuitive. "Digging" (in lack of a better word) is also not intuitive which is why I consider "low rubber" setting more noob friendly than "high rubber" ones.

Not sure how much we should aim for noob friendliness anyway, since this is supposed to be a showcase for free games. I think we should use settings that make for spectacular gameplay. To my experience this is achieved with really low rubber and fairly high speed settings, like the ones on the Nexus9 server for example.

libervisco's picture
Offline
Joined: 2006-05-04
We have to start

We have to start somewhere.. Perhaps some compromises could be made to the spectacle by making it accessible enough for players to join and for the event to be organized. Then perhaps next year we can do something further.

Offline
Joined: 2007-09-10
High rubber is more

High rubber is more tolerant of mistakes and imprecise timing. This makes it noob friendly IMO.

Quick double turns I can live without if necessary, but people should know about it if it is used. I would think it would allow for more interesting strategy. OTOH, limiting turns to one per second does change the strategy which makes things interesting. Both options/modes should be included if possible.

Now. Should we go for noob friendly or not? I think we should try to incorporate both options. We are appealing to noobs. I think that is part of the point of the fest, but you can't leave behind those that make up the current game community nor those who have lots of experience and would like to showcase their stuff. If there is a competition, you'd expect those with skill to have the best chance.

High rubber does neutralize some of the skill the experienced players may have, but it doesn't neutralize all of it.

Anyway, I'd like to see variation in the fights. Maybe we can do one round-robbin round with one set of options and then another with a different set. For the tournament itself, we might flip a coin or something. Maybe the final two rounds (eg, if single elimination) can involve two matches, one with each type of server, and some way to pick the winner of the two sets of matches. Or if the games are short enough, just have double (or triple) matches across the board. This would test the most distinct number of skills.

I do like LMS, but I'd be willing to look at the other modes and may even like them to be included in the tournament. Does anyone have precise rules of the game for these other modes and/or have opinions (or point us to a server that is active so noobs can test it).

libervisco's picture
Offline
Joined: 2006-05-04
I've drafted up the

I've drafted up the Armagetron Advanced tournament rules.

Feel free to comment and make suggestions on details. We could still change some things if most people request so, but the basics can't really change anymore. It will be an LMS tournament with 3 matches in each group for qualifications and one set of 3 matches between everyone qualified in the final stage.

Based on Jose's suggestion above I've also added the possibility of changing the settings of one or two of the last three matches (the final tournament stage) to something more challenging, IF the players involved in such a match agree to do so.

Thank you

Offline
Joined: 2007-09-10
3 small but important rule changes proposed and argued

I would like to change the rules in a few ways. I do like them overall (consider this a vote in favor). I would not object to allowing for a little more variety with things like rubber (35) if participants really want the variation across matches or across qualifying rounds (I'll explain below what I mean by qualifying rounds).

The changes basically total three, and at the bottom of this post, I recap by providing a way to adjust the wording of the tournament rules to incorporate the three suggested changes. The wording is not intended to be great or final, just good enough to convey the message of this post.

First: We can mention that each match will be played with 10 intramatch victories (or some other number). It wasn't mentioned, but we might as well. The rest of what follows assumes a value of 10. Later I'll consider how to adjust things if this value is not 10. By victories I mean that the group will respawn 10 times, leading to 10 cases of one man standing. It is this series of game rounds which I assume constitutes a match. So I would like for the total number to be specified (at least a default value). Again, for most of what follows, I will assume that this value will be 10 for every match.

Second: I would make each group of fixed size 4 with a remainder of 3 forming a new group. Otherwise the remainder (1 or 2) gets added to the existing groups to form (1 or 2) groups of 5. Currently this value is specified as 5 or 6 instead of 4.

Third: After each group plays its 3 matches. I would rebuild the groups (randomly), and redo another set of 3 matches for each group. I would repeat this cycle, of grouping, playing intragroup, and reshuffling, for some number of times. I would suggest we do it seven times or maybe more.

OK, why these changes? Basically to give those playing this tournament a legit chance as winning Ultimate Player. Also we probably want to create more matches or the tournament will end quickly compared to other tournaments (I am guessing here since I am not sure how long other tournaments might be).

In terms of length of the tournament, a group can play out to 10 victories/rounds 3x within a day for sure (or maybe on 3 different occasions). So will that be it? Will the qualifiers be just one day basically? Seven sounds better.

OK, now why I picked 4, why I want random reshuffling, and why seven qualifier rounds (alternate explanation).

Contrasting 1 vs 1 with LMS:
-- In 1-1, if we assume all players are about equal, the average number of points per game will be 2.5 (or less if you forfeit, which would be rare for good players likely to be candidates for Ultimate Player). Basically, you win one, lose one, win one, lose one, etc. Your winning percentage is .500. Your points every two games would be 4+1=5. So your average points per game would be 2.5.

-- If we keep AA LMS as orginally, you win once, you come in second once, you come in third once, you come in fourth once, and you come in fifth once (for the case of 5). The total points are 4+3+2+1+1=11. The average is 11/5=2.2. If however, we put people in groups of 4, then we end up with 10/4=2.5.

What does this mean? Well, assuming everyone in all tournaments is about equal, AA LMS winners will end up with a lower average point per game. And this average is basicaly what determines who wins the Ultimate Champion (assuming all tournaments are stacked about the same). So 5 players per group leads to an average of 2.2. Instead for 4, AA LMS would have an average of 2.5 just like a 1-1 based tournaments.

Yeah, but ....

OK, look at things this way. To take an extreme case to make the picture clear, if you are in a group of 100 people that are about your same skill, you shouldn't expect to win except once every 100 times; whereas against those same people in a 1-1 tournament setup, you would win 1 lose 1 win 1 lose 1, etc. No matter the format of the tournament (LMS or 1-1) you can rank 1 through 100 with equal probability, but with the 100/group setup, you would expect to end up with *many* fewer points than with someone winning a 1-1 tournament because you would not be first or second or third except once every blue moon. You would basically get 1 point per match, even if you were the top player among a competitive field of 100. Thus, while putting 100 per group doesn't really hurt who wins AA LMS because the best player will eek out a few more points than all the rest (by coming in 1, 2, or 3rd a little more than average), it will bias this AA LMS tournament winner with *many* fewer points than the comparable 1-1 situation (where you would expect to get about as many 4 pt contributions as 1 point contributions).

Note that I assumed everyone was about equal. I suspect that other assumptions will on average still end up with a bias against the 5/group AA LMS setup. The bias is slight I think, but I also think 4/group would even things out a little more (ties and forfeits would probably not affect this analysis too much and those are likely to be very uncommon). So the conclusion is that while 5 per group does not lead to that much of a bias, 4 per group is better. 3 per group or fewer would give AA LMS an advantage over other tournament winners. This would not be fair either. [FYI/FWIW, note that this last case of 3 or 2 is so because AA LMS scoring gives second place 3 points. In 1-1, second place is called the loser: 1 point.]

Thus I recommend we try to get 4/group if at all possible.

Note, that this analysis (and what follows) can be generalized to apply to **ALL** LMS type point systems to make them comparable to 1-1 round-robin systems (I am assuming round-robin. Other formats can be analyzed if they exist in the tournament.)

Now, why redo the whole thing 7 times with each time having a different group (in general, if the randomization is decent)? Well, lets contrast 1-1 with AA LMS.

In a round-robin 1-1 set-up, the best player plays the second best player one time for every time s/he plays the worst player, the next worse, the next worse after that one, etc. In other words, good players (likely to be candidate for Ultimate Player) as well as any other player really, get to play a really tough match for every really easy match played. This means that 1-1 balances things so that good players as well as next best player, etc, will win many matches.

The way AA LMS is currently set up, if two really good players end up in the same group (eg, #1 and #2 end up in the same group), their point totals will suffer realative to 1-1 tournament players. In particular, number 2 might damage the ability of the number 1 player to win. For example, the best player of AA LMS may win match one, get second in match two, and get third in match 3. This may be good enough to win the whole thing because no other player did that well, but the clean sweep was destroyed because the second best player in the AA LMS tournament was there to give #1 heck. Thus #1 gets 4, 3, and 2 points. That average is good, but not great. Meanwhile, during qualifiers of a typical 1-1 tournament, the top player swept through with a perfect 4.0, having only had to play #2 once in the tournament finals (or maybe once during a qualifier, which #1 won in a tight competition). So to repeat, the AA LMS current setup is not fair any more than it would be fair to risk having 1-1 matches occur by putting number 1 to play against number 2 over and over and over (if their draw was unlucky) without any opportunities for easy matches against the number 14 and number 15 ranked players (assuming 15 players). In fact, with only 3 groupings of 5 (or 3 of 4) the odds are 1 in 3 that #1 and #2 of AA LMS will end up in the same basket, married to each other for the entire duration of the qualifiers, making their lives miserable and destroying their point totals.

OK. So if we randomize and play another round, number one has some small chance to end up in a group with number 2, but likely this won't happen too frequently. There will be cases where number 1 will have a walk in the park and easily win all 3 matches within the group (racking up a perfect 4.0 average every so often). Eventually, number one wil lose here and there. That is as things should be. This would happen when seven qualifiers with randomized groupings are played. Win a bunch and lose occasionally. The current situation is: almost guaranteed to lose one or two of the 3 shots within the single qualifiers.

I know that there is a chance that #1 will have a walk in the park, but think about it. You have to expect that even being #1, there are 4 other players that can spoil the fun. In 1-1 there is only one other to spoil the fun and most of the time, #1 will be ranked much higher. Anyway, I do like the seven round system better. Though not thoroughly proven mathematically (not even to myself), I have a strong hunch that it would be more fair. In the process, it provides a way to give more games to AA LMS participants since 3 matches total for qualifiers is probably very little.

Why randomize? Well, this is the way to attempt fairness without resorting to some thorough but complex schedule to make sure that all combinations of groups play each other (yuck). It's a simple way so as not to worry about scheduling combinations without missing any or scheduling 100's of games perhaps in order to get all combinations (or using a very complex formula/ computer simulation to do a subset of all combinations that would be statistically representative.. double yuck).

And why specifically *seven* qualifier rounds? Well, I just picked that number. Basically, for the regrouping to have meaning, you need to do it a few times. We can do five or four, eg, but is that enough games for the tournament? [See the reply to why 7 at the very top of the post.]

Hope this all makes sense.

Please give feedback if you don't like any of the suggestions. I really think these tweaks (if I haven't made an error) are minor but improve things for AA LMS players that want to play a healthy number of games and want as fair a shot as anyone else at winning Ultimate Player.

Finally, why 3 matches per qualifier round (as the rules suggest)? Why 10 rounds per match (as suggested at the top)? [Why 7 qualifier rounds (as suggested here also)?] Well, these numbers are rather arbitrary, but they are probably as good as any and better than many, assuming 15 players and allowing for a healthy amount of game play time. I think all of these can be adjusted without too many problems, but we have to pick something. .. oh, yes. If we don't accept the 10 but use something else, then that affects the amount of game play time if we keep the other numbers the same. If we lowered that below 10, I would suggest we increase the 3 or the 7.

Recap of suggestions, not necessarily in the order presented above.

1 -- Use groupings of 4.

"Qualifications: All players will be split into groups of 5 or 6 players (depending on the number overall number of players registered and confirmed)."

=>

Qualifications: All players will be split into groups of 4 players (depending on the number overall number of players registered and confirmed).

[If you want, add in the explanation given near the top about how the remainder will be divided to a new group of 3 or else to produce one or two 5 player groups from the existing four player groups.. you can adjust that though.. the main point is to try and stay as near to 4 as possible.. higher will bias against AA LMS; lower will bias in the favor of AA LMS and against the other tournament's Ultimate Player hopefulls.]

2 -- Do seven qualifier rounds. Groups (of 4) are selected randomly for each round. Only at the end of the seven rounds will the lowest third be eliminated.

"Based on these points a final qualifications chart will be determined once each group played 3 matches. The bottom third (worst performers) are to be disqualified while the rest proceed to the main tournament."

=>

After doing seven qualifying rounds, each round consisting of a set of groups generated randomly (ie, each group during each qualifying round to play 3 matches), the bottom third (worst performers) are to be disqualified while the rest proceed to the main tournament.

[This is just one wording that pops into my mind now. You can probably work the idea in better.]

3 -- Specify the number of rounds per match as 10

"Server configuration: ..."

=>

Server configuration: ...
10 rounds per match

[10 is just a number I remember being set up in the servers I played at. The point is to specify some number in order to be more complete. This isn't too important, but maybe it would help participants get an idea of how long each match will last (eg, to schedule around their other activities). Also, you may just want to leave this up to those competing to decide on the spot (if that is even possible to do).]

Offline
Joined: 2007-09-10
Just checked the

Just checked the openarena/nexuiz rules which shows that the groups there will be of 5. That is fine. The odds basically are still of gettng 4 points "half" the time and 1 point the other "half". That is different than here in AA LMS where the larger the group the lower your average points per match is likely to be because of the increased competition for the top spots (since only the top three spots get more than 1 point).

For reference: the Ultimate Player formula boils down to comparing the average points won per match during the respective tournaments won by the tournament winners. The idea is that great players would score about the same no matter what point system is in place.

So that is why I suggested the 4 per group (necessary only for AA LMS).

Group reshuffling is suggested to make up for the fact that #1 and #2 may end up in the same group (roughly 1 in 3 chance when using the 13 player scenario) which effectively means they would play against each other in *every* match of the qualifiers if we only have that single round of 3 matches (vs the suggested seven). This is unfair as #2 may be able to beat #1 almost half the time. 1-1 never puts #1 vs #2 except *at most once* in the qualifiers and at most twice in the tournament (if double elimination and if they are that evenly matched). Actually AA LMS would put #1 vs #2 throughout the *entire tournament/qualifiers* if they were in the same group originally (since they would necessarily be in the same group in the tournament round).

Offline
Joined: 2007-09-10
Uh, oh

[From the rules..]
>> Final Tournament: Those who are qualified proceed for another round of 3 matches among themselves. The points will be awarded based on the same system as for qualifications, but will not be summed with the points earned in the qualification stage. The one with the most points wins the tournament.

I just realized that the rules say that the total points for determining the winner of AA LMS will only involve the 3 matches in the tournament round.

First point is that the very long argument I made above is inapplicable now insofar as the points during qualifiers do not count towards determining the winner of AA LMS (I presume this means they won't count for computation of the Ultimate Player average either).

The second point is that this seems silly to ignore what happened in qualifiers in order to compute Ultimate Player. It's not completely silly. I can see some logic, but this does mean that the winner's average will only be based on tough competition, and only in three matches (meaning 12 maximum points).

Are the other tournaments going to be that limiting also? Will the other tournaments' points to be used for Ultimate Player calculations also be that hard to come by (ie, be won against such tough competition)? If those other tournaments don't do the same, that would really bias against the AA LMS winner.

Assuming all tournaments would base their winner's points on only a few tough play-off matches, is this really what we want? This can work out, but I just hope that people are clear on this. Basically Ultimate Player will be determined on only three matches where the points per match are the discrete values 1, 2, 3, or 4. I worry this will result in ties or in a single mistake during one match effectively disqualifying you from Ultimate Player.

Again, that can be OK, but it just wasn't what I expected. I was looking at Ultimate Player a bit like winning most valuable player during the season (major league baseball) or winning season batting average, as opponsed to being MVP of the play-offs or getting the highest batting average in the play-offs only.

Either is fine. I was just surprised. Basically, players should know that UP will be based on their performance in 3 matches only (if the rules stay as they are).

And I am assuming that these three matches will be fair for AA LMS (ie, 4 players per group match) and that the other tournaments' players are also in the hot seat of having their limited points coming only from a few matches, from a few tough play-off matches, where it will be very difficult to get 4 points in any match.

Libervisco, sorry to be reading some of these details so late. I definitely respect the work you are putting into this. As someone on the sidelines, it is difficult to keep up (you know, since I have no pressure on me to keep up.. except for now with the deadline hours away Smiling ).

Offline
Joined: 2007-09-10
I should make clear, the 3

I should make clear, the 3 final tournament matches, if run with more than 4 players per match, would seem to me to be biased against the AA LMS winner when computing averages for Ultimate Player.

The same argument used in the long post to argue for 4 players per group would apply in the final tournament, too (it would apply in any situation where the points from such a match would be used in the UP calculation).

libervisco's picture
Offline
Joined: 2006-05-04
Let me know when you're

Let me know when you're finished Jose. Sticking out tongue

I'll then take a deep breath, when I dare, and dive in, because I have to.. consider everyone's input.

Thank you

Offline
Joined: 2007-09-10
Small amount of playing time for the tournament

Looking over the rules for the various tournaments, it seems that not that many matches are being planned. In some games that might be fine, but in others wouldn't a single match last a very short time? Having only a handful of matches would make the tournament/qualifiers a bit short I think.

Wouldn't it be possible to play several matches in a single day in a stretch of an hour or two in most cases? This would make the time playing in the tournament be just a few hours total though maybe thinned out across numerous days.

Maybe there is a plan to have the game times or frag counts or rounds be a very large number. If this is the case and was stated in the rules I missed it.

Anyway, just speaking up my preference for a longer amount of tournament playing time. After all of that hard work.......

BTW/FYI/FWIW, this is my seventh post on this thread today (in case you missed some and want to read them). In the first I was fairly clueless (a meaningless post). The other six start with the very large post and address/consider the current rules with specificity.

Offline
Joined: 2007-09-10
Last comment I really hope

I didn't expect this, either. Let me try and help.

The topics considered were these:

-- The tournaments seem short
-- The point systems for determining Ulitmate Player (UP) seem to be based on a very small number of matches (at least for AA LMS, and it would be unfair to have AA LMS be the only one where this is the case)
-- AA LMS groups should try to be of size 4 for matches where the points will be used for computing UP averages.

Now, the first main set of posts did not consider the first and second item above, but assumed that there would be many matches from which would come the points (for all tournaments).

If the various tournament rules were to stay exactly as they are right now as concerns the number of matches and the points related to those matches, then the only important part of what I posted is the part about having 4 players per group when doing the final tournament round of AA LMS. You may just want to narrow things down to that if you don't plan on adding more matches. In this case, I am not sure how to change the rules. Maybe state that if there are more than 4 players qualifying for the end (instead of 7 or any other number), there would be a post-qualifier match to remove the weakest players of the bunch to leave only 4. Afterwards we would have the 3 final matches discussed in the rules.

I'm playing no matter what. Don't get me wrong. I just wanted more matches. [and I thought I was contributing a few small changes to the rules to make the dream of being UP a realistic goal for AA LMS. I now see I may have stepped into a tar pit. Oh, well. That's what? .. a two frag penalty I think. Smiling ]

I am really really going to try and post no more for a while. [I'm busy for a large part of tomorrow. Trying to speak my peace before it's too late.]

Good work with the tournament, btw, no matter how critical I seem at times.

Offline
Joined: 2007-09-10
Last comment + 1

>> I'll then take a deep breath, when I dare, and dive in, because I have to.. consider everyone's input.

Poor you. That's the price for organizing a show and trying to be fair Sticking out tongue

OK. This post is a serious post. I considered carefully how posting again today on this thread was not to be taken lightly at all. Here goes....

****

I wanted to try and explain how one of the main issues I raised ties in to the discussions we had before about a point system with adjustments and such. This may help reconcile what may appear to be inconsistencies.

Every detail of a point system basically provides room to create a unique point system. Putting 2, 3, 4, 5, 6, or 7 players per group creates 6 (or however many) distinct point systems [AA LMS]. It may not seem so, but these numbers affect the effective range of points the players are likely to get and so constitute a different point system. Ex, a point system where everyone basically gets 4 points per match is not the same as one where everyone basically gets 1 point per match, even if it is *possible* to get a 1 or a 4 in each of the two point systems. Alright, so if we believe that every detail affects the distribution of points. And if we buy that these distributions should be matched up (adjusted) across tournaments, then we see that we can probably use an adjustment (a scaling multiplier) in this case to more or less solve any problem of point system distribution mismatch. But we want to avoid explicitly using an adjustment if possible. That was the intent in designing all point systems to be based on a model using basically 1 through 4 points no matter the type of tournament. [correct?]

To make it clearer how a detail like the number of players per group affects the distribution of points, all else remaining equal (eg, with the players being of the same quality in all cases), consider this. If we have 2 players per group in an AA LMS match, there is a 50% chance for a player to get 4 points (to first approximation.. go with the flow for a sec). There is a 100% chance of getting a 3 or a 4 however because the worst you can do is finish second (3 points). Now consider a case of 1000000 in a group. In all likelihood, you will not manage to win after a modest number of matches are played even if you are very good. There are just too many other competitors all fighting for a very few spots (1st, 2nd, or 3rd). Over the long run, you will win a few games, but your average would be close to 1.0. [All of these numbers are based on the current LMS point system.]

Now, in both cases the quality of the players and their game play was not assumed to change. Only the point system changed. So what I am saying is that while the point system discription in the general rules for 1-1 and LMS are very good models, it is not enough. It's like saying that I will give you 4 bucks if you win whether you use game type A or game type B, but in one scenario you have tho climb Mount Everest to "win" and in the other scenario you just have to spell your name correctly to "win". Clearly, simply using the same scale is not enough if the difficulty of getting the same value is very different among the different games. [In the general case, all point systems use the same scale from 0 to infinity.]

Anyway, the explanation for the 2 types of point systems (1-1, LMS) is very decent and logical. It makes an initial good attempt at creating compatible point systems, but we can't put further obstacles. In other words, the unspoken assumption with the discription of those two point systems was that a win in one system is as easy as a win in the other.

OK, so we can resolve the problem by using adjustments. Eg, if the likely average in one point system is 1.06 but it's 2.5 in the other, we can scale one to level the playing field. BUT we were trying to avoid using this sort of adjustment explicitly (or any kind of adjustment for that matter).

OK, so what I was looking at was that while putting 2, 3, 4, 5,... players per group creates different point systems (ie, different distributions of points, different averages, etc), it just so happens that the case of 4 per group matches very closely the standard 1-1 point system in terms of likelihood of getting a similar average point per game (ie, a similar likelihood overall of getting similar points whether you win or lose or finish 2nd, etc). I am making assumptions and trying to use my gutt to define "fair" and to tie down probabilities. I presented the argument so that others can look it over and agree or disagree. I make no claim to have been rigorous and have presented proof beyond a shadow of a doubt.

So the short of this is that 4 players per group gives a point system for AA LMS (IMO) that is about as close as we can easily design ahead of time to match the point system of a typical 1-1 round-robin point system. In this particular case of 4 per group, the adjustment would be 1 to get it to match the other point system. In other words, there is effectively no adjusting to be done.

Hope this makes sense and helps tie in the earlier discussions to the problem I have with 5 or 6 or 7 etc players per group as currently mentioned in the rules for AA LMS.

****

I had other issues, but they are preferences I have (eg, to have more matches). I don't think the fairness is damaged as long as roughly a similar number of matches is used to determing the points for UP across all tournaments and these matches are of similar difficulty (ie, all tournaments use play-off matches or they all use qualifier+playoff matches, etc).

Warning: If the 3 (sets(?) of) qualifier matches for AA LMS are used for UP points, then (in addition to 4/group) I do suggest that various qualifier rounds be used (and not just a single qualifier round) with random groupings to help even out the odds that the #1 player will have both difficult and easy matches just as with 1-1 competitors. Otherwise, once again we should find that 4 points on AA LMS will be more difficult to achieve than 4 points in the typical 1-1 round-robin setup. This would happen, eg, if the AA LMS #1 player ends up in the same group as the #2 and #3 players. Whoever wins out in that group would have had their points take a beating as there likely would have been different players holding the top spot in all 3 matches. Contrast that with 1-1 where it is impossible by design to have #1 play #2 except once for every round-robin round (meaning that there would be easy matches in there to pad the average). Doing several qualifier rounds with random groupings [AA LMS] would help smooth out anomolies, increasing the odds that the good players will have rounds where they can pad their averages, too.

Offline
Joined: 2007-11-11
just when i thought i could

just when i thought i could make long replies on my forum...

please forgive me i dont read everything.

-Q

edit: bad me, i just did read everything...
ill come with a reply soon.. (/me is tired of reading Sticking out tongue)

libervisco's picture
Offline
Joined: 2006-05-04
Jose wrote:First point is
Jose wrote:

First point is that the very long argument I made above is inapplicable now insofar as the points during qualifiers do not count towards determining the winner of AA LMS (I presume this means they won't count for computation of the Ultimate Player average either).

I am reading and reading... all decent so far, and then I hit that sentence. Wrong! Smiling The qualifier points are counted just as well as the final tournament points for the Ultimate Winner. They're just not counted for the winner of AA LMS. Why? Because qualifiers are for qualifying and nothing else. Hence once we have our qualified players we shouldn't let their qualifying results influence the final tournament winner. It must be a clean slate.

However, for the Ultimate Winner of the fest, points are tracked and counted for EACH match played regardless of which tournament it is and which stage it is.

I'll reply further on as I read.

libervisco's picture
Offline
Joined: 2006-05-04
Alright Jose. Those were

Alright Jose. Those were some interesting observations and suggestions. I like to keep things clear and organized, mostly for myself, but I think it helps others too, so I'll lay down the issues you seem to be addressing:

  • 1. AA tournament doesn't have enough matches compared to other tournaments making it harder for an AA player to win the fest
  • 2. More than 4 players per group is too much because it then becomes harder for a player to gain a point.

Your suggestions:

  • 1. 1 match = 10 rounds, regroup and repeat 3-matches 7 (or other amount) of times in qualifiers
  • 2. Have each qualifier group have 4 (or as close to 4) players as possible.

My reply is: well done. Smiling

I may have perhaps missed some of the finer grained points, but I think I'm getting the gist of what you're trying to say. I agree to both of these suggestions, mostly. I'm just not sure about regrouping, especially doing it 7 whole times, but some balance could be stroke there.

But I definitely agree that one match should constitute a multitude of rounds (up to 10 at most) and that each qualifications group should be about 4 players.

Edit: One way we perhaps don't need to limit qualification groups to only 4 players in each is actually revealed in a post from an earlier topic:

libervisco wrote:

Of course, if we are to stick with a "you must get at least 1 point for each match you participate in" then we can extend that 1 point for the fourth best in a fragfest to all of the rest.

So this way, while in groups larger than four it still becomes harder to compete for points more than "1", mere participation, simply playing a match already guarantees you 1 point.

Qbrain, and anyone else who may be reading, what do you think? This should be the last chance to change the rules.

Thank you

Offline
Joined: 2007-11-02
Qbrain was exaggerating

Qbrain was exaggerating about the one turn per second, a cycle delay of 0.1 (normal in fortress/sumo) lets you make ten turns per second (which makes the playing level between those who doublebind and those who dont even).

And i'll have to agree with Maz about the intuitive thing. Beside those points already mentioned by him, with lowrubber settings you can actually see the turns in a wall, in a highrubberserver a guy can just grind against a wall, do fifteen 180's and 4 adjusts and you wouldn't see the difference with someone who just did a grind. Being able to know how walls are placed is far more important for a beginner than not dying even if you hit a wall (which is counterintuitive anyway if you just started to play). People expect to die if they make a bad turn, but they don't expect to die if they already outgrinded someone and suddenly die halfway in the tunnel.

Another problem I see with highrubber is all the extra rules you need to make. For example, while I don't deny that camping exists in lowrubber, camping is a much bigger problem in highrubber because you only need a very limited space to be able to live 'forever'. Next to a no-camping rule, you'll also need to get a consensus about instants, closing, speeding, etc.

About servers, there's the Nexus9 server Maz mentioned (which, to be honest, might have a bit too little rubber for complete noobs + it has a Turbo, which is a big nono imho).
There's also World of Hate which mimics a server which used to be pretty popular before there was a big gap between those who played high and lowrubber, but I'm not sure if it's any popular nowadays. You might want to take a look at shrunkland too (even though it says it's for experienced players only).

Offline
Joined: 2007-09-10
>> I am reading and

>> I am reading and reading... all decent so far, and then I hit that sentence. Wrong! Smiling The qualifier points are counted just as well as the final tournament points for the Ultimate Winner. They're just not counted for the winner of AA LMS.

Whew! That's quite a relief.

I went back to the rules and could not find it stated one way or the other.

And speaking of AA LMS rules, the only link to those rules I found were from the comment you wrote to which I replied with the "long post." I didn't see a link from the main Rules page, and if you go from the Tournament's page you have to first go to this thread and then search until coming across that link in your post http://www.nuxified.org/topic/organizing_armagetron_advanced_tournament#comment-11068 .

Taking a sneak peak at your next reply, it looks pretty good from my perspective. That was quite a performance to keep up with the long twists and turns I put out at the eleventh hour. I hope we don't meet until the finals of AA LMS!!!

libervisco's picture
Offline
Joined: 2006-05-04
Interesting points. I'd

Interesting points. I'd like to see a response from others, like QBrain or Zion about these issues. Maybe we can do some sort of a compromise.

Q did mention that the no camping rule would be necessary here. As for the rest I'm not sure yet.

libervisco's picture
Offline
Joined: 2006-05-04
Talking in-game with QBrain

Talking in-game with QBrain on a DarkSyndicate server we came to the conclusion that it would be best to have 4 players per group when playing all in one, but regrouping doesn't sound very exciting. Sticking out tongue

Basically, the idea is as follows. We have groups of 4 each playing all-in-one matches 4 times (basically 1 match per player). Then based on the charts we get we disqualify 2 from each group. If we have 3 groups that means 6 players go into the tournament. Then we split these 6 into groups of 3 which again play all in one matches until we disqualify one more from each, being left with 2 players in each group.

Then the lower two and the higher two both play 1vs1. This gives us the winner, second place and third place.

All in all, with 3 groups of 4 for qualifications playing 4 all-in-one's we have 12 qualification matches. If we continue to follow the rule of one match per player in a group, the tournament would then count 8 additional matches. All in all it's 20 matches.

Now if that seems low I suppose we can further adjust the numbers. I proposed, partly based on what Jose says, to match up the number of matches to the one they'd play if it was 1vs1 all the time. In that case, one group of 4 would inevitably play 6 matches so 3 groups would count 18 matches in qualifications. After the tournament we'd then have total of 26 matches.

However, the disadvantage of that would apparently be that playing 6 all in one matches with same people would quickly get old..

But anyway, these are the number's we're working with if we have 12 players. They'd obviously increase if we have more. It's possible we'll have 16 or at best even 20 players in which case we'd have 4 or 5 groups of four which gives further opportunity for matches.

Offline
Joined: 2007-09-10
Overview of this reply: I

Overview of this reply:

I split the reply into parts that attempt to focus on an issue at a time.

In the first section, I address one of your summary points (#1) because it is off a bit. I try to clarify a few things. The most important point is that 4 players per group fixes most of the imbalance as concerns Ultimate Winner calculations. You seem to agree with this, but you may not understand why I like 4. In a later section I redo a little of the math with extra discussion.

In the second section, I try to clean up a mistake I made in the earlier arguments. Basically, I went a bit too far with an earlier analogy and that may partly explain some of the ensuing confusion. This section is best skipped on first reading.

In the third section, I try to explain the reasons I think groups should be redone randomly at least a few times (eg, prior to starting each new qualifier round).

In the fourth section I address a comment you made that confused me. As a result, I cover a little more the reasons I have for chosing 4 players per group. I do briefly go over an example of having 7 players per group.

In the fifth and final section, I simply point out that ties are not a part of the AA LMS point system and I query on how the winner of a (10? round) match will be determined.

****

I think your summary captures almost everything fairly accurately (once we factor out the confusion I had over how Universal Winner (UW) is calculated), but I am going to seek clarification on two items: your first summarizing statement and (later on) on why and what randomizing groupings between qualifying rounds helps.

>> 1. AA tournament doesn't have enough matches compared to other tournaments making it harder for an AA player to win the fest

The main reason I think an AA player is disadvantaged (with current rules) is with the 5/6 sized grouping instead of 4. Fixing this I think balances things fairly well.

Separately, I also do think that there are too few matches in AA LMS. Too few matches means little playing time, meaning you would be close to getting your money's worth (ie, the $0 entrance charge Smiling ).

I want to recognize a mistake I made, but first let me emphasize from my recent post today that UW calculations are based on points from qualifiers and tournament matches, not just from tournament matches as I assumed during some of the arguments. From here forward, let's try to forget most of what I said in the latter batch of comments that relied on this incorrect assumption. The focus should be on the initial long post that proposed 3 suggested changes to the rules.

You may want to skip this next section since it is a bit of a tangent and slightly more confusing than other parts of the reply. Maybe read it later on.

****

**** [Skip?]

The mistake...

I did use the example that #1 has a greater chance of meeting up with (eg) #2 throughout many of #1's matches. [Reference the paragraph starting: "The way AA LMS is currently set up...."] I said this would be similar to having #1 vs #2 matches repeated frequently in 1-1 tournaments. Well, the idea was correct but I said ...

>> This may be good enough to win the whole thing because no other player did that well, but the clean sweep was destroyed because the second best player in the AA LMS tournament was there to give #1 heck. Thus #1 gets 4, 3, and 2 points. That average is good, but not great. Meanwhile, during qualifiers of a typical 1-1 tournament, the top player swept through with a perfect 4.0, having only had to play #2 once in the tournament finals (or maybe once during a qualifier, which #1 won in a tight competition).

If we have groups of size 4, then the above is inaccurate. If we run many experiments we should see (using the probability simplifications I made), that while most of the time the 1-1 scenario will result in #1 ending up with a higher score (eg, close to 4.0), every now and them, the score would be lower than in the LMS case. In these less common cases, the lower score would be significanly lower because the 1-1 point system does not have the 3 point option.

In short, I think the calculation done for 4 per group where we get 2.5 points per game, factors everything in. I muddled the water by introducing the analogy above. The analogy came from my gut and shows that AA LMS is worse off in most cases.. but not in all cases, not in the case of 4 or less per group. I failed to realize the analogy fell short. The discussion which proceded was thus worded in a way that would confuse the issue.

Anyway. All of this going back and forth, assumptions necessarily having to be made, and subtle flaws (hopefully there aren't many more) certainly made the reading difficult to grasp.

>> I may have perhaps missed some of the finer grained points

As just shown, apparently so did I! Smiling

**** [EndSkip]

****

>> I'm just not sure about regrouping, especially doing it 7 whole times, but some balance could be stroke there.

OK. One reason to have more than one qualifier round is as a way to increase the game time. If no one else has suggestions, I will echo once again that I think more than one round is very desireable. Is 7 the "right" number? I don't know. We'd have to look carefully at the playing times involved here and in the other tournaments. We'd have to see what we want from the tournament. Anyway, 3 or 4+ qualifier rounds would be a very nice improvement (note that I only have a vague idea about the total playing times among the various tournaments).

The main reason to do something like randomizing is to prevent one odd group drawing from destroying the overall UW balance. In some cases, the AA LMS player that ultimately competes for UW might be in too unfair a hole. In other cases, this player would end up with better odds than the other tournaments' candidates.

Really, there is no inherent advantage or disadvantage to reshuffling. We can just accept the fact that lottery draws will necessarily hurt some and help others. I'll agree. My point would be that if randomizing can be achieved with little extra effort, it means that the odds of getting a *very* bad or a *very* good draw are diminished significantly since in this case such a particularly good/bad draw would have to occur by chance after every randomization.

By way of anology, think of Russian roulette [ http://en.wikipedia.org/wiki/Russian_roulette ]. Playing 10 rounds with a bullet that is 1 tenth as likely to kill you if it hits you may be deemed desirable if you really don't want to die. Yes, you are more likely to end up partially hurt, but you are less likely to die (assuming it would take all 10 shots to kill you). So the idea would be to lessen the odds that one bad draw would spoil the fun for the AA LMS players (or for everyone else). Across several draws, odds are much greater that the good would cancel with the bad more times than not. In the Russian roulette analogy, many draws would bring you closer to the average of 1 deadly bullet and 5 blanks. [That average is still painful -- fortunately, for this tournament, the average is "neither good nor bad," ie, no advantage / no disadvantage. In this sense, the Russian roulette analogy was overkill (lol). ..Yes, this analogy was horrible, yet efficient, all at the same time Smiling ]

What consitutes a *very* good/bad draw? I am not sure we would be able to know without the ability to play many sample tournaments to see which draw ended in a very unlikely final outcome/point total. *If* we could identify it, then we could just trash the grouping made if it would be a very good/bad grouping, or perhaps just not select such a grouping in the first place. For example, if we had an idea of the quality of the players (assuming no player had a significant weakness specifically against some other player that was ranked much lower), then we would want to distribute the groupings so that generally you had some "nice mix" of good and bad players. I don't know what a good mix would look like, and it might not be a good idea to make assumptions about the quality of players given that there have been no qualifiers. Randomization simply works to our favor without us having to think very much about the details. It's main goal is to reduce the impact of a very lucky or a very unlucky draw.

****

****

>> >> Of course, if we are to stick with a "you must get at least 1 point for each match you participate in" then we can extend that 1 point for the fourth best in a fragfest to all of the rest.

>> So this way, while in groups larger than four it still becomes harder to compete for points more than "1", mere participation, simply playing a match already guarantees you 1 point.

I'm not sure if you understand why I think 4 is ideal.

Let me return to your second point at the top of the post

>> 2. More than 4 players per group is too much because it then becomes harder for a player to gain a point.

Well, it becomes more difficult for a player to gain 4, 3, or 2 points. Everyone already gets 1 point (at least according to the draft of the rules I read).

Let me re-explain where "4" came from.

I assumed that a random player had an equal chance of finishing 1st as finishing in any other spot. [I'll speak a little bit more on this assumption below.] For four players this leads to an average of 2.5 points per game (I'll rework the math below). This matches exactly the average for the 1-1 tournaments. The calculation for the case of 4 already factors in that at least 1 point will be given to everyone that participates in the match. Is this what you are saying above? I think it is already factored in.

Let's say we have 7 in a group. Then, with equal probability, you can get 4,3,2,1,1,1 or 1 point. On average, you will have (4+3+2+1+1+1+1)/7 = 1.85 points per game. This is below the 2.5 from the 1-1 tournament. This means that the average of all point totals scored for AA LMS would be 1.85 while it would be 2.5 for the 1-1 tournaments. This *seems* unfair to me, because it *sort of* means that the AA LMS winner will already be in a hole and will have to rise that much more above the average just to end up where others from other tournaments will be with much less effort. Heck, someone with a losing record, scoring say 2.14 average (note that it's below the 2.5), in a 1-1 tournament is still scoring higher than most people in the AA LMS tournament.

There are various ways we can try to even things out. I mentioned elsewhere that I think Z-scores would be used in a situation like this. To give an idea of just what is a Z-score, I quote from the google translation of the german wikipedia page for Z-score. Keep in mind I haven't really taken any significant stats math course and also do not want to complicate this tournament's point system.

>> Z-scores in some cases make it possible to samples from different populations in a meaningful way (through their Z-scores versions) comparable.

Anyway, my simplified approach to "harmonizing" AA LMS points with the other tournament point systems would be to find the group size so that the average points per game (across all players.. ie, for a random player) would be 2.5 to match that average for 1-1 tournaments.

How did I get this average?

For 1-1, I ignore ties and forfeits and assume a random player is as likely to win as to lose. This last part is true if we look at the average across all players since for every win there is a loss. [I ignore ties and forfeits, which I presume would be unlikely, maybe rare; this isn't a perfect assumption, but I wasn't sure how to simply factor that in to the calculations. Plus, the LMS point system doesn't even allow for ties! We may need to correct that oversight. I'll mention more on this last point below.]

1-1:
With equal probability you can get 4 or 1 point. The average then is (4+1)/2 = 2.5 points per game.

4 per group:
With equal prob you can get 4, 3, 2, 1 points. Ave: (4+3+2+1)/4 = 2.5 points per game.

7 per group (compare this with the 4 case):
With equal prob you can get 4, 3, 2, 1, 1, 1, 1 points. Ave: (4+3+2+1+1+1+1)/7 = 1.85 points per game.

[Note that I use "random player" sort of as a substitute expression for "the average player." The bottom line though is that a great AA LMS player has to stand out that much more from the average, to match, point-wise, a typical just-better-than-average player in any of the 1-1 tournaments.]

OK. Well, this re-explains my analysis to some extent. You said you liked the choice of 4, but maybe now it will be clearer. I compared the averages for the two point systems. Only with 4 per group in AA LMS, do you get a comparable (equal) result with the 1-1 case (eg, 7 per group gives a lower average.. as would any value higher than 4). This set of calculations do factor in giving 1 point for "the rest."

****

****

Now, let's look at the matter of ties for AA LMS.

First of all, will the winner of a match (say after 10 rounds) be determined by us by adding up individual wins or will we let the AA server do it? The easiest is to let the server do it. If so, will the server ever yield a tie (we might want to ask on a dev forum or mailing list .. unless someone knows where the code is that handles that or is very sure of the answer). If we do the computation manually (eg, player 1 won 4 rounds and the other 3 players won 2 rounds each, so player 1 wins the match), then we have to consider ties. The server may mostly avoid ties if it uses a more elaborate system to pick the winner. For example, it might award points for being LMS, for "frags," for being 2nd-to-last LMS, etc.

Will we do something about ties? What?

****

I am now going to read your last post and reply to that, but I'll try not to repeat anything here (only mention it) since you weren't able to review this comment prior to posting.

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

OK. One reason to have more than one qualifier round is as a way to increase the game time. If no one else has suggestions, I will echo once again that I think more than one round is very desireable. Is 7 the "right" number? I don't know. We'd have to look carefully at the playing times involved here and in the other tournaments.

Wouldn't it be easier to just increase the number of rounds per match then? You initially said 10 rounds, but, guess what, the DS server on which I met with QBrain today has 100 rounds! Laughing out loud

The sky is the limit, if we want more play time let's just play longer matches.

I know you say it's not too complicated to reshuffle, but I beg to differ. We'll have to group everyone taking their location into account (so that we can easier deal with ping issues). This already makes any further regrouping more difficult. And it's just that one additional pesky variable in the whole picture that if we really don't have to do it may be better not done.

As for your clarification of why you think 4 players per group are better.. So it's because of making the probability of getting more points same as with other 1on1 tournaments. Something tells me this might also have something to do with standardizing the amount of effort to move forward between tournaments of hypothetically same number of players.

But let's close that chapter. It's 4 players/teams per group (or as close to it as possible) everywhere, in every tournament, including AA. You can also see here how that also helps standardize the number of games between same-sized tournaments.

Jose wrote:

First of all, will the winner of a match (say after 10 rounds) be determined by us by adding up individual wins or will we let the AA server do it? The easiest is to let the server do it.

If AA server can do it then it should do it. The only case I can imagine in which we wouldn't let the server determine match winner is when we have rules that differ from the rules which server itself uses for this determination. This isn't necessary in this case.

I'd be interested to hear if AA servers ever recognize ties, although fro