Author |
Topic: 2014 World Championship Format (Read 7574 times) |
|
aaaa
Forum Guru
Arimaa player #958
Posts: 768
|
|
Re: 2014 World Championship Format
« Reply #15 on: Feb 6th, 2013, 8:49pm » |
Quote Modify
|
on Feb 6th, 2013, 7:09pm, Fritzlein wrote:A simple compromise would be to discard the unbiased tournament performance rating and rely entirely on the seeded tournament performance rating. I'm not sure how that would work out, but we could keep an eye on it to see whether it is an "even-up" mechanism that serves the intended purpose without over-reacting. |
| I would oppose that, because that would make the seeds directly affect the final standings. My position was and remains that no appeal should be made to any evidence external to the tournament except where it would resolve any indifference by the internal evidence. I may be biased, but right now I haven't gotten any clear indication of the scheduling going astray with respect to what would be desirable. Three lives for each player should be sufficient for the highly dynamic strengths of schedule to become close to respectable by the time survival is on the line for a real contender. Remember that, as enhanced as it may be, this is still an elimination tournament, so don't expect the final result to represent the "truth" with a confidence akin to that of a round-robin league.
|
|
IP Logged |
|
|
|
Janzert
Forum Guru
Arimaa player #247
Gender:
Posts: 1016
|
|
Re: 2014 World Championship Format
« Reply #16 on: Feb 7th, 2013, 1:17pm » |
Quote Modify
|
on Feb 6th, 2013, 7:09pm, Fritzlein wrote:A simple compromise would be to discard the unbiased tournament performance rating and rely entirely on the seeded tournament performance rating. |
| For me this isn't a compromise but closer to the ideal situation, so long as seeding was by EWHR. At the same time though while I'm mildly worried about the current system pairing it hasn't so far seemed problematic. I'm disappointed though that we haven't tried out these different tournament systems in a simulator to see how they compare. Janzert
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2014 World Championship Format
« Reply #17 on: Feb 7th, 2013, 9:50pm » |
Quote Modify
|
on Feb 7th, 2013, 1:17pm, Janzert wrote:I'm disappointed though that we haven't tried out these different tournament systems in a simulator to see how they compare. |
| If you simulated tournaments to compare different systems, what criteria would you use to say one system is better than another? One beloved criterion is that the tournament should have a high percentage chance of crowning the truly strongest player as champion. However, Omar showed that, under reasonable assumptions, the tournament format with the best chance of crowning the truly strongest player is the format that awards the title to the highest-rated player without playing any games. Of course it is absurd to give the pre-tournament rating that much weight. On the other hand, unlike aaaa, I don't hold it as a first principle that the effect of seeding should be minimized as much as possible. I think it would be right and fitting for seeding to have more influence than it has under the current format, particularly if that made the "even-up" mechanism work less randomly than it has worked so far. Quote:Three lives for each player should be sufficient for the highly dynamic strengths of schedule to become close to respectable by the time survival is on the line for a real contender. |
| Yes, three lives gives a lot of time for seemingly unfair pairings to even out, but even so, why not minimize the unfairness in the first place? If I took your position to the extreme, I could argue that we shouldn't use pre-tournament ratings at all, that the early rounds should be randomly paired when UTPR makes no distinction, and that there will be plenty of time for strength of schedule to even things out before any serious contender gets his third loss. That's just as logical as your argument, only a bit more extreme in the a priori value placed on not using seeds. To me it seems we could use the pre-tournament ratings to make the early-round pairings better, not only reducing the need for "even up" pairings later on, but also making the "even up" mechanism effective at an earlier date rather than needing to wait several rounds for an accurate strength of schedule to be established.
|
|
IP Logged |
|
|
|
Hippo
Forum Guru
Arimaa player #4450
Gender:
Posts: 883
|
|
Re: 2014 World Championship Format
« Reply #18 on: Feb 8th, 2013, 2:27am » |
Quote Modify
|
Once again, I don't think there is anything wierd in our current system. Chessandgo received bye ... the easiest pairing. This should be compensated by him receiving stronger opponent next round. So this is generaly fair for others, but it could look unfair for his next opponent... The problem is we want to scale him high for his opponents, but we want to give him difficult matchups at the same time. It seems to me current solution is good enough.
|
|
IP Logged |
|
|
|
Boo
Forum Guru
Arimaa player #6466
Gender:
Posts: 118
|
|
Re: 2014 World Championship Format
« Reply #19 on: Feb 8th, 2013, 3:06am » |
Quote Modify
|
Quote:If you simulated tournaments to compare different systems, what criteria would you use to say one system is better than another? |
| Theoretically most of the time each player should take the final place = to the initial strehgth (WHR) seeding. So I would compare the normal distribution of the final places of each player, trying to answer questions: how fast does a mean of the normal distribution approach its expected value (initial seeding) for any player? which system has a less standard deviation for any player? Quote:why not minimize the unfairness in the first place? |
| I was going to ask the same question, but you were the first to do that Quote:Once again, I don't think there is anything wierd in our current system. Chessandgo received bye ... the easiest pairing. This should be compensated by him receiving stronger opponent next round. So this is generaly fair for others, but it could look unfair for his next opponent... The problem is we want to scale him high for his opponents, but we want to give him difficult matchups at the same time. |
| Chessandgo should have played a strong player who had an easy path as well, not the one who had a difficult path.
|
|
IP Logged |
|
|
|
Janzert
Forum Guru
Arimaa player #247
Gender:
Posts: 1016
|
|
Re: 2014 World Championship Format
« Reply #20 on: Feb 8th, 2013, 9:54am » |
Quote Modify
|
on Feb 7th, 2013, 9:50pm, Fritzlein wrote: If you simulated tournaments to compare different systems, what criteria would you use to say one system is better than another? One beloved criterion is that the tournament should have a high percentage chance of crowning the truly strongest player as champion. |
| In measuring the effectiveness of any tournament system there are multiple conflicting criteria, each one a valid and justifiable concern. I do not believe one criteria should be placed as "the one" concern that overrides all others. Every tournament system will be a compromise balancing the various criteria against themselves. Also at least some, maybe many of the criteria cannot reasonably be empirically or objectively measured. So my disappointment is merely that the few criteria that are relatively easy to measure, some of which were measured and compared in the past, have not been looked at. Janzert
|
|
IP Logged |
|
|
|
browni3141
Forum Guru
Arimaa player #7014
Gender:
Posts: 384
|
|
Re: 2014 World Championship Format
« Reply #21 on: Feb 8th, 2013, 6:20pm » |
Quote Modify
|
on Feb 7th, 2013, 9:50pm, Fritzlein wrote: If you simulated tournaments to compare different systems, what criteria would you use to say one system is better than another? One beloved criterion is that the tournament should have a high percentage chance of crowning the truly strongest player as champion. However, Omar showed that, under reasonable assumptions, the tournament format with the best chance of crowning the truly strongest player is the format that awards the title to the highest-rated player without playing any games. |
| I agree that that is the most important criterion, but I disagree that the tournament format with the best chance of crowning the truly strongest player is simply giving the award to the highest rated player. What about a match elimination format where each round each pair of players play best of 10,000 games, with the last standing player being crowned? Surely this would be more likely (assume the players both have the stamina) to award the strongest player the title than using ratings only. Ratings are subject to manipulation and are not always accurate. Edit: I meant to ask for more details on what omar showed.
|
« Last Edit: Feb 8th, 2013, 6:30pm by browni3141 » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2014 World Championship Format
« Reply #22 on: Feb 8th, 2013, 8:06pm » |
Quote Modify
|
on Feb 8th, 2013, 6:20pm, browni3141 wrote:Edit: I meant to ask for more details on what omar showed. |
| Right, best of the ones under consideration, i.e. better than triple elimination.
|
|
IP Logged |
|
|
|
chessandgo
Forum Guru
Arimaa player #1889
Gender:
Posts: 1244
|
|
Re: 2014 World Championship Format
« Reply #23 on: Feb 9th, 2013, 5:08am » |
Quote Modify
|
on Feb 8th, 2013, 3:06am, Boo wrote: Chessandgo should have played a strong player who had an easy path as well, not the one who had a difficult path. |
| strong = having had a difficult path, so I'm (was) weak for the pairing algorithm.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2014 World Championship Format
« Reply #24 on: Feb 10th, 2013, 10:22am » |
Quote Modify
|
Note to 2014 TD: it would be wise for the rules to explicitly set a time frame for players to appeal to change the result of a game. Currently the rules don't say anything on this score, so I have to make a ruling on my own judgment.
|
|
IP Logged |
|
|
|
Boo
Forum Guru
Arimaa player #6466
Gender:
Posts: 118
|
|
Re: 2014 World Championship Format
« Reply #25 on: Feb 11th, 2013, 2:41am » |
Quote Modify
|
Quote:strong = having had a difficult path, so I'm (was) weak for the pairing algorithm. |
| This is currently how pairing algorithm works. I was talking about how it should be changed. Sorry if I was not clear enough. After the second round the 0=loss player group according to the path difficulty was sorted this way (I don't remember for sure, this is for illustrative purposes only): 1. omar 2. Hippo 3. supersamu 4. rabbits 5. Thiagor 6. Boo 7. Adanac 8. 99of9 9. novacat 10. Alfons 11. chessandgo 12. bye According to WHR the same group can be sorted this way: 1. chessandgo (2623.9) 2. Alfons (2430.7) 3. Adanac (2371.0) 4. Thiagor (2336.4) 5. Boo (2283.1) 6. Hippo (2210.0) 7. 99of9 (2178.2) 8. rabbits (2167.3) 9. novacat (2048.7) 10. supersamu (1978.3) 11. omar (1931.0) 12. bye (0000) And by following the rule "the player, who had the most difficult path so far, should be matched with the player, who is expected to perform the weakest in the future (has the lowest seeding). " the pairing would be: 1) 1. omar - 12. bye (0000) 2) 2. Hippo - 10. supersamu (1978.3) 3) 4. rabbits - 9. novacat (2048.7) 4) 5. Thiagor - 7. 99of9 (2178.2) 5) 6. Boo - 3. Adanac (2371.0) 6) 10. Alfons - 1. chessandgo (2623.9) This is what I meant by saying "Chessandgo should have played a strong player who had an easy path as well, not the one who had a difficult path. "
|
« Last Edit: Feb 11th, 2013, 2:44am by Boo » |
IP Logged |
|
|
|
chessandgo
Forum Guru
Arimaa player #1889
Gender:
Posts: 1244
|
|
Re: 2014 World Championship Format
« Reply #26 on: Feb 12th, 2013, 7:20am » |
Quote Modify
|
gotcha, sorry.
|
|
IP Logged |
|
|
|
Hippo
Forum Guru
Arimaa player #4450
Gender:
Posts: 883
|
|
Re: 2014 World Championship Format
« Reply #27 on: Feb 12th, 2013, 10:52am » |
Quote Modify
|
on Feb 11th, 2013, 2:41am, Boo wrote: This is currently how pairing algorithm works. I was talking about how it should be changed. Sorry if I was not clear enough. After the second round the 0=loss player group according to the path difficulty was sorted this way (I don't remember for sure, this is for illustrative purposes only): 1. omar 2. Hippo 3. supersamu 4. rabbits 5. Thiagor 6. Boo 7. Adanac 8. 99of9 9. novacat 10. Alfons 11. chessandgo 12. bye According to WHR the same group can be sorted this way: 1. chessandgo (2623.9) 2. Alfons (2430.7) 3. Adanac (2371.0) 4. Thiagor (2336.4) 5. Boo (2283.1) 6. Hippo (2210.0) 7. 99of9 (2178.2) 8. rabbits (2167.3) 9. novacat (2048.7) 10. supersamu (1978.3) 11. omar (1931.0) 12. bye (0000) And by following the rule "the player, who had the most difficult path so far, should be matched with the player, who is expected to perform the weakest in the future (has the lowest seeding). " the pairing would be: 1) 1. omar - 12. bye (0000) 2) 2. Hippo - 10. supersamu (1978.3) 3) 4. rabbits - 9. novacat (2048.7) 4) 5. Thiagor - 7. 99of9 (2178.2) 5) 6. Boo - 3. Adanac (2371.0) 6) 10. Alfons - 1. chessandgo (2623.9) This is what I meant by saying "Chessandgo should have played a strong player who had an easy path as well, not the one who had a difficult path. " |
| This is interesting proposal. I am not sure how it would work if a player has intentionaly lowered his rating.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2014 World Championship Format
« Reply #28 on: Feb 13th, 2013, 7:52am » |
Quote Modify
|
It would be great if the pairing software could allow for players to drop out and rejoin later, as in the case of UMDRevan who lost Internet connection for days due to snowstorm. In technical terms, this would mean allowing for zero-point byes instead of only one-point byes.
|
|
IP Logged |
|
|
|
mistre
Forum Guru
Gender:
Posts: 553
|
|
Re: 2014 World Championship Format
« Reply #29 on: Feb 13th, 2013, 9:58am » |
Quote Modify
|
I'd like to propose a change for next year: "If a game ends due to a disconnection the tournament performance ratings of the two players are not effected." The Browni3141 - Novacat result had an impact on tournament seeding that likely could have been minimized if performance ratings were not effected. And this was a mild example. What if Chessandgo had lost due to a disconnection to the bottom-ranked player in the first round? I don't think this change would cause people to time out on purpose as you are still getting a L result. Thoughts?
|
« Last Edit: Feb 13th, 2013, 10:01am by mistre » |
IP Logged |
|
|
|
|