Author |
Topic: 2010 World Championship (Read 16504 times) |
|
Fritzlein
Forum Guru
Arimaa player #706
Gender: 
Posts: 5928
|
 |
Re: 2010 World Championship
« Reply #135 on: Feb 7th, 2010, 7:46pm » |
 Quote  Modify
|
on Feb 7th, 2010, 7:19pm, 99of9 wrote:Tuks, I acknowledge that I have seen your communication that you may change your preferences before the next scheduling is run (i.e. you have my permission - even though it's now against the rules  ). |
|
Without permission from either woh or Omar, I have updated my time preferences to reflect the fact that, for this week only, Saturday is marginally better for me than Sunday. But if this is ruled invalid and I have to keep the currently scheduled time, no problem, because it is a second-best even on my changed preferences. In fact, even if Omar rules that he will re-run the scheduler at the normal time, woh can use the information I have just imparted to unilaterally insure that the scheduler picks the same time on the next run. 
|
|
 IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #136 on: Feb 8th, 2010, 11:05am » |
Quote Modify
|
Woh contacted me offline and we agreed to time slot 70, so all is well!
|
|
IP Logged |
|
|
|
Simon
Forum Guru
Arimaa player #1198
Gender:
Posts: 125
|
|
Re: 2010 World Championship
« Reply #137 on: Feb 8th, 2010, 5:09pm » |
Quote Modify
|
I have unilaterally changed my preferences too. I tried emailing Adanac but got no response, and for all I know (without further information from Adanac) the only reason the game wasn't scheduled on the weekend could be that he has forgotten to update his times from being away the previous weekend.
|
|
IP Logged |
|
|
|
99of9
Forum Guru
Gnobby's creator (player #314)
 
Gender:
Posts: 1413
|
|
Re: 2010 World Championship
« Reply #138 on: Feb 8th, 2010, 6:08pm » |
Quote Modify
|
on Feb 7th, 2010, 6:08pm, omar wrote:Here are the final standings after round 5.
Adanac 8, 1.592, 2319, 14981327
Fritzlein 8, 1.442, 2568, 5740510
chessandgo 8, 1.442, 2561, 85081596
99of9 8, 0.787, 2227, 57429415
Tuks 6, 2.247, 2199, 34425107
The_Jeh 6, 2.034, 1987, 83071905
woh 6, 2.034, 1851, 25689908
Simon 4, 3.558, 1799, 36616487 |
|
It's interesting that the top 6 qualifiers were all in the top 7 last year (and only differ because arimaa_master did not enter this year). Woh only just missed out last year, after placnig 9th in the preliminaries (having been 7th in 200 . So Simon is the lone finals representative of the non WC-veterans. Even in a year where we saw so many newcomers to the server and indeed the tournament.
This suggests a lot more stability than some speculated about in anticipation of the release of the sets and book. I guess it goes to show that arimaa theory already has a significant depth to it, and that even the most brilliant newcomers have to put in the study/hours to competitively take on our brilliant current and past world champions. Personally I'm very glad of the fact that arimaa experience doesn't expire too rapidly . My optimistic aim is to come 4th again this year. If that happens I'll be putting the hard word on those of you who say that humans are still advancing rapidly, because I find it hard to believe that I've advanced much (subconsciously?) since last playing in the 2009 WC.
Nevertheless, it was very refreshing to see so many committed newcomers in the Open Classic this year. I agree with Fritz that the lack of forfeits, and oodles of close games, make it a very respectable tournament, and fun to take part in. Thanks everyone.
|
| « Last Edit: Feb 8th, 2010, 6:27pm by 99of9 » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #139 on: Feb 8th, 2010, 7:50pm » |
Quote Modify
|
on Feb 8th, 2010, 6:08pm, 99of9 wrote:| My optimistic aim is to come 4th again this year. If that happens I'll be putting the hard word on those of you who say that humans are still advancing rapidly [...] |
|
You make a good point about the amount of continuity from last year to this year. I tend to agree that Arimaa has already built up quite a bit of depth that must be assimilated over time by newcomers, such that asking someone who only learned of the game at the time of its release in August to have risen to the top by now is unrealistic. There is too much to learn in six months, which is a good thing for Arimaa's prestige and staying power as a game.
Nevertheless, I will still stake my claim as one of those who says that Arimaa theory is still advancing rapidly, something like 100 Elo points per year. I was less sure of this a month or two ago than I am today, but this tournament has increased my confidence that we are still running forward.
Given the wide spread of ratings at the top of the scale, such a great change in skill need not produce a change in ranks. If the current gameroom ratings are accurate, I could gain 100 points without passing chessandgo, and Adanac could likewise gain 100 points without passing me. Arimaa ratings are not compressed at the top like chess ratings where twenty-five active players are within 100 points of the top-rated player.
When you took fourth last year, you were rated a bit more than 100 points ahead of arimaa_master, the fifth-place finisher then. If you would like to forward a fourth-place finish by yourself as evidence that Arimaa theory has not advanced much in the past year, then you would have to grant that a fifth-place finish by you proves that theory has advanced every bit as much as I claim. How sure are you that you are still ahead of Tuks?
But it is a bit silly to stake so much on the outcome of a single tournament. Last year you lost to arimaa_master in the preliminaries but finished ahead of him in the finals. This year you might finish behind Tuks in the finals after having beaten him in the preliminaries. We will be basing the performance of each of you on approximately nine games; a very small sample.
Consider instead that last year when you played chessandgo in the 2009 finals, you were rated 262 points behind him, whereas presently you are rated 346 points behind. If you want to argue that Arimaa theory hasn't advanced by those 84 rating points, you would have make the case that his rating is artificially high. Yet there are no signs that the system as a whole has undergone rating inflation in the past year, and chessandgo's rating in particular has been established against human opposition, not inflated by bot-bashing. I regard your distance from the top to be at least as reliable an indicator as your placement in the World Championship tournament.
Note that I'm not claiming that I personally am 100 points better than I was last year. I rather think my own pace of improvement has slowed down. But hopefully even as I tiptoe into my forties I can still hang on to the #2 ranking for a while to come. Like you, I'm hoping that I don't instantly become obsolete.
|
|
IP Logged |
|
|
|
99of9
Forum Guru
Gnobby's creator (player #314)
Gender:
Posts: 1413
|
|
Re: 2010 World Championship
« Reply #140 on: Feb 8th, 2010, 9:14pm » |
Quote Modify
|
I don't trust the stability of the ratings anywhere near enough to measure this year upon year. You say there's no evidence of ratings inflation, but is there evidence that there's none? If there's none amongst the top players as a group, then ratings do not show that we've advanced our understanding?? So you sort of require inflation and no-inflation (amongst different populations) to prove your point with ratings.
I agree with you that the sparsity at the top makes it hard to measure anything except the order in a short tournament. However this particularly applies at the very top. To get a good measure of "humanity", it's not good to use Jean as your metric. If your claim is that "Jean has advanced by 100 points of understanding", I'll concede that this may be true, but Magnus Carlsen probably also has/is. Young players are bound to gain while they are active, but at some point they leave the game (e.g. arimaa_master?) or get old, so if we're talking about an overall population effect, it's a very sawtooth/noisy measure if you just use a gap or rating for the top player.
4th position is better, but admittedly not much better. I'd rather use around 8th. It's just that our other non-learning standards (bots) are so far out of the league that we can't use them to measure this at all (and suffer from the predictability-inflation effect).
Regarding your ratings argument about arimaa_master (if we're going to talk ratings): Last year going into the finals there were two humans within the 100 points below me (just), so a straightforward 100 point advance would most likely place me 6th not 5th. Further, the reason my methodology doesn't work when negated is that it's a reasonable assumption to make that I have not improved, but it's not quite as reasonable to assume that I have not gone backward with ten months off.
Regarding Tuks: I have no idea, we'll find out this week . But that's why my plan to give you a hard time is only going to come into effect if I actually achieve 4th! (although since you wanted to start early, I'm happy to indulge  )
|
| « Last Edit: Feb 8th, 2010, 9:18pm by 99of9 » |
IP Logged |
|
|
|
99of9
Forum Guru
Gnobby's creator (player #314)
Gender:
Posts: 1413
|
|
Re: 2010 World Championship
« Reply #141 on: Feb 8th, 2010, 9:28pm » |
Quote Modify
|
on Feb 8th, 2010, 7:50pm, Fritzlein wrote:| Consider instead that last year when you played chessandgo in the 2009 finals, you were rated 262 points behind him, whereas presently you are rated 346 points behind. If you want to argue that Arimaa theory hasn't advanced by those 84 rating points, you would have make the case that his rating is artificially high. |
|
As an example of how unreliable arguments using ratings are, going into the finals last year (the corresponding point) I was rated 329 points behind. My official rating uncertainty is almost 120!
|
|
IP Logged |
|
|
|
Manuel
Forum Guru
Arimaa player #4020
Gender:
Posts: 58
|
|
Re: 2010 World Championship
« Reply #142 on: Feb 8th, 2010, 11:46pm » |
Quote Modify
|
I thought that final games are played on a 90s time scale, but I notice they are scheduled as 60s games. I guess that is a mistake?
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #143 on: Feb 8th, 2010, 11:54pm » |
Quote Modify
|
on Feb 8th, 2010, 9:14pm, 99of9 wrote:| So you sort of require inflation and no-inflation (amongst different populations) to prove your point with ratings. |
|
Yes, the non-inflation is going on with the fixed-performance bots. In fact, the last time I measured, they had deflated significantly because we had nailed down ArimaaScoreP1 to 1000 rating and had started newcomers at 1300. Because of that measured deflation, we bumped up the ratings of newcomers to 1400. I haven't measured again since, but my rough impression is that the fixed-performance bots are now near their 2005 level, i.e. before the inflation/deflation cycle. But this reminds me that the end of 2009 is a good time to repeat my measurement rather than relying on rough impressions.
If it is true that the ratings of fixed-performance bots are relatively stable, then it would suggest that any increased ratings among top players reflect a genuine increase in skill.
Quote:| However this particularly applies at the very top. To get a good measure of "humanity", it's not good to use Jean as your metric. If your claim is that "Jean has advanced by 100 points of understanding", I'll concede that this may be true, but Magnus Carlsen probably also has/is. |
|
The flaw in your analogy is that Jean was already the top-rated Arimaa player before his 100-point advance, whereas Magnus Carlsen was not the top-rated chess player until after his 100-point advance. The rating of the top chess player has made a jagged graph indeed, but slowly rising, and only including the last few points of Carlsen's advance. The rating of the top Arimaa player is probably even more jagged, but nevertheless shows a much more rapid advance, including Jean's huge rise this year.
Quote:| Young players are bound to gain while they are active, but at some point they leave the game (e.g. arimaa_master?) or get old, so if we're talking about an overall population effect, it's a very sawtooth/noisy measure if you just use a gap or rating for the top player. |
|
Yes, I agree it is noisy to just measure the rating of the top player, mostly because of the inaccuracy of ratings.
Quote:| 4th position is better, but admittedly not much better. I'd rather use around 8th. |
|
I'm curious why the skill level of the 8th-best player in the world is a better indicator of the state of the art than the skill level of the best player. By your measure Jean and I could learn arbitrarily much about Arimaa without advancing Arimaa theory at all! I'm not sure why improved skill should only count if it is among players below your own rank.
Quote:| Further, the reason my methodology doesn't work when negated is that it's a reasonable assumption to make that I have not improved, but it's not quite as reasonable to assume that I have not gone backward with ten months off. |
|
Hmmm... Does it happen in science that an experiment that comes out positive strengthens a hypothesis when that same experiment coming out negative doesn't weaken the hypothesis at all? I guess if one can design such one-way experiments, it's the best type to get funded, because they can never fail!
It is telling that you don't mention PMertens' failure to qualify as evidence that the general level of play has risen. Yes, it is a plausible explanation that he is merely rusty, but it is also a plausible explanation that fire is caused by phlogistin moving from the wood into the air.
Rustiness, however, is not a very plausible explanation for why Omar has just had his worst World Championship finish ever, because Omar has played reasonably actively in the past few months. Should we not cast a broader net for evidence than just your own final ranking?
Quote:| Regarding Tuks: I have no idea, we'll find out this week . But that's why my plan to give you a hard time is only going to come into effect if I actually achieve 4th! (although since you wanted to start early, I'm happy to indulge ) |
|
Well, it's good to start giving each other a hard time early, so we can consider various outcomes on an equal footing. What will you say if you take 3rd this year? Will it prove that the rest of us have gotten worse at Arimaa? I would submit that if you came in 3rd you would explain it by saying there is a fair bit of random variation in a short trial, e.g. you got a lucky win over Adanac. (And by the way, I don't take Adanac's #1 seed in the final as much evidence that he has passed up Jean and me in skill.) But why would the "random variation" explanation apply only to a 3rd place finish and not to 4th place or a lower finish?
Quote:| As an example of how unreliable arguments using ratings are, going into the finals last year (the corresponding point) I was rated 329 points behind. My official rating uncertainty is almost 120! |
|
Ah, that's a very good point. Your rating has fluctuated 118 points just within this year's tournament! That effectively blows apart my argument by rating gap between you and Jean. This year you were briefly rated about 500 points below him. I should perhaps have argued instead from the absolute increase in Jean's rating, coupled with an argument that ratings in general are not inflating.
It is always so much easier to pick apart an opposing position than to make a coherent positive case. If I may say so, you have done a much better job of arguing for the meaninglessness of ratings than for the meaningfulness of the placement of a single player (you) in a single tournament.
|
| « Last Edit: Feb 9th, 2010, 12:15am by Fritzlein » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #144 on: Feb 8th, 2010, 11:56pm » |
Quote Modify
|
on Feb 8th, 2010, 11:46pm, Manuel wrote:| I thought that final games are played on a 90s time scale, but I notice they are scheduled as 60s games. I guess that is a mistake? |
|
Nice catch. This is indeed a mistake.
|
|
IP Logged |
|
|
|
Hippo
Forum Guru
Arimaa player #4450
Gender:
Posts: 883
|
|
Re: 2010 World Championship
« Reply #145 on: Feb 9th, 2010, 12:00am » |
Quote Modify
|
The stability was not so obvious I thing. I still hope I could win the game with woh and Nombril could win the last game (playing anybody). But I agree the human x human games training will help much. I started play humans on the end of december ...
|
|
IP Logged |
|
|
|
PMertens
Forum Guru
Arimaa player #692
Gender:
Posts: 437
|
|
Re: 2010 World Championship
« Reply #146 on: Feb 9th, 2010, 7:41am » |
Quote Modify
|
on Feb 8th, 2010, 11:54pm, Fritzlein wrote:
It is telling that you don't mention PMertens' failure to qualify as evidence that the general level of play has risen. Yes, it is a plausible explanation that he is merely rusty, but it is also a plausible explanation that fire is caused by phlogistin moving from the wood into the air. |
|
I wonder what you want to say here
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #147 on: Feb 9th, 2010, 8:22am » |
Quote Modify
|
I guess it complicates my debate with 99of9 that I have a very fuzzy notion of what I mean by "humanity" advancing by 100 points in Arimaa skill. Obviously some players will not have advanced that much, or even have gotten worse due to layoffs. Other will have advanced more, such as Tuks (the most active HvH player) improving by over 200 rating points, and Simon improving by over 300. Some players will have dropped out of the field while other have joined in, making unmeasurable ratings changes.
My main thought in assessing the advancement of "humanity" is our ability to defend the Arimaa Challenge against steady advances in the strength of computer players. For this purpose the most relevant gains are at the very top of the scale. Ultimately it will come down to just one player who is strong enough to beat back the bots, and at that time nothing but his playing level matters. In the mean time, however, it is very important how large the Arimaa community is, and how deeply Arimaa knowledge permeates that community, because Jean could at any moment get tired of Arimaa and drop out forever.
For measuring the depth and strength of the potential future bot-beating community, it does seem relevant that this year's World Championship pool of participant is stronger and more serious than ever. I do grant that if 99of9 finishes fourth this year, just like last year, after having been inactive for a whole year, it is some evidence that the field is not stronger than ever. But likewise if he finishes lower than 4th it would be some evidence that we are making headway. However strong the evidence is in either case is debatable; I would say in both cases it is not as strong as evidence from ratings.
The ratings of bots in general (both fixed performance and variable performance) deflated over the course of 2009, while the ratings of top active humans increased. This is rather indirect evidence given that the best humans lead the best bots by more than a rating class, but it is evidence nonetheless, and I think the most solid evidence we have.
|
| « Last Edit: Feb 9th, 2010, 10:21am by Fritzlein » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #148 on: Feb 9th, 2010, 8:40am » |
Quote Modify
|
on Feb 9th, 2010, 7:41am, PMertens wrote:| I wonder what you want to say here |
|
Apparently you conceded yourself that opening theory had advanced, at least to the extent that you adopted an unbalanced setup. But you were inactive for two years rather than one, which both gave you more time to get rusty and gave more time for the Arimaa community as a whole to move forward. Both explanations are stronger due to your long layoff.
My guess is that even if you were "in form" you would be on the bubble of the top eight qualifiers, making it to the finals more often than missing, but still in that range. Adanac has gotten palpably better than he was when you were trading punch for punch with him, to say nothing of Tuks et. al.
What I'm really hoping for is not that you disappear again and next year come back early enough to get in form, but rather that you take up Arimaa seriously enough to improve to levels above your previous best. I miss your unique and creative approach to Arimaa, and as fun as it is to have you back for two months of the year, it would be far more fun to have you back year-round. That's "what I want to say here".
|
| « Last Edit: Feb 9th, 2010, 8:46am by Fritzlein » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #149 on: Feb 9th, 2010, 10:18am » |
Quote Modify
|
on Feb 8th, 2010, 9:14pm, 99of9 wrote:| [...] so if we're talking about an overall population effect, it's a very sawtooth/noisy measure if you just use a gap or rating for the top player. |
|
That measure, admittedly very noisy, has risen 400 points in the last five years. At the beginning of 2005 you were the top player, rated around 2200, and now Jean is rated around 2600. Of course, even if that noisy number reflects "humanity" having risen by 400 points in playing strength, which is arguable, it doesn't say how much of the 400-point rise occurred in the last six months since the release of Arimaa sets. Furthermore, it suggests the trend is a bit under 100 points per year. But thanks to fixed performance bots, I do think we can at least say ratings in 2010 are about on the same scale they were in 2005.
|
| « Last Edit: Feb 9th, 2010, 10:26am by Fritzlein » |
IP Logged |
|
|
|
|
Home
Help
Search
Members
Login
Register
Reply
Notify of replies
Send Topic
Print