Don wrote:
Still, the better program has a significant advantage. It's not really correct to say, "may the luckiest program win" - because if you look at the 32 programs that started this tournament and the 4 that are left, is it that much of a shock who are left?
Yes, the better programs have a better chance of winning. Of course, a 2700 elo program thrown in there would almost certainly not win. But you cannot deny that there's a huge part of luck.
Critter eliminated and Quazar qualified should make that clear enough. Critter is about 200 elo stronger than Quazar, according to CCRL 40/40. And that result is relatively precise (see error bars and numbers of games).
Luck can perfectly decide that Komodo, Stockfish, Rybka or Vitrivius will win instead of Houdini, which we know is the strongest.
Yes, there is clearly an element of statistical noise. In fact in Stage 2a Houdini at one point was in a little trouble of not qualifying. A strong finish salvaged that but had the second half gone the same way as the first half not even Houdini would have been there.
The only anomalies in my opinion is Komodo being there instead of Critter. It's not clear to me how strong Vitruvius is so it's hard to comment on that one. It's not that Komodo is not a incredibly strong program but Komodo was playing on 1 core and the entire rest of the field was playing on 16 cores except for Chiron which was removed anyway.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
Don wrote:
It was unexpected to see Critter not get through but it was playing in a tough section. Komodo over-performed considering it was one of the few single core programs and still gets to move on to stage 4. Critter should be there in it's place but such are the fortunes and misfortunes of tournament competition. Other than the Critter anomaly I believe the 4 programs that survive to stage 4 are the ones that deserve to be there.
A shame for Critter. Very nice engine, with incredible tactical abilities. Unfortunately its superior tactics are not very useful at long time control. For example Critter destroys Stockfish when I play games at 10"+0.1" on my machine, but on CCRL 40/40 they are about equal.
But I don't think that's the real explanation. In a tournament with so few games, luck plays an enormous role. It's also what makes the tournaments more fun. Almost any of the finalists could win, given enough luck. If we know that only the strongest can win and we know who the strongest is (Houdini), then it's less fun, and becomes a rating list rather than a tournament.
May the luckiest win
Yes, indeed. I may go back to my old Ludo instead. Chess is just a slightly more ellaborate form of a die game!
As per my last post:
if you look at the 32 programs that started this tournament and the 4 that are left, is it that much of a shock who are left?
Really, I don't think Rybka, Stockfish, Komodo and Houdini finishing in the top 4 is a major shocker nor does it qualify as being the same as picking 4 of the 32 by rolling a dice.
I wasn't thinking about your post, i was only thinking about the words "may the luckiest win", independantly. It was not meant to be taken too seriously.
Yes, i appreaciate the results are quite just, if not 100%.
Yes, I agree. It's far from 100% clear who will finish. You could play this exact tournament 1000 different times and would probably get every one of the 16 qualifying programs in there at least once. And Houdini, clearly the strongest program, would not always appear in the final 4.
There is a saying a tennis player in my old club used to utter (but only when he was winning) which is, "the cream always rises to the top!" Statistically it is true but it might rise too slowly - and if there are not enough games a worthy program may not have time to recover from a couple of unexpectedly bad results.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
Don wrote:There is a saying a tennis player in my old club used to utter (but only when he was winning) which is, "the cream always rises to the top!" Statistically it is true but it might rise too slowly - and if there are not enough games a worthy program may not have time to recover from a couple of unexpectedly bad results.
Exactly, sometimes rises too slowly.
e.g. if it had been stopped after 28 games, Houdini would not have been even in the top 3.
chessmann wrote:It happen second time that Critter has "weak-anomaly" on 8 or more cores,
it first time happened in Graham's >> THE CHAMPIONS 2012 8CPU << tournament.
Just bad luck again ? Or maybe Critter seriously has some problems on high amount of cores / longer time controls?
Critter was eliminated from Stage 2a, 3 cores were used then.
Don wrote:
Still, the better program has a significant advantage. It's not really correct to say, "may the luckiest program win" - because if you look at the 32 programs that started this tournament and the 4 that are left, is it that much of a shock who are left?
Yes, the better programs have a better chance of winning. Of course, a 2700 elo program thrown in there would almost certainly not win. But you cannot deny that there's a huge part of luck.
Critter eliminated and Quazar qualified should make that clear enough. Critter is about 200 elo stronger than Quazar, according to CCRL 40/40. And that result is relatively precise (see error bars and numbers of games).
Luck can perfectly decide that Komodo, Stockfish, Rybka or Vitrivius will win instead of Houdini, which we know is the strongest.
Yes, there is clearly an element of statistical noise. In fact in Stage 2a Houdini at one point was in a little trouble of not qualifying. A strong finish salvaged that but had the second half gone the same way as the first half not even Houdini would have been there.
The only anomalies in my opinion is Komodo being there instead of Critter. It's not clear to me how strong Vitruvius is so it's hard to comment on that one. It's not that Komodo is not a incredibly strong program but Komodo was playing on 1 core and the entire rest of the field was playing on 16 cores except for Chiron which was removed anyway.
To reach the Superfinal, now we have approximately these odds: H3 80%, K5 (if MP), SF, RY 35-45%. For H3 it's harder to reach the Superfinal (80%) than actually to win it (97%), because it plays only 18 games in stage 4, but 48 games in the Superfinal.
