Thanks again! It seems GnuChess sits nicely on its own evolutionary branch,
BTW: My branch of GnuChess suports UCI so WB2UCI should normally not be necessary.
Pairwise Analysis of Chess Engine Move Selections
Moderator: Ras
-
Michel
- Posts: 2292
- Joined: Mon Sep 29, 2008 1:50 am
-
Adam Hair
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: Pairwise Analysis of Chess Engine Move Selections
That is what I get for assuming instead of reading or typing uci at the prompt.
-
Adam Hair
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: Pairwise Analysis of Chess Engine Move Selections
I have managed include more engines for comparison. Here is a quick analysis of 83 engines where no assumptions of relatedness are made.
This skews the data some, due to the fact we know some engines are related to Fruit. But, I thought it would be interesting to drop all assumptions and see what the data shows. More in-depth analysis and trees will be included in the future. And hopefully more engines.
Here are the 83 engines:
This produces 3403 data points. The distribution of the data can be approximated by a normal distribution, but Student's T is a better fit, so I am using it instead. The Student's T curve that best fits the data has these parameters:
degrees of freedom : 3.904
location : 46.293
scale : 2.542
Here is some of the data. The last column, "Percentage", gives the probability of a given data point occuring ( given the possible distribution of the data).
This skews the data some, due to the fact we know some engines are related to Fruit. But, I thought it would be interesting to drop all assumptions and see what the data shows. More in-depth analysis and trees will be included in the future. And hopefully more engines.
Here are the 83 engines:
Code: Select all
1) Alaric 707 (time: 355 ms scale: 1.0)
2) Alfil 8.1.1 Optimized (time: 485 ms scale: 1.0)
3) Arasan 12.3 (time: 761 ms scale: 1.0)
4) Aristarch 4.50 (time: 686 ms scale: 1.0)
5) Baron 2.23 (time: 816 ms scale: 1.0)
6) Bison 9.11 (time: 288 ms scale: 1.0)
7) Bobcat 20110220 (time: 577 ms scale: 1.0)
8) Booot 5.1.0 (time: 113 ms scale: 1.0)
9) Bright 0.5c (time: 166 ms scale: 1.0)
10) Cerebro 3.03d (time: 1236 ms scale: 1.0)
11) ChessTiger 2007 (time: 288 ms scale: 1.0)
12) Colossus 2008b (time: 408 ms scale: 1.0)
13) Crafty 23.4 (time: 149 ms scale: 1.0)
14) Critter 1.0 (time: 21 ms scale: 1.0)
15) CuckooChess 1.10 (time: 1004 ms scale: 1.0)
16) Cyrano 0.6b17 (time: 422 ms scale: 1.0)
17) Daydreamer 1.75 (time: 343 ms scale: 1.0)
18) Shredder 11 (time: 102 ms scale: 1.0)
19) Delfi 5.4 (time: 320 ms scale: 1.0)
20) Dirty24APR2011 (time: 597 ms scale: 1.0)
21) Dragon 4.6 (time: 1940 ms scale: 1.0)
22) Et Chess 130108 (time: 408 ms scale: 1.0)
23) Frenzee Feb 08 (time: 299 ms scale: 1.0)
24) Fruit 2.1 (time: 381 ms scale: 1.0)
25) Gaia 3.5 (time: 1689 ms scale: 1.0)
26) Gandalf 6.0 (time: 485 ms scale: 1.0)
27) GarboChess 2.20 (time: 538 ms scale: 1.0)
28) Gaviota 0.83 (time: 577 ms scale: 1.0)
29) Glass 1.6 (time: 1420 ms scale: 1.0)
30) Glaurung 1.2.1 (time: 381 ms scale: 1.0)
31) GnuChess 5.07 1705b TCEC (time: 597 ms scale: 1.0)
32) Greko 8.0 (time: 1874 ms scale: 1.0)
33) Gull 1.2 (time: 59 ms scale: 1.0)
34) Hamsters 0.7.1 (time: 618 ms scale: 1.0)
35) Hannibal 1.0a (time: 89 ms scale: 1.0)
36) Hermann 2.6 (time: 970 ms scale: 1.0)
37) HIARCS 12 (time: 113 ms scale: 1.0)
38) Houdini 1.5 (time: 10 ms scale: 1.0)
39) Jonny 4.00 (time: 209 ms scale: 1.0)
40) Junior 10.1 (time: 211 ms scale: 1.0)
41) Komodo 1.3 (time: 39 ms scale: 1.0)
42) Ktulu 8 (time: 234 ms scale: 1.0)
43) LambChop 10.99 (time: 1749 ms scale: 1.0)
44) List 512 (time: 577 ms scale: 1.0)
45) Little Goliath Evolution (time: 1004 ms scale: 1.0)
46) Loop 2007 (time: 166 ms scale: 1.0)
47) Movei 00.8.438 (time: 453 ms scale: 1.0)
48) N2 0.4 (time: 597 ms scale: 1.0)
49) Naraku 1.31 (time: 381 ms scale: 1.0)
50) Naum 2.0 (time: 279 ms scale: 1.0)
51) Naum 4.2 (time: 40 ms scale: 1.0)
52) Nejmet 3.07 (time: 1874 ms scale: 1.0)
53) Onno 1.0.4 (time: 113 ms scale: 1.0)
54) Pepito 1.59.2 (time: 1200 ms scale: 1.0)
55) Petir_4.999999 (time: 686 ms scale: 1.0)
56) Pharaon 3.5.1 (time: 577 ms scale: 1.0)
57) Philou 3.5.1 (time: 874 ms scale: 1.0)
58) Protector 1.4.0 (time: 77 ms scale: 1.0)
59) Pseudo07c (time: 874 ms scale: 1.0)
60) Pupsi2 v0.08 (time: 1004 ms scale: 1.0)
61) RedQueen 0.98 (time: 577 ms scale: 1.0)
62) RobboLito 0.085g3 (time: 19 ms scale: 1.0)
63) Rotor 0.6 (time: 640 ms scale: 1.0)
64) Ruffian 2.1.0 (time: 502 ms scale: 1.0)
65) Rybka 4 (time: 18 ms scale: 1.0)
66) Sjeng WC2008 (time: 109 ms scale: 1.0)
67) SlowChess Blitz WV 2.1 (time: 502 ms scale: 1.0)
68) SmarThink 1.20 (time: 209 ms scale: 1.0)
69) SOS 5.1 (time: 640 ms scale: 1.0)
70) Spark 1.0 (time: 61 ms scale: 1.0)
71) Spike 1.4 (time: 65 ms scale: 1.0)
72) Stockfish 2.1 (time: 18 ms scale: 1.0)
73) Strelka 2.0 B (time: 288 ms scale: 1.0)
74) Tao 5.6 (time: 1280 ms scale: 1.0)
75) The King 3.50 (time: 355 ms scale: 1.0)
76) Thinker 5.4d Inert (time: 80 ms scale: 1.0)
77) Tornado 4.4 (time: 243 ms scale: 1.0)
78) Twisted Logic 20100131x (time: 178 ms scale: 1.0)
79) Ufim 8.02 (time: 844 ms scale: 1.0)
80) Umko 0.9 (time: 178 ms scale: 1.0)
81) WildCat 8 (time: 437 ms scale: 1.0)
82) Yace 9987 (time: 1040 ms scale: 1.0)
83) Zappa Mexico II (time: 102 ms scale: 1.0)degrees of freedom : 3.904
location : 46.293
scale : 2.542
Here is some of the data. The last column, "Percentage", gives the probability of a given data point occuring ( given the possible distribution of the data).
Code: Select all
% Matching Moves t(alpha,d.f.) Percentage
Fruit 2.1 Loop 2007 71.13 19.31 0.03036
Loop 2007 Onno 1.0.4 67.88 16.78 0.04609
Fruit 2.1 Onno 1.0.4 67.83 16.74 0.04641
Strelka 2.0 B Thinker 5.4d Inert 62.07 12.26 0.11678
Houdini 1.5 RobboLito 0.085g3 61.87 12.11 0.12126
Naum 4.2 Strelka 2.0 B 61.37 11.72 0.13351
Hannibal 1.0a Twisted Logic 20100131x 60.74 11.23 0.15140
Philou 3.5.1 Stockfish 2.1 60.43 10.99 0.16138
Loop 2007 Umko 0.9 60.33 10.91 0.16479
Alaric 707 Fruit 2.1 60.28 10.87 0.16653
Alaric 707 Loop 2007 59.88 10.56 0.18135
Colossus 2008b Fruit 2.1 59.69 10.41 0.18901
Fruit 2.1 Umko 0.9 59.42 10.20 0.20064
Onno 1.0.4 Umko 0.9 59.38 10.17 0.20244
Alfil 8.1.1 Optimized Glaurung 1.2.1 59.07 9.93 0.21718
Critter 1.0 RobboLito 0.085g3 59.04 9.91 0.21868
Alaric 707 Onno 1.0.4 58.49 9.48 0.24880
Naum 4.2 Thinker 5.4d Inert 58.40 9.41 0.25424
Colossus 2008b Loop 2007 58.23 9.28 0.26496
Fruit 2.1 Strelka 2.0 B 58.18 9.24 0.26822
Delfi 5.4 Loop 2007 57.65 8.83 0.30636
Loop 2007 Naraku 1.31 57.65 8.83 0.30636
Delfi 5.4 Fruit 2.1 57.50 8.71 0.31845
Critter 1.0 Houdini 1.5 57.40 8.63 0.32685
Colossus 2008b Onno 1.0.4 57.37 8.61 0.32943
Loop 2007 Strelka 2.0 B 57.11 8.41 0.35296
Onno 1.0.4 Strelka 2.0 B 57.10 8.40 0.35391
Fruit 2.1 Naraku 1.31 57.00 8.32 0.36358
Alaric 707 Rotor 0.6 56.68 8.07 0.39699
Colossus 2008b Naraku 1.31 56.68 8.07 0.39699
GarboChess 2.20 Strelka 2.0 B 56.68 8.07 0.39699
Alaric 707 Strelka 2.0 B 56.65 8.05 0.40033
Naraku 1.31 Onno 1.0.4 56.57 7.99 0.40941
Delfi 5.4 Onno 1.0.4 56.48 7.92 0.41996
Fruit 2.1 N2 0.4 56.05 7.58 0.47561
Hamsters 0.7.1 Loop 2007 56.02 7.56 0.47984
Fruit 2.1 SmarThink 1.20 55.89 7.46 0.49879
Loop 2007 Rotor 0.6 55.60 7.23 0.54476
Fruit 2.1 Rotor 0.6 55.49 7.15 0.56366
Fruit 2.1 Hamsters 0.7.1 55.38 7.06 0.58343
Loop 2007 SmarThink 1.20 55.33 7.02 0.59272
Naraku 1.31 Umko 0.9 55.32 7.02 0.59460
Colossus 2008b Umko 0.9 55.26 6.97 0.60605
Fruit 2.1 Naum 4.2 55.22 6.94 0.61385
N2 0.4 Onno 1.0.4 55.01 6.78 0.65703
Hamsters 0.7.1 Rotor 0.6 54.98 6.75 0.66352
Loop 2007 Naum 4.2 54.98 6.75 0.66352
RobboLito 0.085g3 Rybka 4 54.98 6.75 0.66352
Loop 2007 N2 0.4 54.96 6.74 0.66789
-
Sven
- Posts: 4052
- Joined: Thu May 15, 2008 9:57 pm
- Location: Berlin, Germany
- Full name: Sven Schüle
Re: Pairwise Analysis of Chess Engine Move Selections
Code: Select all
% Matching Moves t(alpha,d.f.) Percentage
Fruit 2.1 Loop 2007 71.13 19.31 0.03036
Loop 2007 Onno 1.0.4 67.88 16.78 0.04609
Fruit 2.1 Onno 1.0.4 67.83 16.74 0.04641
Strelka 2.0 B Thinker 5.4d Inert 62.07 12.26 0.11678
Houdini 1.5 RobboLito 0.085g3 61.87 12.11 0.12126
Naum 4.2 Strelka 2.0 B 61.37 11.72 0.13351
Hannibal 1.0a Twisted Logic 20100131x 60.74 11.23 0.15140
Philou 3.5.1 Stockfish 2.1 60.43 10.99 0.16138
Loop 2007 Umko 0.9 60.33 10.91 0.16479
Alaric 707 Fruit 2.1 60.28 10.87 0.16653
Alaric 707 Loop 2007 59.88 10.56 0.18135
Colossus 2008b Fruit 2.1 59.69 10.41 0.18901
Fruit 2.1 Umko 0.9 59.42 10.20 0.20064
Onno 1.0.4 Umko 0.9 59.38 10.17 0.20244
Alfil 8.1.1 Optimized Glaurung 1.2.1 59.07 9.93 0.21718
Critter 1.0 RobboLito 0.085g3 59.04 9.91 0.21868
Alaric 707 Onno 1.0.4 58.49 9.48 0.24880
Naum 4.2 Thinker 5.4d Inert 58.40 9.41 0.25424
Colossus 2008b Loop 2007 58.23 9.28 0.26496
Fruit 2.1 Strelka 2.0 B 58.18 9.24 0.26822
Delfi 5.4 Loop 2007 57.65 8.83 0.30636
Loop 2007 Naraku 1.31 57.65 8.83 0.30636
Delfi 5.4 Fruit 2.1 57.50 8.71 0.31845
Critter 1.0 Houdini 1.5 57.40 8.63 0.32685
Colossus 2008b Onno 1.0.4 57.37 8.61 0.32943
Loop 2007 Strelka 2.0 B 57.11 8.41 0.35296
Onno 1.0.4 Strelka 2.0 B 57.10 8.40 0.35391
Fruit 2.1 Naraku 1.31 57.00 8.32 0.36358
Alaric 707 Rotor 0.6 56.68 8.07 0.39699
Colossus 2008b Naraku 1.31 56.68 8.07 0.39699
GarboChess 2.20 Strelka 2.0 B 56.68 8.07 0.39699
Alaric 707 Strelka 2.0 B 56.65 8.05 0.40033
Naraku 1.31 Onno 1.0.4 56.57 7.99 0.40941
Delfi 5.4 Onno 1.0.4 56.48 7.92 0.41996
Fruit 2.1 N2 0.4 56.05 7.58 0.47561
Hamsters 0.7.1 Loop 2007 56.02 7.56 0.47984
Fruit 2.1 SmarThink 1.20 55.89 7.46 0.49879
Loop 2007 Rotor 0.6 55.60 7.23 0.54476
Fruit 2.1 Rotor 0.6 55.49 7.15 0.56366
Fruit 2.1 Hamsters 0.7.1 55.38 7.06 0.58343
Loop 2007 SmarThink 1.20 55.33 7.02 0.59272
Naraku 1.31 Umko 0.9 55.32 7.02 0.59460
Colossus 2008b Umko 0.9 55.26 6.97 0.60605
Fruit 2.1 Naum 4.2 55.22 6.94 0.61385
N2 0.4 Onno 1.0.4 55.01 6.78 0.65703
Hamsters 0.7.1 Rotor 0.6 54.98 6.75 0.66352
Loop 2007 Naum 4.2 54.98 6.75 0.66352
RobboLito 0.085g3 Rybka 4 54.98 6.75 0.66352
Loop 2007 N2 0.4 54.96 6.74 0.66789Sven
-
Graham Banks
- Posts: 44012
- Joined: Sun Feb 26, 2006 10:52 am
- Location: Auckland, NZ
Re: Pairwise Analysis of Chess Engine Move Selections
Any chance that you could post a list that includes all done to date?Adam Hair wrote:I have managed include more engines for comparison. Here is a quick analysis of 83 engines where no assumptions of relatedness are made.
This skews the data some, due to the fact we know some engines are related to Fruit. But, I thought it would be interesting to drop all assumptions and see what the data shows. More in-depth analysis and trees will be included in the future. And hopefully more engines.
gbanksnz at gmail.com
-
Adam Hair
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: Pairwise Analysis of Chess Engine Move Selections
Yes. Though, many of the other engines I have tested are actually earlier versions of engines listed above.
My main focus at this point is to see if every group of engines that tend to choose similar moves at a higher rate include an open source engine that preceeded the other engines.
My main focus at this point is to see if every group of engines that tend to choose similar moves at a higher rate include an open source engine that preceeded the other engines.
-
Gian-Carlo Pascutto
- Posts: 1260
- Joined: Sat Dec 13, 2008 7:00 pm
Re: Pairwise Analysis of Chess Engine Move Selections
Is the percentage (really: probability) in the last column corrected for the fact that you did 3403 simultaneous comparisons instead of 1? If not, the number is wrong and tremendously overstating the real probability.Adam Hair wrote: This produces 3403 data points. The last column, "Percentage", gives the probability of a given data point occuring ( given the possible distribution of the data).
-
Gian-Carlo Pascutto
- Posts: 1260
- Joined: Sat Dec 13, 2008 7:00 pm
Re: Pairwise Analysis of Chess Engine Move Selections
Nevermind this - you used student-t but didn't actually do a pairwise comparison, you just used that as the distribution you guessed, so this isn't directly relevant to the quoted percentage.
I'm not sure where the 3403 data points come from, but you probably still suffer from the same flaw wrt false positives just due to the sheer amount of possible pairs. If you pick 3403 samples from a distribution, you'll expect about 170 of them to have a 5% or less probability of occurring even if there is nothing particular about them. This should be taken into account when interpreting those percentages.
The validity of your probability also hinges on the question if the distribution is correct and how accurate the estimation of it is. You found that student-t is a better fit than the normal distribution, despite there being no particular reason why this data shouldn't have been normally distributed. Note that student-t differs from the normal distribution in that it has bigger sidelobes (more outliers). So another explanation is: not all data points included are actually from the same, "null hypothesis" distribution of engines that are unrelated to each other. You mentioned this in your text, but it obviously affects what distribution is going to fit your data best. Quite possibly, if you combine the distribution of unrelated engines with that of engines with a common origin, your best fit will actually be a bimodal distribution!
So basically, I'd take the "percentage" numbers with a few cups of salt until a more rigorous analysis is done.
I'm not sure where the 3403 data points come from, but you probably still suffer from the same flaw wrt false positives just due to the sheer amount of possible pairs. If you pick 3403 samples from a distribution, you'll expect about 170 of them to have a 5% or less probability of occurring even if there is nothing particular about them. This should be taken into account when interpreting those percentages.
The validity of your probability also hinges on the question if the distribution is correct and how accurate the estimation of it is. You found that student-t is a better fit than the normal distribution, despite there being no particular reason why this data shouldn't have been normally distributed. Note that student-t differs from the normal distribution in that it has bigger sidelobes (more outliers). So another explanation is: not all data points included are actually from the same, "null hypothesis" distribution of engines that are unrelated to each other. You mentioned this in your text, but it obviously affects what distribution is going to fit your data best. Quite possibly, if you combine the distribution of unrelated engines with that of engines with a common origin, your best fit will actually be a bimodal distribution!
So basically, I'd take the "percentage" numbers with a few cups of salt until a more rigorous analysis is done.
-
Adam Hair
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: Pairwise Analysis of Chess Engine Move Selections
Hi Gian-Carlo,
Thanks for responding to this thread. Any valid criticism is welcomed.
refers to the comparison of the move selections of pairs of engines.
For 83 engines, the number of pairs is 82*83/2 = 3403. The actual
data is derived from the number of moves each pair of engines chose
in common when using the similarity tester. I used the website
http://zunzun.com/ to find the statistical distribution that best fit the data.
The second best fit to the data ( using the negative log likelihood to judge
between models ) was a Student's T distribution with the parameters I
gave above. The best fit was labelled Johnson SU, which I have never
heard of and involved more parameters and was only marginally a better
fit. A normal distribution with mean = 46.310 and st.dev = 3.538 was
14th best. The percentage in the last column was derived from using
the Student's T distribution with the given parameters. One thing that I
left out is that the percentages calculated were for both tails, not just the
upper tail ( the pairs that fall in the lower tail involve Sjeng and Junior).
So, since we might be more interested in the upper tail, those percentages
should be divided by 2. Hence, the probability that two engines would
match 71.13% of their moves is 0.01518%. Assuming, of course, that
this is an accurate model for the population of engines. I am not ready to
make any assumptions like that.
And the highest percentage in the table above is 0.66789%, which should
have been 0.33395%. Out of 3403 samples, we would expect to find 11
to 12 occuring in the upper 0.33395% tail, whereas there are 49 pairs
listed above. All of which is irrelevant if the actual distribution for the
population of engines differs from this estimate.
agree with you. More rigorous analysis is needed. That analysis is waiting
on the accumulation of more data.
Thanks for the comments,
Adam
Thanks for responding to this thread. Any valid criticism is welcomed.
I should have been more explicit. The use of "pair-wise" in the titleGian-Carlo Pascutto wrote:Nevermind this - you used student-t but didn't actually do a pairwise comparison, you just used that as the distribution you guessed, so this isn't directly relevant to the quoted percentage.
refers to the comparison of the move selections of pairs of engines.
For 83 engines, the number of pairs is 82*83/2 = 3403. The actual
data is derived from the number of moves each pair of engines chose
in common when using the similarity tester. I used the website
http://zunzun.com/ to find the statistical distribution that best fit the data.
The second best fit to the data ( using the negative log likelihood to judge
between models ) was a Student's T distribution with the parameters I
gave above. The best fit was labelled Johnson SU, which I have never
heard of and involved more parameters and was only marginally a better
fit. A normal distribution with mean = 46.310 and st.dev = 3.538 was
14th best. The percentage in the last column was derived from using
the Student's T distribution with the given parameters. One thing that I
left out is that the percentages calculated were for both tails, not just the
upper tail ( the pairs that fall in the lower tail involve Sjeng and Junior).
So, since we might be more interested in the upper tail, those percentages
should be divided by 2. Hence, the probability that two engines would
match 71.13% of their moves is 0.01518%. Assuming, of course, that
this is an accurate model for the population of engines. I am not ready to
make any assumptions like that.
You are right. I would like to note that 536 pairs fell in the upper 5% tail.Gian-Carlo Pascutto wrote: I'm not sure where the 3403 data points come from, but you probably still suffer from the same flaw wrt false positives just due to the sheer amount of possible pairs. If you pick 3403 samples from a distribution, you'll expect about 170 of them to have a 5% or less probability of occurring even if there is nothing particular about them. This should be taken into account when interpreting those percentages.
And the highest percentage in the table above is 0.66789%, which should
have been 0.33395%. Out of 3403 samples, we would expect to find 11
to 12 occuring in the upper 0.33395% tail, whereas there are 49 pairs
listed above. All of which is irrelevant if the actual distribution for the
population of engines differs from this estimate.
I agree with you 100%.Gian-Carlo Pascutto wrote: The validity of your probability also hinges on the question if the distribution is correct and how accurate the estimation of it is. You found that student-t is a better fit than the normal distribution, despite there being no particular reason why this data shouldn't have been normally distributed. Note that student-t differs from the normal distribution in that it has bigger sidelobes (more outliers). So another explanation is: not all data points included are actually from the same, "null hypothesis" distribution of engines that are unrelated to each other. You mentioned this in your text, but it obviously affects what distribution is going to fit your data best. Quite possibly, if you combine the distribution of unrelated engines with that of engines with a common origin, your best fit will actually be a bimodal distribution!
I would probably take a couple of cups of salt less than you, but I doGian-Carlo Pascutto wrote: So basically, I'd take the "percentage" numbers with a few cups of salt until a more rigorous analysis is done.
agree with you. More rigorous analysis is needed. That analysis is waiting
on the accumulation of more data.
Thanks for the comments,
Adam
-
Adam Hair
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: Pairwise Analysis of Chess Engine Move Selections
Sven, could I ask you how you did that?Sven Schüle wrote:Just for even more readabilityCode: Select all
% Matching Moves t(alpha,d.f.) Percentage Fruit 2.1 Loop 2007 71.13 19.31 0.03036 Loop 2007 Onno 1.0.4 67.88 16.78 0.04609 Fruit 2.1 Onno 1.0.4 67.83 16.74 0.04641 Strelka 2.0 B Thinker 5.4d Inert 62.07 12.26 0.11678 Houdini 1.5 RobboLito 0.085g3 61.87 12.11 0.12126 Naum 4.2 Strelka 2.0 B 61.37 11.72 0.13351 Hannibal 1.0a Twisted Logic 20100131x 60.74 11.23 0.15140 Philou 3.5.1 Stockfish 2.1 60.43 10.99 0.16138 Loop 2007 Umko 0.9 60.33 10.91 0.16479 Alaric 707 Fruit 2.1 60.28 10.87 0.16653 Alaric 707 Loop 2007 59.88 10.56 0.18135 Colossus 2008b Fruit 2.1 59.69 10.41 0.18901 Fruit 2.1 Umko 0.9 59.42 10.20 0.20064 Onno 1.0.4 Umko 0.9 59.38 10.17 0.20244 Alfil 8.1.1 Optimized Glaurung 1.2.1 59.07 9.93 0.21718 Critter 1.0 RobboLito 0.085g3 59.04 9.91 0.21868 Alaric 707 Onno 1.0.4 58.49 9.48 0.24880 Naum 4.2 Thinker 5.4d Inert 58.40 9.41 0.25424 Colossus 2008b Loop 2007 58.23 9.28 0.26496 Fruit 2.1 Strelka 2.0 B 58.18 9.24 0.26822 Delfi 5.4 Loop 2007 57.65 8.83 0.30636 Loop 2007 Naraku 1.31 57.65 8.83 0.30636 Delfi 5.4 Fruit 2.1 57.50 8.71 0.31845 Critter 1.0 Houdini 1.5 57.40 8.63 0.32685 Colossus 2008b Onno 1.0.4 57.37 8.61 0.32943 Loop 2007 Strelka 2.0 B 57.11 8.41 0.35296 Onno 1.0.4 Strelka 2.0 B 57.10 8.40 0.35391 Fruit 2.1 Naraku 1.31 57.00 8.32 0.36358 Alaric 707 Rotor 0.6 56.68 8.07 0.39699 Colossus 2008b Naraku 1.31 56.68 8.07 0.39699 GarboChess 2.20 Strelka 2.0 B 56.68 8.07 0.39699 Alaric 707 Strelka 2.0 B 56.65 8.05 0.40033 Naraku 1.31 Onno 1.0.4 56.57 7.99 0.40941 Delfi 5.4 Onno 1.0.4 56.48 7.92 0.41996 Fruit 2.1 N2 0.4 56.05 7.58 0.47561 Hamsters 0.7.1 Loop 2007 56.02 7.56 0.47984 Fruit 2.1 SmarThink 1.20 55.89 7.46 0.49879 Loop 2007 Rotor 0.6 55.60 7.23 0.54476 Fruit 2.1 Rotor 0.6 55.49 7.15 0.56366 Fruit 2.1 Hamsters 0.7.1 55.38 7.06 0.58343 Loop 2007 SmarThink 1.20 55.33 7.02 0.59272 Naraku 1.31 Umko 0.9 55.32 7.02 0.59460 Colossus 2008b Umko 0.9 55.26 6.97 0.60605 Fruit 2.1 Naum 4.2 55.22 6.94 0.61385 N2 0.4 Onno 1.0.4 55.01 6.78 0.65703 Hamsters 0.7.1 Rotor 0.6 54.98 6.75 0.66352 Loop 2007 Naum 4.2 54.98 6.75 0.66352 RobboLito 0.085g3 Rybka 4 54.98 6.75 0.66352 Loop 2007 N2 0.4 54.96 6.74 0.66789
Sven