Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0

[Utzinger Kurt]

Geschrieben von:/Posted by: Utzinger Kurt am 01 Februar 2000 23:41:09:
In a rapid 60 match under WinBoard 4.05 running on AMD K6_II 350/64 with 16 MB hash tables for each program and no tablebases I could hardly believe the most surprising result of 9,0 - 1,0 2 draws, 8 wins) in favour of Crafty 1.74 against the latest Phalanx 22 version. On request I can send the games in PGN-format.

[Alexandre Côme]

Geschrieben von:/Posted by: Alexandre Côme am 02 Februar 2000 00:07:27:
Als Antwort auf:/As an answer to: Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0) geschrieben von:/posted by: Utzinger Kurt am 01 Februar 2000 23:41:09:

In a rapid 60 match under WinBoard 4.05 running on AMD K6_II 350/64 with 16 MB hash tables for each program and no tablebases I could hardly believe the most surprising result of 9,0 - 1,0 2 draws, 8 wins) in favour of Crafty 1.74 against the latest Phalanx 22 version. On request I can send the games in PGN-format.

Hello,
It's very hard to believe.
May be something is broken in your setings or....
Best Regards
Alexandre Côme

[Dann Corbit]

Geschrieben von:/Posted by: Dann Corbit am 02 Februar 2000 01:46:03:
Als Antwort auf:/As an answer to: Re: Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0) geschrieben von:/posted by: Alexandre Côme am 02 Februar 2000 00:07:27:

In a rapid 60 match under WinBoard 4.05 running on AMD K6_II 350/64 with 16 MB hash tables for each program and no tablebases I could hardly believe the most surprising result of 9,0 - 1,0 2 draws, 8 wins) in favour of Crafty 1.74 against the latest Phalanx 22 version. On request I can send the games in PGN-format.

Hello,
It's very hard to believe.
May be something is broken in your setings or....

Not at all difficult to imagine. If the chess programs were about equal in strength, this result would not be unexpected. Since crafty is somewhat stronger, it should not be surprising at all. Even a Phalanx 22 victory of 9-1 would not be incredibly unlikely.
Such things happen. That's why we go ahead and play the games.



My FTP site

[Aaron]

Geschrieben von:/Posted by: Aaron at 03 February 2000 05:58:25:
Als Antwort auf:/As an answer to: Re: Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0) geschrieben von:/posted by: Dann Corbit at 02 Februar 2000 01:46:03:

In a rapid 60 match under WinBoard 4.05 running on AMD K6_II 350/64 with 16 MB hash tables for each program and no tablebases I could hardly believe the most surprising result of 9,0 - 1,0 2 draws, 8 wins) in favour of Crafty 1.74 against the latest Phalanx 22 version. On request I can send the games in PGN-format.

Hello,
It's very hard to believe.
May be something is broken in your setings or....

Not at all difficult to imagine. If the chess programs were about equal in >strength, this result would not be unexpected. Since crafty is somewhat >stronger, it should not be surprising at all. Even a Phalanx 22 victory of 9->1 would not be incredibly unlikely.

But the more games the less likely it is right? So a 90-10 Crafty victory over Phalanx would increase the certainty that Crafty is that much stronger?
(In my 5 mins Blitz games, I've played close to 60 games for this matchup and I believe the results are very close with Crafty just ahead by 10 games or so)

Is there a formula to calculate this? I remember someone on CCC saying that a Binomial function would not be accurate since it allows only 2 outcomes Win and Loss ...
Also is there a simple way to figure out ELO differences based on results?
I've taken a basic statsics course, so I probably familar with most terms like chi-square etc..But I just need reminding

[Dann Corbit]

Geschrieben von:/Posted by: Dann Corbit at 04 February 2000 03:19:17:
Als Antwort auf:/As an answer to: Re: Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0) geschrieben von:/posted by: Aaron at 03 February 2000 05:58:25:

In a rapid 60 match under WinBoard 4.05 running on AMD K6_II 350/64 with 16 MB hash tables for each program and no tablebases I could hardly believe the most surprising result of 9,0 - 1,0 2 draws, 8 wins) in favour of Crafty 1.74 against the latest Phalanx 22 version. On request I can send the games in PGN-format.

Hello,
It's very hard to believe.
May be something is broken in your setings or....

Not at all difficult to imagine. If the chess programs were about equal in >strength, this result would not be unexpected. Since crafty is somewhat >stronger, it should not be surprising at all. Even a Phalanx 22 victory of 9->1 would not be incredibly unlikely.

But the more games the less likely it is right? So a 90-10 Crafty victory over Phalanx would increase the certainty that Crafty is that much stronger?
(In my 5 mins Blitz games, I've played close to 60 games for this matchup and I believe the results are very close with Crafty just ahead by 10 games or so)
Is there a formula to calculate this? I remember someone on CCC saying that a Binomial function would not be accurate since it allows only 2 outcomes Win and Loss ...
Also is there a simple way to figure out ELO differences based on results?
I've taken a basic statsics course, so I probably familar with most terms like chi-square etc..But I just need reminding

You can download ELOSTAT and it will calculate the probabilities for you (IIRC).
Just feed it the PGN.


My FTP site

[Kurt Utzinger]

Geschrieben von:/Posted by: Kurt Utzinger at 06 February 2000 16:04:42:
Als Antwort auf:/As an answer to: Crafty174 vs Phalanx22; Rapid 60 AMD K6_II 350/64 (9,0 - 1,0) geschrieben von:/posted by: Utzinger Kurt at 01 Februar 2000 23:41:09:
The match is being repeated with the following changes:
1)
Phalanx22 plays with the usual small opening book. I have the feeling that Phalanx22 did not well with the big opening book.
2)
In the winboard.ini the option "-l+" is deleted.
By the way I am not at all surprised that the tactical monster Phalanx gets much better results against Crafty in blitz games.