Page 22 of 23
Re: TCEC 10
Posted: Wed Dec 06, 2017 1:55 am
by Milos
Guenther wrote:AdminX wrote:1 Houdini 6.03 3185 43.0 79 1548.00 ············· + 13 = 60 - 6
2 Komodo 1970.00 3232 36.0 79 1548.00 + 6 = 60 - 13 ·············
20 Games Remaining
Funny that Komodo has now +3 -0 =8 out of the last 11 games.
It starts mirroring the first quarter result...
Funny how immediately after you comment suggesting Komodo's comeback it loses a game or two or even three in a short period
.
Houdini is statistically better engine and actual chances for Komodo to win even if match lasted 200 games are pretty grim.
Re: TCEC 10
Posted: Wed Dec 06, 2017 9:20 am
by Guenther
Milos wrote:Guenther wrote:AdminX wrote:1 Houdini 6.03 3185 43.0 79 1548.00 ············· + 13 = 60 - 6
2 Komodo 1970.00 3232 36.0 79 1548.00 + 6 = 60 - 13 ·············
20 Games Remaining
Funny that Komodo has now +3 -0 =8 out of the last 11 games.
It starts mirroring the first quarter result...
Funny how immediately after you comment suggesting Komodo's comeback it loses a game or two or even three in a short period :D.
Houdini is statistically better engine and actual chances for Komodo to win even if match lasted 200 games are pretty grim.
Still in your usual mood? Those 'manners' get a bit boring over time,
especially your childish abuse of smilies, which statistically increased
a lot over the last months.
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
Re: TCEC 10
Posted: Wed Dec 06, 2017 10:00 am
by Milos
Guenther wrote:Still in your usual mood? Those 'manners' get a bit boring over time,
especially your childish abuse of smilies, which statistically increased
a lot over the last months.
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
Always same bla, bla, no substance. It's clear you are approaching argument like a fan and not from a rational point of view. So I'll leave you to cheer.
Re: TCEC 10
Posted: Wed Dec 06, 2017 12:32 pm
by sovaz1997
There is not much difference in statistics between Houdini and Komodo:
http://www.sp-cc.de/long-thinkingtime.htm:
Houdini-Komodo: 741 (+118,=529,- 94)
Re: TCEC 10
Posted: Wed Dec 06, 2017 5:05 pm
by kranium
Seems very likely Houdini is stronger...
games = 741
win% = 0.516194
elo = +11.2569
los = 0.945045
Re: TCEC 10
Posted: Wed Dec 06, 2017 6:11 pm
by kranium
Guenther wrote:
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
You don't need a large # of games if the margin of victory is big enough...
http://www.computerchess.org.uk/ccrl/40 ... 4-bit_4CPU
Houdini 6 vs Komodo 11.2
(+10−1=46)
games = 57
win% = 0.578947
elo = +55.3211
los = 0.994617
Re: TCEC 10
Posted: Wed Dec 06, 2017 6:22 pm
by Guenther
kranium wrote:Guenther wrote:
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
You don't need a large # of games if the margin of victory is big enough...
http://www.computerchess.org.uk/ccrl/40 ... 4-bit_4CPU
Houdini 6 vs Komodo 11.2
(+10−1=46)
games = 57
win% = 0.578947
elo = +55.3211
los = 0.994617
This is cherry picking a 57 games interim match result.
Why not posting the full results table with other games included?
Code: Select all
2 Houdini 6 64-bit 4CPU 3412 +22 -22 69.7% -128.3 54.5% 664
3 Komodo 11.2 64-bit 4CPU 3398 +19 -19 67.4% -113.9 53.8% 864
Nothing to conclude with that error bars - we still know nothing and
this is just for 4 cpus and not TCEC hardware...
Re: TCEC 10
Posted: Wed Dec 06, 2017 6:45 pm
by kranium
Guenther wrote:kranium wrote:Guenther wrote:
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
You don't need a large # of games if the margin of victory is big enough...
http://www.computerchess.org.uk/ccrl/40 ... 4-bit_4CPU
Houdini 6 vs Komodo 11.2
(+10−1=46)
games = 57
win% = 0.578947
elo = +55.3211
los = 0.994617
This is cherry picking a 57 games interim match result.
Why not posting the full results table with other games included?
Code: Select all
2 Houdini 6 64-bit 4CPU 3412 +22 -22 69.7% -128.3 54.5% 664
3 Komodo 11.2 64-bit 4CPU 3398 +19 -19 67.4% -113.9 53.8% 864
Nothing to conclude with that error bars - we still know nothing and
this is just for 4 cpus and not TCEC hardware...
The point was LOS calculation, which can be applied to 2 engines only
Weren't you talking about Houdini vs Komodo, like now in TCEC...?
If so, there's plenty of data like the 2 examples I posted above.
http://www.cegt.net/40_40%20Rating%20Li ... ion/8.html
Houdini 6.0 x64 1CPU -3360 vs Komodo 11.2 x64 1CPU - 3332
+ 22 = 66 - 12
games = 100
win% = 0.55
elo = +34.8601
los = 0.950873
Re: TCEC 10
Posted: Wed Dec 06, 2017 8:43 pm
by corres
[quote="Guenther"]
There is no statistical evidence that Houdini or Komodo is better.
No one so far produced a significant amount of games that would proof one of them being better.
[/quote]
You are right.
But the one question is "What is the better engine?" and the other question is
"What engine may be the winner of TCEC?"
To give the right answer to the question one we would get lot of thousand games.
To give the right answer to the second question we have enough games yet.
In my opinion there is a small difference between Houidini and Kimodo in Elo number but the advantage is in the side of Houdini.
Re: TCEC 10
Posted: Thu Dec 07, 2017 9:08 am
by AdminX
Congratulations Robert on winning the TCEC Season 10.