A Critter 1.4 Match To Satisfy My Own Mind!!

[geots]

Critter 1.4 32-bit vs. Strelka 5.1


Seems we have some people on different forums who feel that Strelka is the No. 2 engine now in 1CPU 32bit. I admit that any engine in the top 6 or 7 might, if the moon is in the right position, play Critter right down to the final game to decide the outcome. Shit happens. But I have seen a couple of results where Strelka destroyed Critter. That much shit does not happen. Let me be very clear here: Houdini and Critter at this point in time are just on another planet from the rest.

Hence the running of this match. I will admit that Strelka performed a slight bit better against Critter than I had expected.



1CPU/32bit
128MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


Critter 1.4 32-bit.......+56..........+16/-8/=26........58.00%.......29.0/50

Strelka 5.1................ -56...........+8/-16/=26.......42.00%.......21.0/50



If anyone thinks that after 10 more matches Strelka would show to be much closer to Critter than this- you are more than welcome to your beliefs. I'm not buying it.


Gotta go- my other computer waits,

g
e
o
r
g
e

[Desperado]

Critter 1.4 32-bit vs. Strelka 5.1


Seems we have some people on different forums who feel that Strelka is the No. 2 engine now in 1CPU 32bit. I admit that any engine in the top 6 or 7 might, if the moon is in the right position, play Critter right down to the final game to decide the outcome. Shit happens. But I have seen a couple of results where Strelka destroyed Critter. That much shit does not happen. Let me be very clear here: Houdini and Critter at this point in time are just on another planet from the rest.

Hence the running of this match. I will admit that Strelka performed a slight bit better against Critter than I had expected.



1CPU/32bit
128MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


Critter 1.4 32-bit.......+56..........+16/-8/=26........58.00%.......29.0/50

Strelka 5.1................ -56...........+8/-16/=26.......42.00%.......21.0/50



If anyone thinks that after 10 more matches Strelka would show to be much closer to Critter than this- you are more than welcome to your beliefs. I'm not buying it.


Gotta go- my other computer waits,

g
e
o
r
g
e

Hello, George,

dont get me wrong now, it is easy for me to _believe_ that Critter
is stronger than Strelka (but i dont know it).

To prove it by presenting a match with (only !) 50 games is _clearly_
the wrong way to convince people of your _belief_.

best regards,

Michael

[geots]

Critter 1.4 32-bit vs. Strelka 5.1


Seems we have some people on different forums who feel that Strelka is the No. 2 engine now in 1CPU 32bit. I admit that any engine in the top 6 or 7 might, if the moon is in the right position, play Critter right down to the final game to decide the outcome. Shit happens. But I have seen a couple of results where Strelka destroyed Critter. That much shit does not happen. Let me be very clear here: Houdini and Critter at this point in time are just on another planet from the rest.

Hence the running of this match. I will admit that Strelka performed a slight bit better against Critter than I had expected.



1CPU/32bit
128MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


Critter 1.4 32-bit.......+56..........+16/-8/=26........58.00%.......29.0/50

Strelka 5.1................ -56...........+8/-16/=26.......42.00%.......21.0/50



If anyone thinks that after 10 more matches Strelka would show to be much closer to Critter than this- you are more than welcome to your beliefs. I'm not buying it.


Gotta go- my other computer waits,

g
e
o
r
g
e

Hello, George,

dont get me wrong now, it is easy for me to _believe_ that Critter
is stronger than Strelka (but i dont know it).

To prove it by presenting a match with (only !) 50 games is _clearly_
the wrong way to convince people of your _belief_.

best regards,

Michael

I think you missed my point. In this case, for example, I said I ran the match because I had seen where a match with similar time controls had given the result: 19 - 6 in favor of Strelka. So Strelka beat Critter 1.4 by 13 games in a 50 game match. You bet I will say that did not happen and never will happen. The above result doesn't necessarily tell you even which engine is better, but for me it pretty much seals the deal on my above statement. But I knew that before I ran the match.

I don't think an engine I have run has less than 5- 50 game matches. Most have more. And I still can't be totally sure of much of anything. It's hard to do any really structured testing with a 32bit box and 1CPU. Tho I am as we speak trying to load engines onto a new Win7 64bit quad I recently bought.

But I have never tried to convince anyone of much of anything. I give them the results, at times telling them what I think. In the end they have to decide for themselves if it is worth anything. But I appreciate you responding, and of course what you say is true about "50 games".


Best To You (thanks for the input),

george


PS: I still say Houdini and Critter are on another planet at the moment.

PSS: Any particular engine you would like to see Nemo play in a 50 game match?

[Tomcass]

Hi, George.

My test seem to suggest -obviously not enough games to prove anything- that Strelka 5.1 and Critter 1.4 are substantially even at 2 min + 1 sec/game. Komodo seems much more uncomfortable at this time control. It is better at almost pure incremental time control.

Previous test with Strelka 5.0 and Critter 1.2

20110115Strelka5w64 2011

Strelka 5w64 - Critter 1.2 64-bit SSE4(x1) 39.5 - 35.5 +18/=43/-14 52.67%
Strelka 5w64 - Deep Rybka 4.1 SSE42 x64(x1) 44.5 - 30.5 +25/=39/-11 59.33%
Strelka 5w64 - Houdini 2.0 Pro x64(x1) 36.5 - 38.5 +18/=37/-20 48.67%
Strelka 5w64 - Komodo64 3(x1) 46.5 - 28.5 +28/=37/-10 62.00%


Latest test:


201201Critter Strelka Komodo-1 2012

1 2 3
1 Critter 1.4 64-bit SSE4(x1) 49.0 - 51.058.5 - 41.5** 107.5/200
2 Strelka 5.1(X1) 51.0 - 49.056.0 - 44.0 ** 107.0/200
3 Komodo64 SSE Version 4 41.5 - 58.544.0 - 56.0 ** 85.5/200

Regards from Barcelona.

Tom.

[geots]

Hi, George.

My test seem to suggest -obviously not enough games to prove anything- that Strelka 5.1 and Critter 1.4 are substantially even at 2 min + 1 sec/game. Komodo seems much more uncomfortable at this time control. It is better at almost pure incremental time control.

Previous test with Strelka 5.0 and Critter 1.2

20110115Strelka5w64 2011

Strelka 5w64 - Critter 1.2 64-bit SSE4(x1) 39.5 - 35.5 +18/=43/-14 52.67%
Strelka 5w64 - Deep Rybka 4.1 SSE42 x64(x1) 44.5 - 30.5 +25/=39/-11 59.33%
Strelka 5w64 - Houdini 2.0 Pro x64(x1) 36.5 - 38.5 +18/=37/-20 48.67%
Strelka 5w64 - Komodo64 3(x1) 46.5 - 28.5 +28/=37/-10 62.00%


Latest test:


201201Critter Strelka Komodo-1 2012

1 2 3
1 Critter 1.4 64-bit SSE4(x1) 49.0 - 51.058.5 - 41.5** 107.5/200
2 Strelka 5.1(X1) 51.0 - 49.056.0 - 44.0 ** 107.0/200
3 Komodo64 SSE Version 4 41.5 - 58.544.0 - 56.0 ** 85.5/200

Regards from Barcelona.

Tom.

As far as I can tell, your games are all run with 64bit exes. If so, I cannot comment one way or the other. Except to say that your testing has always been extremely accurate.I will start 64bit testing soon, but I have no games at present. All mine are 32bit.

The only thing I know is that 64bit can vary from a little to a lot different, depending on which one the author gave the most attention to- and how much more.

Based on your past testing I have seen, I will have to go with your results at this time. It will be interesting to see how close we are. It's not my business, but I wish you were running 40/3 repeating- that would be one variable we could throw out.


Thanks and best regards,

george