I can't say anything to explain it. Brent does excellent work- I do excellent work. I really don't see 40/3 repeating and 10'10" enought time diff. to really matter. I doubt "incremental" is the answer for a tight match and a slaughter. It is early in my 64bit testing- but 32 or 64, Strelka has had about 25 matches. The only one anywhere near close was a loss to Critter in a 200 game 32bit match: 103 - 97. Brent always has accurate results- as do I. And I know he would concur with that statement. If I had time, which I may try to find- and can run a 200 game match at 40/20 repeating, probably backed off to adapt to CCRL (meaning a fast machine would be maybe 40/15 repeating), I will run it. I would have no idea, but if Strelka can score 60% or better in that match- it would be case pretty much closed on my end. And that still doesn't mean Brent's are not accurate. But one thing I can promise you- time will clear it all up to everyone's satisfaction. Have patience, and thanks for your comments and interest.
Richard Allbert wrote:50 Games says nothing, really. After 2000 or so the result may be accurate to +/- 10 elo.
Will you be continuing the test?
Regards Richard
Me? Oh yes. With 2 separate 1CPU 32bit rigs, and this 64bit system with 4 true cores, it gives me the flexibility I need. Right now it's all in deciding the best way to maximize my cpus and cores. Which I guess is a good problem to have.
Thanks for stopping and commenting. Your thoughts are always of interest to me.
geots wrote:...I really don't see 40/3 repeating and 10'10" enought time diff. to really matter. ...
Well I would respectfully disagree here and say there is a significant time difference since my machine runs at 3.7 GHz and has ponder=on. Where your control is clearly blitz, I would not say that for mine. From my site: "My machine benchmarks almost twice as fast as the 'standard' 2.4 GHz PC, using Crafty Bench 19.17, so my computer processes the same amount of data in 40/21 as a 'standard' 2.4 GHz PC does in 40/40. The total thinking time each engine of mine will have in a typical 80-move game is 2,800 seconds, or about 47 minutes. If I were to participate in a 40/40 list, my computer's thinking time would be 42 minutes for 80 moves. Forty-two minutes = 2,520 seconds, which is still about 300 seconds less than the 2,800 seconds my engines will think in 80 moves using my time controls."
Either way, the results are interesting. Thanks for all your testing, George.
geots wrote:...I really don't see 40/3 repeating and 10'10" enought time diff. to really matter. ...
Well I would respectfully disagree here and say there is a significant time difference since my machine runs at 3.7 GHz and has ponder=on. Where your control is clearly blitz, I would not say that for mine. From my site: "My machine benchmarks almost twice as fast as the 'standard' 2.4 GHz PC, using Crafty Bench 19.17, so my computer processes the same amount of data in 40/21 as a 'standard' 2.4 GHz PC does in 40/40. The total thinking time each engine of mine will have in a typical 80-move game is 2,800 seconds, or about 47 minutes. If I were to participate in a 40/40 list, my computer's thinking time would be 42 minutes for 80 moves. Forty-two minutes = 2,520 seconds, which is still about 300 seconds less than the 2,800 seconds my engines will think in 80 moves using my time controls."
Either way, the results are interesting. Thanks for all your testing, George.
That is exactly what my new system benchmarks at as well- 40/21 for 40/40. And "ponder on". Not sure how much diff. it makes, if any. But it is an added variable for us, as mine is ponder OFF.
I agree 100%. I love to follow your results. I like your site, and I know I am looking at good results. I am just damn well pissed because I did not take the time to find out about your site until recently. Now I go every day. And I truly thank you for your testing as well. I NEVER get caught up in that "whose results are better" bullshit, because they are both accurate. And accurate is accurate and there is nothing to add. I will admit there are a couple testers whose results I am often skeptical of- but that nowhere near includes you.
It's a hobby that we enjoy, and the friends I have made along the way may be the best part of it.
Longer time controls seem to really help Komodo, but at the faster time controls Houdini, Critter, are Strelka are just better! I while ago I ran a long time control match between Strelka and Komodo, Komodo won. However, I played a round robin tournament a few weeks ago with a time control of sudden death in 3min. Komodo came in dead last behind Houdini, Stelka, Critter, and Vitruvius C & H. The whole family of Ipp* engines are almost unbeatable at blitz when compared to the other strong engines. Interesting!
Yeah, George, I don't see how anyone can claim their results are better than another's. The different controls offer insights to each engine. I think none of us has the whole picture, and we (the testers and the people who view the results) all benefit from having multiple testing bodies.
In regards to the differences between our matches, to put it simply: In an 80-move game, your engines will think for 6 minutes, where mine will think for 47. And that is why I think we see the discrepancy in our tests. Like I said, all worthwhile. None is better, just different, yet still helpful.
Longer time controls seem to really help Komodo, but at the faster time controls Houdini, Critter, are Strelka are just better! I while ago I ran a long time control match between Strelka and Komodo, Komodo won. However, I played a round robin tournament a few weeks ago with a time control of sudden death in 3min. Komodo came in dead last behind Houdini, Stelka, Critter, and Vitruvius C & H. The whole family of Ipp* engines are almost unbeatable at blitz when compared to the other strong engines. Interesting!
I am not sure exactly what diff. people mean when they say long time controls. But let me say this to be very clear on where I stand. I consider 40/20 repeating a long time control. I have never had results that varied more than a slight slight bit from 40/20 to 40/40. And Komodo has not done much ag. the top 3 engines at 40/20. And I have never and will never run even 1 match in my life that starts with 40 moves in the first 2.5 hours. I had rather watch paint dry.
