Houdini 4 still better than Stockfish at very fast time :(

[felix]

Yes

Just to make the math, a 100 move game take 100*.1 = 10seconds + 5 seconds = 15 seconds *2 = 30 seconds each game.

I know is a very short time. But the engines play very good.

I remember four months ago where I was testing stockfish with Houdini 1.5 at 40/4 too!

Now is not necessary such a long time control

[Nay Lin Tun]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

[felix]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

For me is very useful.

I have faith. If Houdini can, Stockfish will do better some day.

[gerold]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

For me is very useful.

I have faith. If Houdini can, Stockfish will do better some day.

Thanks for the info. I test 32 verison at 5min+3 sec. and get about the same score.

[felix]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

For me is very useful.

I have faith. If Houdini can, Stockfish will do better some day.

Thanks for the info. I test 32 verison at 5min+3 sec. and get about the same score.

You got the same results at 5 min + 3 sec and 5 sec + 100ms?

[gerold]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

For me is very useful.

I have faith. If Houdini can, Stockfish will do better some day.

Thanks for the info. I test 32 verison at 5min+3 sec. and get about the same score.

You got the same results at 5 min + 3 sec and 5 sec + 100ms?

I test at 5 min + 3 sec. Houdini is about 35-40 elo stronger than SF. I have run over 1000 game at this TC.

[S.Taylor]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

For me is very useful.

I have faith. If Houdini can, Stockfish will do better some day.

I don't WANT Stockfish to be good at ultra fast time controls, because it would come at the expense of long time controls.

[S.Taylor]

Ultra fast time should be used in testing between the master vs dev version only. For two different engines, the one which has fast move generator and easy move generator will get significant advantage. For example if H4 needs only 0.001 sec to generate a move and SF need 0.01 sec then the former one will get significant advantage. However that is almost negligible even in 5'+3' blitz.

Time will come where ultra fast time controls for all us human beings will be enough to test ideas.

I don't even think that a human himself can think up his own ideas in just a few seconds.
So why would he be in such a rush to get answer from computer. Would he, the human, be able to digest everything that the computer is pointing out to him, in ultra fast time?

[S.Taylor]

And if now, he would test an idea by leaving the computer computing for an hour, and watching as it changes its top lines and evaluations, then when all that takes only one second, will the top engines have reached 4000 elo or more? And if it did, then _I_ would want to see what happens after an hour analysis of THAT level!

[felix]

And if now, he would test an idea by leaving the computer computing for an hour, and watching as it changes its top lines and evaluations, then when all that takes only one second, will the top engines have reached 4000 elo or more? And if it did, then _I_ would want to see what happens after an hour analysis of THAT level!

Well. I use this ultra fast time control to test a lot of interest positions.

I learnt to play R + P vs B +P. How to win AND how to defend.

Queens endings too. Of course Rook endings.

I use them to test some opening ideas. BOTH colours.

I just set up the position and hit the start button and pay attention. Patterns starts to rise.

You can receive result of dozens of games in a matter of minutes.

I love too TCEC time control, but I don't have the time to wait 100 games at long time controls. Even 5 minutes games is too much.

Have you tried to play this monsters at instant time for them and any time for you? They will win you as they win me or at least you will have a hard time drawing the game.

15 years ago I was a fairly strong player at LTC and very strong at STC. Now I am sure, at least at STC I am stronger. Just testing ideas at this STCs.

Faster computers make good engines be useful at STC. Shorter and shorter every time.