TalkChess.com

Martin Thoresen wrote:Thanks alot for your input, guys!

Honestly, I think 48 games are too much. Not statistically, of course, but I am thinking of cutting it down to 32 considering the time it takes just to let 1 game play through.

I also prefer to test Stockfish next so this will be my choice. I think it will be a fairly even match between SF and Houdini.

If the new (or even the old) Stockfish seems to be clearly stronger than Houdini, then why not go straight vs Rybka 4?

Or, do a short one first, to find out if one is clearly stronger. (24 games? 12 games?)
[Or are people suggesting that a different one, (like firebird 1.3?) may already be stronger than Stockfish]?

Rubinus wrote: Maybe use larger book. So far over short -> example E9 -> 7.de5 (stupid exchange variant). Optimal length cca 24-36 halvmoves, my guess.

Hello Pavel!

I will not use a book. I prefer opening positions from a PGN that are fairly balanced.

My goal with these kind of long TC matches are to let the engines play opening, middle and endgame without assistance.

The positions selected in the PGN are based on what most GM's would/could play.



Great news about SF 1.8, it will definitely have a shot vs Houdini next after this match. I will use the same positions and same match length, only with 2 GB hash instead of 1 GB.

S.Taylor wrote:If the new (or even the old) Stockfish seems to be clearly stronger than Houdini, then why not go straight vs Rybka 4?

Or, do a short one first, to find out if one is clearly stronger. (24 games? 12 games?)
[Or are people suggesting that a different one, (like firebird 1.3?) may already be stronger than Stockfish]?

Since Fire is based on Ippo it's stronger than SF 1.7.1 at least. And since people think Houdini is based on Ippo as well I will not test other Ippo-family engines.

SF 1.8 will play Houdini first, then R4.

Martin Thoresen wrote:Since Fire is based on Ippo it's stronger than SF 1.7.1 at least. And since people think Houdini is based on Ippo as well I will not test other Ippo-family engines.

SF 1.8 will play Houdini first, then R4.

If there is no reason to believe that any other engine (or even IPPO) may be stronger than Stockfish 1.8, then fine. Otherwise, it seems a shame. Rybka seems to be clearly stronger than Houdini, and one can feel this even from the drawn games. Maybe it's due to TB's, or bugs in Houdini? I don't know. But if there is not a very powerful upstart around, it looks like Rybka 4 still holds its position on the charts and may continue to do so for atleast another 6 months.

S.Taylor wrote: If there is no reason to believe that any other engine (or even IPPO) may be stronger than Stockfish 1.8, then fine. Otherwise, it seems a shame. Rybka seems to be clearly stronger than Houdini, and one can feel this even from the drawn games. Maybe it's due to TB's, or bugs in Houdini? I don't know. But if there is not a very powerful upstart around, it looks like Rybka 4 still holds its position on the charts and may continue to do so for atleast another 6 months.

I am not sure I follow you.

Houdini is probably stronger than SF 1.8, and R4 is stronger than both of those, so what is your point exactly?
I am not interested in running only R4 matches.

Besides, to find out which engine is clearly stronger than another engine, a lot more games than 32 or 48 are needed.

These matches exist for excitement only, not for some rating list.

edit: I might run R4 vs SF 1.8 next if that is what people want.

Best Regards,
Martin

Martin Thoresen wrote:

S.Taylor wrote: If there is no reason to believe that any other engine (or even IPPO) may be stronger than Stockfish 1.8, then fine. Otherwise, it seems a shame. Rybka seems to be clearly stronger than Houdini, and one can feel this even from the drawn games. Maybe it's due to TB's, or bugs in Houdini? I don't know. But if there is not a very powerful upstart around, it looks like Rybka 4 still holds its position on the charts and may continue to do so for atleast another 6 months.

I am not sure I follow you.

Houdini is probably stronger than SF 1.8, and R4 is stronger than both of those, so what is your point exactly?
I am not interested in running only R4 matches.

Besides, to find out which engine is clearly stronger than another engine, a lot more games than 32 or 48 are needed.

These matches exist for excitement only, not for some rating list.

edit: I might run R4 vs SF 1.8 next if that is what people want.

Best Regards,
Martin

Sure it's excitement OK!
At the momment, the biggest excitement (besides the chess itself), is to see if something dominates R4!

A match between two possible number 2's is also quite exciting. And it's also interesting to see what Houdini does vs something else.

Oh no! Black just now played 42...Kh6 (game 38). I had overlooked this!

(thinking the bishop was lost. I get the impression that there is no longer any such thing, with these engines, of any tactical line, however long, being entered into, if it would end up with losing anything).

Houdini MIGHT be taking this one (41) too!

S.Taylor wrote:Houdini MIGHT be taking this one (41) too!

It certainly seems so. Very nice game by Houdini, Rybka has been playing from a cramped position for almost the whole game.

Broadcast is back up after another GUI crash.

It looks to me like, perhaps...... Houdini is doing not too badly in this game, 42, too.

If he gets this game then just think:

It is very likely that it may end up that the last 30 games, if it had all started at game 19, would then have been a match victory for Houdini.

The first 18 games may just have been a little lucky for Rybka 4, and all its success in this match is from them.

Therefore, the feelings of many people, like on another thread around here, that Houdini is the stronger engine, might yet be the case.

(To actually overcome the deficit from the first 18 games, would require almost straight wins from now, vs. Rybka)