OpenChess

Why can an engine smart enough to instantly (more or less ) to find blunders that stronger engines make and cant find themselves not beat those engines when they're clearly smarter

Houdinis really dumb ( other than in end game play ) despite its strength , its can find many better moves than it made in20 minute game play with simple 5 minute analysis , so why did it not make those moves

in a recent 20 game it led as black by -0.80 on book closing ,, few moves later it was losing by that same margin , and went from bad to worse , when game was analyzed (by itself ) it found some better moves late on in game but did NOT find a blunder

Stockish 1.5 JA found the blunder more or less instantly and the move gave lead of -1
3 other engines failed to find any blunder , ergo to my simple brains logic stockfish is smarter engine and should be able to beat houdini

but it doesn't

answers in non geek speak please

veritas wrote:Why can an engine smart enough to instantly (more or less ) to find blunders that stronger engines make and cant find themselves not beat those engines when they're clearly smarter

Houdinis really dumb ( other than in end game play ) despite its strength , its can find many better moves than it made in20 minute game play with simple 5 minute analysis , so why did it not make those moves

in a recent 20 minute game it led as black by -0.80 on book closing ,, few moves later it was losing by that same margin , and went from bad to worse , when game was analyzed (by itself ) it found some better moves late on in game but did NOT find a blunder

Stockish 1.5 JA found the blunder more or less instantly and the move gave lead of -1
3 other engines failed to find any blunder , ergo to my simple brains logic stockfish is smarter engine and should be able to beat houdini

but it doesn't

answers in non geek speak please

cant edit post for some dumb reason im told i can no longer edit it
so game was a 20 minute one , fritz gui , Analise were 5 minute blunder checks on same gui

Pretty difficult to follow you if you don't show the precise game...

One blunder in one position means nothing. There is probably plenty of positions where Houdini finds Stockfish blunders. Besides that I believe Houdinis playing strength doesn't necessarily mean that it is a good analysis engine. Houdini without contempt is a a slightly different engine.

Will gladly post game if i ever figure out how to

last poster is missing whole point which is............
Houdin is DUMB.( other than in end game play where its smartest of all engines ) most engines are but Houdini takes cake in this , allow it to Analise any of its game ,even short 5 min analysis and it find it could have made a lot of better moves, not one , not even 2 but often many . the games are not ultra fast ones engines had more thinking time in game play than in analyses
i use houdini in book tests and often in real 20 minute game and have 1000's of examples not just one , throw a dart in data base and the Analise will show , many or several better moves, but as posted it cant find certain losing moves (blunders)
i find it strange that certain other engines notably stock fish (not the strongest of engines ) are far far smarter, seems no one else here does , so no point in perusing topic i guess

try it for yourselves

maybe" Astrro Boy" will pull his head from telescope and immortal forum to gob off about this as he does most things on ivanhoe and firebuggy there

AnthonyTheSage wrote:Besides that I believe Houdinis playing strength doesn't necessarily mean that it is a good analysis engine.

Agreed, at least if you already use Rybka. There are some positions where Houdini will just suggest a poor move, but otherwise Rybka would have suggested the other good moves. So, analyzing with both, you get a lot of redundancy and moves that you already have.

To get rid of the redundancy you have to choose one to use, but then Houdini would be the riskier choice, as it tends to pick losing moves in equal positions, because it's sometimes too overoptimistic in some endgames that are lost.

Probably with the learning features coming for Houdini 2.0, it's Rybka who'll become obsolete for analysis, unless their redundancy comes down.