Winboard Forum

Hi to all!
When I set a critical position for a game between engines , some engines choose a move , some choose another:
for instance , in 2 knights defense
1 e4 e5 2 Nf3 Nc6 3 Bc4 Nf6 4 Ng5 d5 5 exd5 , most engines play 5..Nxd5?
and in Nimzo-indian Leningrad
1 d4 Nf6 2 c4 e6 3 Nc3 Bb4 4 Bg5 h6 5 Bh4 c5 6 d5 , most play
6..Nxd5?
Then , the engines that DON'T PLAY Nxd5 in both cases ( but moves approved by Theory , as Na5 , Nd4 or b5 in 1st case , d6 ,O-O , b5 in 2nd ) are roughly the same , namely Junior , Shredder and Gandalf
Is it just a chance? or is there in their program some feature that prevent them of choosing the wrong move?
Answer is important , since it should lead authors of a new program to include this device

There is probably not a single feature, but a collection of many features. All of these programs have better evaluation, search, etc as compared to the others which play .. Nxd5?

A chess program's play is a sum of many things, and unless we are talking about a tactical move, its a bit difficult to say how a program got around to playing a particular move.

Pallav

Hi Pallav!
If you are right , this would follow the overall strength , and yet nobody can say whether Gandalf , for instance , is stronger than Fritz

Hi Andrew!
One can answer in two ways :
First , engines are used not only for tournaments , but also for analysis and I can't be confident in an engine that falls in these traps ;
if there is a recipe for avoiding these traps , I wish that programmers know it , to give me better programs
The secund reply is that programmers of Hiarcs or Fritz are as clever as those of Junior or Shredder : surely they are aware of this deficiency
( for the Nimzo Leningrad it has been discovered independantly and published by Robin Smith in "Modern Chess Analysis" )
so , if they did not correct this "bug" , they could not , and if they could not , something in the move search prevented them to do it ; they could not change this without deeply modifying the structure of their program
Recently , I downloaded a lot of free engines for which I had no time to verify whether they fall in the traps
By chance , maybe some will be "open source" ,and we shall see whether
my hypothesis is true