Are these engines dead?

[Robert Allgeuer]

... which would not exclude the possibility that a bitboard rewrite of Fruit would be 40% faster on a 64 bit system and thus be even stronger than current Fruit using identical ideas.

Robert

[Uri Blass]

40% faster is not very important in computer chess.

The difference in playing strength between fruit2.1 and the next amateur is clearly more than it so even if you are correct it is not a reason for programmer to go to bitboards because the same time that they use to translate it to bitboard can be used better by testing ideas.

Uri

[Jaime Benito de Valle]

I cannot speak for him, but last time I spoke with Carlos (Pepito's programmer), he wasn't interested in developing his engine any longer, or at least not seriously.

[Ralf Schäfer]

... while bitboards are becoming less popular.

Tord

Is this true? What about the strong newcomers such as Spike, Pseudo, Naum, Scorpio etc. No bitboarders amongst them?
If it were so it would be remarkable: Just at the moment when 64 bit computing is becoming commonly available - and bitboards would benefit most (30% in the Crafty SPEC benchmark) - engine programmers would turn away from it ...

Robert

Hi Robert,

Spike is actually not a bitboard engine, the board representation is based on a sort of 0x88. In fact we use an internal board of 14*16 squares.

Ralf

[Tom King]

As I've been out of the scene for so long, what are the major improvements which programmers have been applying in the last 18 months?

Hi Tom,

As far as I know, there are no recent algorithmic breakthroughs or exciting new ideas. Nevertheless, the general level of strength has improved enormously. I think there are two main reasons for this.

The first reason is that more programmers have realised the immense importance of avoiding bugs. Compared to a couple of years ago, there seems to be less focus on raw speed, and more focus on stability, simplicity and efficient search.

The second reason is that programmers no longer try to copy Crafty as much as before. Until recently, it seemed to be a commonly held belief that a strong chess engine had to resemble Crafty. By now we all know that this is wrong, and people are more willing to do things in their own, independent way.

There are also a few changes I would classify as trends rather than improvements. It seems to me that checks in the qsearch, high null move reduction factors, and late move reductions (my term for history reductions and related techniques) are becoming more popular, while bitboards are becoming less popular.

Tord

Hi Tord,
I do not agree that there are no exciting new ideas.

I think that there are new ideas and you are one of the people who post them.

history reduction is one example of an idea(originally posted by the programmer of smarthink and used in fruit)

Crafty simply does not use many productive ideas like history pruning and it is the reason that significant number of amateurs got above the level of Crafty.

I think that copying from free source programs is one of the reason for the progress.

My opinion is that in part of the cases it is simply cloning other programs and if the cloner is smart enough I doubt if people will find that program X is a clone but in part of the cases it is not cloning and if people understand some ideas that are used in fruit and find that using the same ideas help in their program then it can improve the level of their programs.

Uri

what's this "history reduction" about? I've not heard of that one..

[Uri Blass]

history reduction is reducing depth of bad moves based on the fact that they almost never did not fail high based on the history table.

You expect these moves to fail low and if they fail high you do a research with the original depth.

Uri

[Miguel A. Ballicora]

Hi all,

Just a list of engines the development of which seems to have stopped.

Gaviota

I hope there is still life in at least some of them.

Gaviota has been hibernating for a couple of years. It is not dead, it will come back. I am moving and going to a new job, so as soon as I get some stability, I will slowly start working on Gaviota again. Next phase is king safety, which now is non-existing. When I left Gaviota hibernating, I reached a point in which I believed it was very stable with mos basic functions working (book, protocol II, pondering etc.). I hope Gaviota has been stable these two years for everybody.

Regards to everybody,
Miguel