Critter 0.90 1cpu tests

[Dann Corbit]

do you know that i had begin with EnginMax about a strength of 1500 elo or something ?

and then grow up until 2100 elo , i think improvements over 100 elo each version is possible but this improvements are step by step, and not for the first time comes with 3000 elo engine, like houdini, gull, fruit, rybka, with those engine are i think is something wrong.

then i continue with Tornado but use the same ideas at the begining what i do know from EnginMax allready done, and Tornado 1.0 had only some elo points more then EnginMax like 2200 elo or something, then i working very hard on Tornado and make it until now with a strength of 2700 elo, but this take over 6 years of my life.

do you understand what i mean ?

I think that I understand. I also think you did not mean to attack the Critter author, but simply felt frustration at how difficult it is to compete.

I think if you want to add 600 more Elo, you will have to study the code of:
Fruit
Stockfish
Gull

If you don't want to do that, it will be a lot harder but the path may be more satisfying.

When you consider all the brilliant people who spend a great deal of time on computer chess, it is not surprising that there are a huge number of very strong chess engines. There have been a lot of innovations recently.

Anyone who can write a chess program that even works correctly has done something very special. I think that you should be very proud of your achievements.

[Dann Corbit]

yes i know that is possible making improvements over 80-100 elo from older version to the new one.

what i dont understand is the quickly step what critter grows up in a very short time. who other programmers need over many years.

and i ask me what they doing different work ?

its ok that he reading and understanding other sources, that i am doing too, but how he know that what is a good idea and what is not, he must try this first before implement it in his programm too.

Consider:
The path of Richard Vita is not significantly different than the path of Tord Romstad. Tord's path was a bit more transparent because we got to see the code for his late, great leaps. However, Tord's chess engines also had a meteoric rise. And yet Tord was not just examining other people's innovations, he was creating his own innovations. I guess that Richard is doing the same thing.

[Tord Romstad]

Consider:
The path of Richard Vita is not significantly different than the path of Tord Romstad. Tord's path was a bit more transparent because we got to see the code for his late, great leaps. However, Tord's chess engines also had a meteoric rise. And yet Tord was not just examining other people's innovations, he was creating his own innovations. I guess that Richard is doing the same thing.

Thanks for the compliments, but to be fair, Joona and Marco deserve at least as much credit as me for the strength of Stockfish.

Richard: Congratulations on the release of Critter 0.90! Of course I would have preferred it to be open source (like with any program), but I don't doubt for a moment the originality of your work. Well done! I hope you'll be able to maintain this rate of progress.

[Engin]

i dont doubt the speed increase strength of critter, i want learn what make it so fast better.

and of course i am learning from stockfish too, i implement the nearly same idea of LMR and prunings what is doing there, but i do my own ideas into and my evalution is very different, only what i am not do is the singular extensions idea, that tryed one time but of bad results i removed this quickly again.

did singular extension make any huge strength improvement ?

[Engin]

i do study fruit, stockfish but not gull, mostly i learned from hyatts crafty bitboards.

[Dann Corbit]

i dont doubt the speed increase strength of critter, i want learn what make it so fast better.

and of course i am learning from stockfish too, i implement the nearly same idea of LMR and prunings what is doing there, but i do my own ideas into and my evalution is very different, only what i am not do is the singular extensions idea, that tryed one time but of bad results i removed this quickly again.

did singular extension make any huge strength improvement ?

I guess for most engines it is less than 50 Elo, probably around 20.
From what I have seen also both LMR and also extensions are very fiddly. You will have to spend a lot of effort to figure out which nodes should or should not get the treatment.

Lots of little things at 10 Elo each will add up to a big Elo gain.