Chess Engine Leptir

[Eduard]

Current ranking at this hour on PlayChess:



All day yesterday same. All days the same.

How do ProteusSF that? his main accounts are always at the top. SUPERCOMPUTER with 1,000,000 kns has no chance to top ProteusSF, and MZ (Hypnos author, 64 cores) is even last in the current table. Incredible, is not it?

Or is someone cheating here? Can that be explained differently? Has anyone tested ProteusSF, this new computer game superstar (hahaha)?

[DrEinstein]

When 33 Stockfishes are playing Chess against each other, one Stockfish will be 1st and another Stockfish will end as 33rd. That's just what I would have expected. Some minor random source code changes made in one or the other Stockfish do not play a major role in this (somewhat boring?) game.
But to be honest, I don't know why Stockfish_Prot. is so often at the top. Statistically nearly impossible

[DrEinstein]

Sorry for this link:
https://chessengines.blogspot.com/2023/ ... s.html?m=1
but the picture very nicely illustrates what I have posted just before:
Nine (9) Stockfishes (randomly distributed), than Lc0, Komodo, and the rest.

[Eduard]

Every monkey can get such an Elo if he only plays against himself and wins. I could do the same.

Just fictitious, I don't do anything like that. I create the following accounts:

Main Account "Charisma".
Test account "Charisma test". --> I create this account from my nephew's computer, other house, IP, etc... no one can prove cheating.

Now my nephew plays with my book a whole day and easily manages 2970 Elo.

Now I'm making a weak book for the test account (I'm a master at that, believe me).

From now on, the test account will only play a few games against neutral accounts (this is what ProteusSF does). Then I leave the automatic mode and only challenge my main account. After 5 wins (with a strong book against blunder book) I win 5x16 Elo with the main account, that's 80 Elo. If I start with 2980 Elo, I'll be 3060 Elo very quickly and I'll be the best on the server.

I don't have to do anything now. With a few draws every few days, I'm always the front runner on PlayChess. That's how ProteusSF did it, or pretty much something like that (he has 2 main accounts - see above - and 6 test accounts). In my opinion, this is still cheating.

[Sopel]

sounds like a flawed leaderboard and a waste of electricity

[Eduard]

Personally, I'm not interested in leaderboards. Those who are always on top are cheating. Anyone who knows how to do it can easily be at the top. I just wanted to show how these cheaters do it. I like testing online because it's more exciting than doing it offline. I'm not just testing the engine. I like testing openings for openings books. There should also be people who are also interested in good opening books. I make them in 4 different formats - and if you don't believe it - yes I get mail from people who are interested in it more often. I also occasionally play in the online engine tournaments. For fun, and oh yes, I also get requests about the engines. There are also some good correspondence chess players (even GM). Actually everyone knows where to download Stockfish. Still, some would like to have Solista Attack or Leptir as well. Can't these people test and find out that the Leptir is just Stockfish? How is Leptir Analyzer different from Stockfish dev?

[Peter Berger]

It's obvious that you are putting a lot of effort in whatever you are doing. The question is, what exactly you are doing, and you are getting people to think about it.

My basic assumption is that some random changes to Stockfish won't matter that much if not too huge (then they will be horrible). So we are thinking about a huge pool of mostly identical Stockfish engines playing against each other on and on in some server room. And there obviously is a huge draw rate that has to be taken into account.

There probably is some influence of choosing opponents. You point this out yourself, so you will be doing some of this yourself to not to get "cheated" too badly.

This leaves the opening book. I can imagine this to be important in this environment, as everything you find out will work against several opponents.

sopel is not your nice guy, but I seriously wonder if his point is not just to the point. How is this not a waste of electricity? What is it (besides having fun which would be a very valid thing) that you are achieving? Have you found out things that you think would be interesting for others?

[Eduard]

In my opinion, a hobby should be fun. What do the many Stockfish developers achieve with their tests? Why are you doing the tests?

Let's take the engine rankings. What is the sense if a tester tests engines every day of his life, only to find statistically that engine A has achieved an Elo gain of, say, 5 Elo? Then version B, C etc. is tested. The tester only wastes electricity and time unnecessarily, just to create a ranking list that is no longer of interest to anyone after a year.

New versions of fish cooking have been constantly being developed for decades, and everything is documented with extreme precision. Several hundred testers, testing around the clock. To see after 8 months that a progress of 8 Elo was achieved.

If 200 testers, for 8 months, test something, that makes little progress, what sense does that make in the lives of those testers? Is the purpose of these people's lives just to improve Stockfish by 40 Elo in 5 years? Isn't that a wasted life working with Stockfish?

Is Stockfish meaning of life?

What I do, I do for pure fun, especially during the Corona period I couldn't go out much, I had Corona myself for several weeks.

But I don't intend to deal with computer chess all the time. No worries. That would not be enough for me as a meaning of life.

[Sopel]

In my opinion, a hobby should be fun. What do the many Stockfish developers achieve with their tests? Why are you doing the tests?

Let's take the engine rankings. What is the sense if a tester tests engines every day of his life, only to find statistically that engine A has achieved an Elo gain of, say, 5 Elo? Then version B, C etc. is tested. The tester only wastes electricity and time unnecessarily, just to create a ranking list that is no longer of interest to anyone after a year.

New versions of fish cooking have been constantly being developed for decades, and everything is documented with extreme precision. Several hundred testers, testing around the clock. To see after 8 months that a progress of 8 Elo was achieved.

If 200 testers, for 8 months, test something, that makes little progress, what sense does that make in the lives of those testers? Is the purpose of these people's lives just to improve Stockfish by 40 Elo in 5 years? Isn't that a wasted life working with Stockfish?

Is Stockfish meaning of life?

What I do, I do for pure fun, especially during the Corona period I couldn't go out much, I had Corona myself for several weeks.

But I don't intend to deal with computer chess all the time. No worries. That would not be enough for me as a meaning of life.

That sounds incredibly dumb from someone who clones Stockfish.

[Eduard]

I am currently testing a new version of Leptir (Leptir N1, faster search then Leptir Analyzer). In my new test, this engine manages 103 out of 110 solutions. For a quick check, I tested the engine against Stockfish dev in the bullet.

7500 kns each engine and Ponder ON (this is absolutely important to me because Ponder ON is played on the 3 known servers, PlayChess, InfinityChess, Lichess) - time control 60s + 0.1s.

After 270 games the status is 50%-50%.

My second engine GoldenEye (this engine can learn - but learn is switched off for the test) is only -1 behind Stockfish (Noomen 3move Book).

Sometimes I pause the test when I'm doing other things, currently it looks like this (second run games 59 - 500, Powerfritz 18 GUI) :



I expected a bigger advantage for Stockfish dev. Yes, I know these are too few games (Noomen 3moves book, contains only 250 variants), but all the most important openings of human theory are played - that's enough for me.

Both engines are better than Stockfish dev (only 91 solutions) with 101 and 103 solutions in position tests. It's inexplicable to me. Stockfish dev is not better in games but clearly weaker in position tests. So what are the merits of Stockfish dev compared to my clones?

Download Noomen 3move book PGN:
https://pixeldrain.com/u/2r3iGiP8