Death of chess.

@Clearchesser said in #16:
> I can carry a Magnus Carlsen in my pocket if I invent the right machine. Honey, I shrunk the Kids!
You can't even use computers in actual games.

Plus chess can literally never be solved there's way too many possible positions, in fact I heard there was more chess moves then atoms in the universe. For comprehension, the observable universe is 93 billion light years long. For every meter there are 10^-14 atoms. For every kilometer there are a 1000 meters, so a 1000 times that. There are 9.461e+12 kilometers in a light year, then multiply that all by 93 billion. Maybe then you will under stand, this all multiplies to 100 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 10^80. But that's not just it, the Shannon number, the number of possible chess moves, is 10^120! Good luck trying to lay that out. Chess is still alive, and is alive as it ever will be.

Remember that number, googol, its larger than that!

@vsdgniuz Is stockfish better than alpha zero ? I thought alpha zero beat stockfish in their last match. I don't know, I could be wrong. I found contradictory claims when I did a google search.

A0m has not played since nor has it improved as the project was discontinued. While Stockfish has replaced it's evaluation function with neural net taught same way as A0 was. There is no absolutely no reason to think A0 would still be stronger. But since it no longer exist we cannot really absolutely sure but about 90% sure anyway

@petri999 Thanks for the update. I wasn't keeping up with events, but this is interesting. I assume that the Stockfish programmers used the knowledge from the Alpha Zero project somehow. I guess they must have paid some amount of money for that.

Chess ded? But what? why?how?

No they did pay A0 was described in research paper. The real one i.e original Go-version. First there was Alpha go trained by pro programs and it reached pro strength. then there was AlphaGo Zero that was trained by self play but is used some game specific information like symmetries to reduce load. This exceeded all humans easily. Then there was Alpha zero with zero information except rules sort of and it was trained to play Go, Chess and Shogi.

Based research paper google published from AlpaGo Zero a open source community project Leela Zero was started and it surpasses all humans quite soon. Then when Alpha Zero paper was publish a new open source community effort emerged Leela Chess Zero that took basic neural net stuff from Leela project and built a chess playing version. Then several years later after invention of NNUE (makin neural net use in standard computer feasible) stockfish in co-operation with leela chess zero team added neural net evaluation to Stockfish but keeping the highly tune search part intact.

So now stockfish has traditional treesearch and neural net evaluation and way stronger than it was back then.

Since all is based public papers published by Google there nothing pay for.

@vsdgniuz said in #27:
>
> I think you misunderstand. All I am proving is chess is that chess is not solved yet.
> If I can prove that as on date, no strategy has been found where white can not force a win or draw = black can win even when playing against the best chess engine, it is sufficient to prove that chess is not solved.
>

That is not true. HeX is a solved game for any N x N sized board: it's a win for the first player with perfect play. And the proof is easy: no game can end in a draw, and with a strategy stealing argument you can show that if there's a winning strategy for the second player, there is one for the first.

But for large enough N, we don't know the strategy.

If you have a way to force a win, then the game is solved. But a game can be solved without having a way to a win.