Legal Computer Assistance

@EvilChess

I see, chess as a science so to speak.

But engines by now are so strong that players will leave everything to the engine, knowing that any departure will be an inferior move. You wrote:

> Kasparov stated that a Human+Computer team can be stronger than either one alone.

But that was at the time. A human nowadays cannot usually make a computer stronger. So the "cyborgs" will really be assistants pushing the virtual pieces where the engine tells them. The presence of humans will basically be superfluous.

The cyborg idea works in settings where the human is still irreplacable, and benefits from computer suggestions but still needs to integrate them in the real-life situation. In the small, formal chess world, the human has no relevance any more if we're looking for best play, unless for a few exceptional cases.

I'm more interested in your proposal, namely to use computer assistance as a learning tool. And the idea to see the best moves for activity, material, king safety, etc. would be fun as we could orient the game in crazy directions.

I have another, related idea.
It is often remarked that opening theory had made it more difficult to enter competitive chess as by now one is forced to spend countless hours learning opening theory first.
And accordingly the Fisher chess variant was invented in order to minimize opening theory.
Another way to solve this would be to create a variant where both players have opening book assistance (just the current database-based opening explorer will be fine)!

This would both be a particular way to play chess (where the importance of opening knowledge is reduced, like in Fischer chess), and a great way to learn openings! After playing for some time in this variant, people could go back to normal play with a better understanding of openings, instead of just playing their same repertoire because they are too bored to learn new lines.

Maybe @thibault has an opinion on this. One day we might have computer rooms.

And another idea. I understand that implementing new variants is a difficult choice to make, since it's not obvious they will really work well, attract a sufficient number of players, etc. This prudence is good in order not to corrupt Lichess, however it prevents experiments, from which new worthwhile variants might appear. So I wonder if some sort of "Labs" section could be created where weird things might be attempted. They would not show on the main page, in the list of official variants, etc., but one would access them through that particular section. In this way we could see what works and what doesn't, try new variants, new ideas, new rules (where there are enough people interested in implementing them), and if something really works, gathers a lot of players, etc., then it could be integrated in Lichess proper.

@31

" But engines by now are so strong that players will leave everything to the engine, knowing that any departure will be an inferior move."

The human side of a cyborg can guide the engine what promising opening lines to pursue, what middle game positions to enter to have a higher probability of winning an ending position. I am talking about playing a game with longer TC like 30'+30" or longer.

Basically the human will create variations that are interesting/promising and these variations are constantly being analyzed by the engine tactically, this technique is used to minimize the horizon effect of the engine. Once the evaluation of the engine is fine and the human evaluation is fine too then the cyborg may choose this move or line of play. There are already a lot of knowledge that the human can learn from this method.

I also like to chose openings in engine vs engine games. Trying something offbeat is always fun.

Example game played by cyborgs at TC 30'+30". The game quality is high worthy of a place on your personal game collection especially if the opening is in one of your opening repertoire lines.

en.lichess.org/study/UKewPdTz

Fireworks #33 and chaos3 #29,

The cyborg team's goal is exactly this, to have rooms for computer use.

JacquesD #31,

You have an interesting point. I agree that top engine (Komodo) evaluation is superior than top human (Magnus Carlsten), but I still think that in some positions human may make a better move. For example, there was a game were both players castled short then black begun king-side pawn attack. It was a risky initiative which takes too many moves to prove itself and pay off (like) 40 moves latter and engines like Komodo would take hours to analyze so deep. Yet, even if Komodo says it has seen 40 ply deep (40 ply = 40 half-moves = 20 moves), it does not meen it saw all variations, as this is just an average depth. In that game the Komodo evaluates that as a bad move, pointing out that white is winning, but then it seems to agree with all following moves played by white yet it gradually recognizes black's advantage. And giving more time to the engine doesn't really change things because the relationship between time and depth of analysis is very exponential. Its more like doubled time for each additional move, depending on position complexity.

JacquesD #32, we already have this for correspondence games. You can check statistics from masters opening database during the course of the game. Try it.

In #34, yes, I think Beta-version enhancements could be implemented separately. Yet, for this to work smooth we would need an entirely separate beta-lichess portal. That is to avoid the overload of giving maintenance for shared resources (such as database or browser session).

In regards of the use of engines in opening books, I agree that they can be useful, but only in advanced positions. In initial positions, were both sides have too many good choices, the engines just can't see deep enough. When opening books go all the way into middle-game, there is a point were engine evaluation can be useful and when master's database starts running out of games for that variant. Then a engine's game database could potentially still be rich in number of games, since engines can analyze position non-stop. However, I haven't found yet a engine database so big. Here we have 2 million games from master's database, while ChessBase has 8 million, but the biggest I found from engines is the one below with just 0,6 million games:

http://www.computerchess.org.uk/ccrl/4040/

Currently the only contribution I have access to from chess engines to openings study is the "let's check" feature from ChessBase, which anyone can consult or contribute online for free:

database.chessbase.com/

Its also integrated in Fritz application, were you can integrate other engines such as Stockfish or Komodo. In positions were master's database becomes scarce, sometimes you still see a 35 ply depth evaluation by Komodo and a 42 ply depth evaluation by Stockfish (stockfish looks deeper quicker), contributed by people who left the engine analyzing that position for hours. For instance, this feature is the only resource I have to study Evan's gambit deeper before JusticeBot has another opportunity to crush me with this gambit in his simuls.

#35: We can learn from it, we can direct it in fun or interesting (from our point of view) ways. But with respect to chess strength, the best that the human can do is get out of the way, so in this sense the "cyborg" idea doesn't apply.

Maybe I'm wrong. There's a very simple way to see. Just play, as a cyborg with you and Stockfish max, against Stockfish max alone. If you do significantly better than Stockfish alone against Stockfish alone, then I'm wrong. You probably did it already. What were the results?

#38: Yes, maybe in some rare situations (that the human player won't be able to identify reliably), but not enough for it to be significant and for the human-engine tandem to qualify as a cyborg.

#39: Not sure why you talk about "engines in opening books" -- I was just suggesting access to a game database, for openings. Maybe you're replying to someone else or pursuing a personal line of thought.