I have been messing around with Komodo and the Skill Leve (from 1-25) setting and found that it moves instantly when on reduced skill level...
This is not good for realistic training, because as Lance Perkins (creator of Thinker) said back in 2008
6. Humans are not capable of 10 moves in 1 second. - Lance Perkins (on Human-like chess in engines)
His solution was, when weakening an engine, that:
(6) The engine's minimum move time is 1 second. If it is done searching for less than 1 second, it will sleep for a random amount of time (between 1 to 2 seconds).
Does anyone know if there is a setting in Komodo that makes it "sleep" for a while before making its move in Skill levels below 25?
We discussed this when those levels were added, but didn't (yet) implement it. Chess.com chose to have it move instantly (or maybe 0.1" delay, I forget) rather than have a one second delay on their site, so we didn't do this. We could have it delay one second on the download version and not on the chess.com bots if enough people ask for this. I personally agree with you.
lkaufman wrote: ↑Thu Dec 12, 2019 6:39 am
We discussed this when those levels were added, but didn't (yet) implement it. Chess.com chose to have it move instantly (or maybe 0.1" delay, I forget) rather than have a one second delay on their site, so we didn't do this. We could have it delay one second on the download version and not on the chess.com bots if enough people ask for this. I personally agree with you.
It's a very desirable feature for Human sparring , it is implemented in all 10 of the Honey family of engines , just check the UCI sleep box and it will adjust to what ever time control you are playing at. You must also have the limit strength checked and elect you level of play ( elo or predefined bia the drop down) ( unless you elected the nps play level , then it will automatically kick in)
https://github.com/MichaelB7/Stockfish/releases/tag/XR7 ( each download includes 10 engines for a variety of CPUs). For human play, I recommenrd, Black-Diamond, BlueFish and Weakfish. Weakfish is also suitable for human play without using one of the weakening play functions.
Rebel Decade solved this two decades ago. What the engine can do is checking what move it would have played instantly, but then it continues analyzing paying attention to the time control. When it's its time to move it ignores the extra analysis and makes the instant move anyway. This produces natural looking results (recaptures would still be played fast and such.)
Ovyron wrote: ↑Thu Dec 12, 2019 9:02 am
Rebel Decade solved this two decades ago. What the engine can do is checking what move it would have played instantly, but then it continues analyzing paying attention to the time control. When it's its time to move it ignores the extra analysis and makes the instant move anyway. This produces natural looking results (recaptures would still be played fast and such.)
Although this simulates tournament play, I think that the majority of users, including myself, don't want the engine to take as long as they do, but just take a second or so. I agree that instant response is jarring, and you don't get to see the move and the replies each played on the board. Maybe something like taking 10% of the opponent's time or one second, whichever is more, might be ideal, but just having a one second delay would be a good compromise between keeping the game as fast as possible (given the human thinking time) and avoiding the jarring effect.
Humans tend to imitate their opponents, specially time to move, so if an engine plays in 1 second, the human will be encouraged to play in one second, or much faster than they would if the engine took as long as them. This ends being a bad habit, and I've beaten people at 10 minute chess just because they seemingly play too fast, reaching a lost position with more than half their time on the clock. And engine would not just teach what to do in certain positions, but to spend enough time thinking about the moves (even if it just pretends to do it.)
When it comes to the Turing Test, time management would give an engine away, why are engines designed to imitate weak humans but to not imitate the rate at which weak humans play chess moves?
If the human wants the engine to play fast, they should give less time to the engine.
Ovyron wrote: ↑Thu Dec 12, 2019 8:24 pm
Humans tend to imitate their opponents, specially time to move, so if an engine plays in 1 second, the human will be encouraged to play in one second, or much faster than they would if the engine took as long as them. This ends being a bad habit, and I've beaten people at 10 minute chess just because they seemingly play too fast, reaching a lost position with more than half their time on the clock. And engine would not just teach what to do in certain positions, but to spend enough time thinking about the moves (even if it just pretends to do it.)
When it comes to the Turing Test, time management would give an engine away, why are engines designed to imitate weak humans but to not imitate the rate at which weak humans play chess moves?
If the human wants the engine to play fast, they should give less time to the engine.
Well, as long as the GUI supports separate time for each side, this should work. In the case of the ChessBase GUIs, there is the problem that they don't support fractional minutes for base time or fractional seconds for increments, so the minimum increment setting is 1' + 1", whereas the original post prefers just one second per move, which I also favor. For a Turing Test, yes, it should mimic human time use. But for actual use, how many people when playing a relatively slow untimed game of chess actually want their opponent to take as much or more time than they do? How many people think "I have nothing better to do with my life than spend extra time waiting for my opponent to move?". I certainly don't.
Because they should learn to ponder, definitively against humans you don't sit doing nothing while they're thinking about their move, you calculate while they are thinking, which is an useful skill to develop, and which won't be developed if they play against engines and get used to one second moves that don't happen in real games.
The reason this is important is that, if you managed to predict all of the opponents' moves, you can play your replies instantly and gain an advantage on the clock that can get decisive on the game (very often games are decided by the player with least time having to move quick and blundering). If the engine doesn't allow you to ponder then it's a skill you won't be practicing when you play against it.
