The engine will mirror your skill level. | Developing a new App • page 2/2 • General Chess Discussion • lichess.org

I think it's a great idea- but I prefer setting Shredder to be my rating +200 when I practice. Then after I lose 2/3rds of the time I go backward to find the tipping point and search there for a better idea. It would be great to be able to do both in your app. Also rating is one number that summarizes a constellation of skills. It would be great if it would diagnose your "effective rating" in different aspects of the game.

check-sandwich_1-2-3

#12

I will just play a game and if I lose, most of the time I will analyse it with the lichess engine stockfish and/or look at it with my chess coach and he will find my mistakes and help me with it. I think sometimes human analysis can sometimes be better that computer analysis if you don't understand why a move is a mistake. However if you want to find the best move in a position, it is hard to beat an engine in my personal opinion.

NM MrPushwood

#13

"there is no way an AI can not be taught to help our Chess to improve"

What in the world does that mean?

"there is no way an AI can not be taught to help our Chess to improve" What in the world does that mean?

check-sandwich_1-2-3

#14

well, I get your point but it just means what it says, even if you don't think it is a true statement

NM MrPushwood

#15

Actually, I don't think it's a statement.

pawnedge

#16

In fact I can think of lots of ways an AI can’t be taught to help your chess improve, and very few in which it can. . . .
There’s probably an infinite number of the former, and maybe one or two of the latter.

🤷 Just saying.

In fact I can think of lots of ways an AI can’t be taught to help your chess improve, and very few in which it can. . . . There’s probably an infinite number of the former, and maybe one or two of the latter. 🤷 Just saying.

NM MrPushwood

#17

And regarding the engine "mirroring my skill level"...I was hoping for a bit more from one of those gadgets.

RichardRahl

#18

@swimmerBill you will have an option to set the skill level, but the App will figure out your level after a few games.

@MrPushwood once you import or play a few games, only then the "mirrowing" will be really good. As the App is able to find out your Skill Level from your moves. Maybe I also can generate an opening book from the games, so it plays really similar. One feature I have in mind will help you to prep against another player, by provinding all his public games. In a way you can play against him, before you play against him in person. Not sure if my method without Neural Networks is good enough for that. Otherwise I will add Neural Networks.

@pawnedge I have worked on Machine Learning/Neural Networks before, you would be surprised what is possible. In 2 years every Tesla will be autonomus.

@swimmerBill, @check-sandwich_1-2-3 thats are current possibilities to use the engine, and good ways to improve.

Anyway there are simple ways to make training with the engine really effective, that are not used right now in other Apps.
I want to close the gap. So every reply from you is important, as it shows me what you think is important to train (with the engine) and what features you think are valuable.

@swimmerBill you will have an option to set the skill level, but the App will figure out your level after a few games. @MrPushwood once you import or play a few games, only then the "mirrowing" will be really good. As the App is able to find out your Skill Level from your moves. Maybe I also can generate an opening book from the games, so it plays really similar. One feature I have in mind will help you to prep against another player, by provinding all his public games. In a way you can play against him, before you play against him in person. Not sure if my method without Neural Networks is good enough for that. Otherwise I will add Neural Networks. @pawnedge I have worked on Machine Learning/Neural Networks before, you would be surprised what is possible. In 2 years every Tesla will be autonomus. @swimmerBill, @check-sandwich_1-2-3 thats are current possibilities to use the engine, and good ways to improve. Anyway there are simple ways to make training with the engine really effective, that are not used right now in other Apps. I want to close the gap. So every reply from you is important, as it shows me what you think is important to train (with the engine) and what features you think are valuable.

RichardRahl

#19

Feature (1): Show to the player when to take more time to think, based on past performance. For example if the player always blunders or makes mistakes at a certain part of the game.

Feature (2): Enable/Disable visualization of squares that show which pieces are not defended.

Feature (3): Show your average engine evaluation on move 1,2,3.. n. Gives motivation to beat the average and shows where to improve as well.

Feature (4): Calculate "ELO" for a player, after they played or importet a few games. This ELO is calculated as follows: Analyse all positions for moves 1,2,3.. n for all games. Compare the profile with that of lichess's game dataset grouped by different ELO classes (2000-2100,2100-2200,...), the most similar group is your ELO.

I am just documenting some features I have in mind. As well for me to remember them. Please feel free to add, create or comment on them.

I will be sharing a few games next week, playing with the engine at different levels.

Cheers,
RichardRahl

Feature (1): Show to the player when to take more time to think, based on past performance. For example if the player always blunders or makes mistakes at a certain part of the game. Feature (2): Enable/Disable visualization of squares that show which pieces are not defended. Feature (3): Show your average engine evaluation on move 1,2,3.. n. Gives motivation to beat the average and shows where to improve as well. Feature (4): Calculate "ELO" for a player, after they played or importet a few games. This ELO is calculated as follows: Analyse all positions for moves 1,2,3.. n for all games. Compare the profile with that of lichess's game dataset grouped by different ELO classes (2000-2100,2100-2200,...), the most similar group is your ELO. I am just documenting some features I have in mind. As well for me to remember them. Please feel free to add, create or comment on them. I will be sharing a few games next week, playing with the engine at different levels. Cheers, RichardRahl

RichardRahl

#20

Feature (5): Different skill levels for the engine opponement:
I analysed many games with stockfish for each move I write down the evaluation. For players of different stengths, I group them togehter, for example 2000-2100, 2200-2300. Then I have a probability distribution of the evaluations per move. So each move stockfish looks for a given move that has the randomly selected evaluation.

Already implemented, checking the quality atm.

Feature (6): Extension of (5), Mirror Skill Level:
Play or import games, use same method from (5) to let the engine mirror your strength. Needs enough games from the player to be accurate. Neural Networks are a option if skill levels are to ambigous.

Feature (5): Different skill levels for the engine opponement: I analysed many games with stockfish for each move I write down the evaluation. For players of different stengths, I group them togehter, for example 2000-2100, 2200-2300. Then I have a probability distribution of the evaluations per move. So each move stockfish looks for a given move that has the randomly selected evaluation. Already implemented, checking the quality atm. Feature (6): Extension of (5), Mirror Skill Level: Play or import games, use same method from (5) to let the engine mirror your strength. Needs enough games from the player to be accurate. Neural Networks are a option if skill levels are to ambigous.