The engine will mirror your skill level. | Developing a new App

Hi Lichess Community,

I am working on a new Chess App, mainly because I want to improve my Chess and at the same time leverage the power of Stockfish and Coding . Inspired by Play Magnus I see a opportunity to create something better suited to learn and improve in Chess.

How will this App be different than Play Magnus?

The main difference is that the App will not be a game, but a learning tool. The App will have features to help you improve your game. Play Magnus is to play games, and it does a great job of testing your abilities.

What is the best way to learn?

The best way to learn is by playing against humans, but not for the reasons you might be thinking. Playing against humans will teach you how to exploit weaknesses, which is crucial. Playing against the engine is a good way to let the engine exploit your weaknesses, so that you can figure out how to fix them.

This leads me to the first feature of the new App:
- Self-Play, e.g the Engine mirrors your skill level. That way you must learn to exploit mistakes of the engine.

Another feature that completes this is that the App will be able to figure out your "ELO" just that instead of ELO Points, the quality of your moves is measured and compared with that of different players.

What input will users be able to give?

Users will be able to give input by playing on the board, but also there will be different modes that are especially designed for improving your chess:

- by playing a variation, or even by selecting a move from a list of moves.
- inputing your own evaluation, that can be compared to the engines evaluation, to see how good your intuition is.

Have you ever thought of an feature that you really need for improving your Chess with an Engine but that was not there?
Critical thoughts?

Best regards,

RichardRahl

"Have you ever thought of an feature that you really need for improving your Chess with an Engine but that was not there?"

No. I don't need any of that. I don't need an engine to improve my game except for stockfish. I am still waiting for Lc0 to explain its own moves, this is all I'm interested in concerning engines and A.I.

Your idea sounds to me as if you were playing against yourself. I don't want that kind of future. I want to play against humans.

Thank you xPhilosophusx for your feedback.

I am very passionate about Chess and I would like to learn to play better (by a lot).

A lot of people ask how to improve their game.

The main advice is:

- Play a lot of games against humans.

But honestly, I would be lying if I said that is enough to improve your game.

To improve your game, you must play against humans, but you must play against humans who are slightly better than you.

An Chess Programm has the ability to know your strengths and weaknesses much better than any mentor could.
This is not the same as just playing against stockfish. It will be an opponement that pushes you to the edges of your abilities to barely win.

That being said the App will be for people being serious about learning Chess. It's fine to play Chess for fun.

I'm seeing a long long string of draws.

www.researchgate.net/publication/347464465_2_-_Quantifying_Chess_Skill
Read this article, might help you.

hey I'll be interested to be a beta tester and even contributed to the source code if it happened to be open source in the future. good luck bro!

@RichardRahl Hi Richard, sounds quite fun and interesting, hope it goes open source! :)

How do you plan to make the computer make mistakes in a human fashion? Relying just on evals to play a slight mistake/inaccuracy sometimes lead to a disadvantage that is very hard for a human to find. Is there anything you already thought of in regards to adding a human-like factor to the engine's artificial weaknesses?

Thanks!

@RichardRahl

"A lot of people ask how to improve their game. The main advice is:
- Play a lot of games against humans."

Wrong. The main advice is do tactics, study and play games. No one said just playing games would be enough.

"But honestly, I would be lying if I said that is enough to improve your game."

We agree on that.

"To improve your game, you must play against humans, but you must play against humans who are slightly better than you."

We agree on that too.

"An Chess Programm has the ability to know your strengths and weaknesses much better than any mentor could."

I disagree.

"This is not the same as just playing against stockfish."

I never said I was playing against stockfish. Stockfish is for analysis.

"That being said the App will be for people being serious about learning Chess. It's fine to play Chess for fun."

You are more or less implying that people who are not interested in your idea wouldn't be serious about improving. That's pretty f* up and insulting to be honest.

Anyways good luck, I'm not interested.

#1
Have you seen this paper by the Maia team? I think you will find it interesting and relevant to your pursuits.
Learning Personalized Models of Human Behavior in Chess
arxiv.org/pdf/2008.10086.pdf

@immagnus5yo most important for me is that the app works and actually helps with my chess, I am open to open source. If its really usefull people might help out with donations. Anyway I already have a job as data scientist.

I have a first prototype, it uses chess.js, chessboard.jsx, capacitor, react and stockfish. Similar stack to Lichess, but very light on the server side. You are welcome to beta test the App once its more furnished.

@PinToWin at the moment I do not use Neural Networks to play similar to humans, my approach (in short) is very simple:
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.
This leads to some intresting results, if you blunder the engine is also more likely to play badly, which acutally is good to stay motivated and come back. Normaly you have Zero Chance to come back against a decent engine. Which is really frustrating.

@xPhilosophusx You might want to look into what OpenAi is doing. GPT-3 can basically reproduce human text, there is no way an AI can not be taught to help our Chess to improve. Thank you for wishing me luck.

@raphaelkoszalka @jomega thank you. yeah, great paper! I will not use Neural Networks for now, but great opportunities for later on!