Hacker News new | ask | show | jobs
by manquer 1653 days ago
I can imagine training ML models that can imitate human play closely enough and even your play specifically to increase the odds you win.

Detecting a stockfish moves over enough sample size is easy sure, but detecting a engine which is designed to imitate human play not make the best move everytime is not easy with number of moves a human would play in their lifetime.
1 comments
usehackernews 1653 days ago
Sure, but this isn’t an easy way to cheat.

The original commenter say it’s easy to cheat at chess. While potentially possible, I wouldn’t consider building a ML model to mimic your own play an ‘easy’ cheat method.

I’d also say building a model to mimic your own play consistently would be incredibly difficult. But, that’s for a different conversation.

link
manquer 1653 days ago
While ML model building is bit more difficult , fuzzing a standard engine moves do only what you can understand is certainly possible without too much difficulty.

The point is cheating is bigger concern in online chess than other eSports

link
usehackernews 1653 days ago
Yes, it’s a concern. But that doesn’t mean unfair advantages and cheating doesn’t occur in other esports.

Anyway, I don’t think we’re disagreeing on chess cheating. For me, the small chance someone is cheating doesn’t ruin the esport for me - For you it does, which is a perfectly reasonable response to it.

Secondary thought in the engine idea you had - chess fraud detection, I imagine, goes well beyond just the engine and move likeliness. It will also human-like interaction (Can’t confirm this, but the PM in me has me consumed with thinking about solutions to this problem)

When people play a chess game online, they are frequently evaluating positions. This results in cursor/mouse behavior that’s sporadic. If a user is considering moving the queen, they’ll move their cursor over to it. A user relying on an engine for every move would interact with the board in a very precise manner.

A player with a perfect engine to mimic humans will still get caught as their interaction with the board would differ greatly because only one position is considered for each move.

link