Chess engine questions

Hi. Im not sure if this is correct place to post this, but if someone knows enough it, than could you please answer me something?

If one were to give an engine a database of books to learn from, study, etc, but it stemmed from only 1 player, would the engine play like said player? Like, Tal or Petrosian for example.

NO....

There are some engines that will play like a certain GM.  But its not a real accurate representation of any GM.

interesting. have u tried this before?

Machine learning doesn't learn from books haha

But if you tried machine learning with a database of games from a particular player, that would be interesting. Probably there aren't enough sample games to train a machine to play like Tal, or Petrosian, etc. There are probably around 1000 quality classical games for either of those players, so to make a program that mimics them you'd have to be a little more creative about it.

And remember, not only is it on the order of 1000 games, but that would be across their whole career. As players get older they tend to switch to a more strategic style. Tal did this too.

i understand. so its not exactly impossible, but instead, unlikely that the engine will even play like the player? i mean, it can always find better moves than humans can, so there's that

A "normal" engine, the alpha-beta (AB) search type, plays better, sure.

But you can't train an AB engine anyway, so that's not what you seemed to be asking. You seemed to be asking about machine learning.

As a general rule, sure, you can train a computer to do anything as long as the dataset and computing power is large enough.

Another misconception is not realizing as many as 9 times out of 10 players like Tal and Petrosian (and even lesser GMs who were never world championship contenders) would be playing the same move. It's rare a player has a chance to show their style.

Kasparov had a tweet about this during one of the recent world chess championship matches... that most of the moves would be played by most GMs, but the WC and WC challenger will play a few moves per game differently, and that's what makes the difference.

You would have trouble getting a sufficiently large consistent data set, though. Fischer has less than 1000 games on record, for example, and that includes Blitz games and off-hand games as well as tournament games.