Chess related master's thesis

Hi,

I study computer science and I want to make my master's thesis chess related. 

I want to ask you guys if you have any interesting reading or even recommendation what could be potentially interesting chess/cs/math topic to make a research on. I was thinking maybe about comparison of algorithms used in chess engines (alpha-beta, minmax etc). But it could be also something completely different.

Any idea is appreciated and will help me to make a nice paper and maybe bring some contribution to the chess community. Thank you!

I am not sure, but perhaps  an interesting topic would be something about which openings are scoring better at different ratings. Starting with the hypothesis that uncommon openings will score better at lower levels. Let’s take the Alekhine defense, which is certainly not very common at any rating. Is this defence scoring better than let’s say 1…e5,  1…c5, 1…c6 or 1…e6 at let’s say a chess.com rapid rating of 1000 to 1500? If yes, does  this change at higher levels? You can perhaps take a database (lichess) and see the evaluation after 10 or 12 moves, and work only with these numbers.

Chess UI/UX is interesting to me and how software is used for training purposes.  I don’t know if this is suitable for a graduate thesis.  A timely topic might be FairPlay and anti-cheating software.

Perhaps a paper on how engines calculate positions?  Other than that, I assume it’s just brute force to find the best move.

This is an interesting paper
https://arxiv.org/pdf/2111.09259.pdf

This is also interesting
https://arxiv.org/pdf/2112.09386.pdf 

Related to the post from @tygxc you might just search the arxiv  for chess-related research papers.  That should give you an idea for what the state-of-the art is and what problems researchers are considering.

I would personally be interested in advancing the ability of engines to mimic human play at high levels.  For example, the goal of the Maia engine is to play the most human-like move rather than to play the best move.  However, to train Maia, I think you need lots of example games played by humans for training, and there simply aren't enough at the grandmaster level to robustly train a neural network to play like a human grandmaster. 

If such a high level version of Maia could be trained, then it would be useful in cheat detection.  A score could be attached to a move denoting whether it was a human-like move (based on how much Maia preferred the move) or an engine-like move (based on how much Stockfish preferred the move)