[cepth]

I have to commend the author for taking on a beast of a task. It's not easy to be learning/relearning chess while also trying to program an engine!

I think programming a chess engine that can beat most club players (let's say sub-2000 rating) is not too hard. At that level, human players will make inaccurate moves (as well as blunders at lower ratings). It is however much harder to develop engines that are competing at GM level (~2500+).

This wiki has some good high level articles (https://www.chessprogramming.org/Main_Page). Even something like a chess bitboard representation is non-trivial to code.

Implementing minimax and comparable algorithms may be straightforward, but the actual evaluation of positions for traditional engines is a distillation of thousands of pieces of expert knowledge. See for example the parameters that can be tweaked in Fritz (http://help.chessbase.com/Fritz/16/Eng/index.html?000038.htm).

If you look at the history of chess engines (https://www.youtube.com/watch?v=wljgxS7tZVE), by the 1990s and later most of the top chess engines have had masters, international masters, and grandmasters intimately involved with development. For example, Deep Blue had several consulting grandmasters. Rybka (the world's best engine 2007-2010) had IM Vasik Rajlich as primary author and GM Larry Kaufman closely involved with tweaking its evaluation functions. Kaufman also went on to write the (still) very strong engine Komodo (https://ccrl.chessdom.com/ccrl/4040/).

Traditional chess engines used to have glaring weaknesses like playing poorly in closed positions, being poor at avoiding disadvantageous endgames, etc. By the mid-2000s many of these weaknesses disappeared, but as recently as Fritz 9 there were well known opening sequences where engines could be tricked into playing losing lines.