Sprint 7 - v7 - 16 hrs

v7.1 - 2 hrs

Goals

• Adding Depth limit to Quiescence Search

Logs

Quiescence search was taking up too much time, so I added a depth limit to it. I add the concept of volatility, which is a simple measure of how many captures there are in a position to limit the depth of the quiescence search. After playing a match, I analysed the game through chess.com and was pleasantly surprised. The engine got an estimated rating of 1550!

While manually looking through the game, I noticed that the engine was making some truly horrible blunders, so I have decided to remove quiescence search. This has brought the average move time per move, in a game of 50 moves a side down to a mere 3 seconds per move!

WOW! THERE WAS SOMEHOW A BLIND SPOT FOR MATE IN 1S! I fixed this. New estimated rating: 1700!

v7.2 - 2 hrs

Goals

• A simple UCI interface

Logs

Again, this was not that hard to implement. I just had to write a simple UCI interface CLI, and it worked perfectly first try.

v7.3 - 6 hrs

Goals

• Threading

Logs

Threading is a headache because of lack of automatic memory management in C++, but the results are in. In the Kiwipete position, at a depth of 5:

Version	Time 1 (ms)	Time 2 (ms)	Time 2 (ms)	Average Time (ms/s)
7.2.1	49896	49144	45507	48182.33/48
7.3.1	32530	36706	38876	36037.33/36

Therefore the newer version is on average 33.70% faster.

After integration testing, I realised that this is a lost venture as it makes random moves. I decided to come back to it later.

v7.4 - 4 hrs

Goals

• Optimisations (Docs optimisation, move ordering better, move ordering optimisation, endgame optimisation, transposition tables)

1. Docs Optimisation

   In the library docs (Getting Started | C++ Chess Library), it suggest a better approach instead of using board.isGameOver(), and I am going to try it out.

   This does not work, as I can not get the MiniMax algorithm to use the suggested optimisation.

2. Move ordering optimisations

   Added in the following move ordering optimisations:

   • MVV-LVA: Most Valuable Victim, Least Valuable Aggressor
   • Killer Moves

3. Endgame optimisations

   I added in various endgame optimisations along with a tapered eval function.

4. Transposition table

   Before adding the transposition table, to decide what to use as a key, I did a little test:

   

   As is apparent from the picture, FENs are the fastest, but my function to remove the last few digits from them is not, so zobrist hashing it is.

v7.5 - 2 hrs

Goals

• Time limited searches

Logs

I wrote up a function to implement Time Limited Searches

I added these in a day. I am now going to let it play a few matches against Stockfish and see how it does.

It is horrible! It is making random moves and it is not even playing the opening book. I will have to fix this in the next sprint.