Botham bot possible ? • page 1/1 • Game Analysis • lichess.org

As a side project, I tried to make a bot based on a gotham but this have proven quite challenging.

At the end of the day, bot just obey programing and some behavior are hard or impossible to code.

Only one game so far that would qualify as kind of what we want but even that is subject to debate as I like to play the devil's advocate.

Here both botham and Stockfish are at 2500 elo, botham is black. Why would the sac of the rook not really qualify as botham being successful?

https://lichess.org/D2uYkL57

As a side project, I tried to make a bot based on a gotham but this have proven quite challenging. At the end of the day, bot just obey programing and some behavior are hard or impossible to code. Only one game so far that would qualify as kind of what we want but even that is subject to debate as I like to play the devil's advocate. Here both botham and Stockfish are at 2500 elo, botham is black. Why would the sac of the rook not really qualify as botham being successful? https://lichess.org/D2uYkL57

CamilleQueenOfTheBot edited

The main issue with this bot is that sac of rook leading to an advantage are VERY rare, With SF at 2100/1900 elo those occure at a frequency of around 1 per 100 games. Any strategies to improve those stats?
On Gotham we tried having it look for preparotory move before the sac at 1 and 2 ply but the results were inferior to just look for sac on the first move. Currently it looks for sacrifice on knight/bishop and pawn max 3 square of the king.

For the main engine, I was thinking castle early and have open or semi open file (on our side of course) but stat analysis of hundred of game showed that both conditions were present in more than 93% of the game played and each time you give an objective to the bot you take a risk of lowering its overall elo.

Any input would be welcome

Move 14

https://lichess.org/6eOcmDCO#30

Move 42

https://lichess.org/FrCJ8OY9#82

Move 29

https://lichess.org/4qLj1Xu5

Move 34

https://lichess.org/7XjL1Dey#67

Move 21

https://lichess.org/pxWP4wA6

Gotham is a bot with two engines, the first one is the one against which you play and you control its elo. The second one is Gotham, Gotham only look for move where a rook sacrifice would lead to an adavantage and it is always at 2500 elo. So if you are around 1500 elo, a rook sac probably mean you have just lost the game. If no rook sac leading to a gain of XXX cp, gotham does nothing. The main issue with this bot is that sac of rook leading to an advantage are VERY rare, With SF at 2100/1900 elo those occure at a frequency of around 1 per 100 games. Any strategies to improve those stats? On Gotham we tried having it look for preparotory move before the sac at 1 and 2 ply but the results were inferior to just look for sac on the first move. Currently it looks for sacrifice on knight/bishop and pawn max 3 square of the king. For the main engine, I was thinking castle early and have open or semi open file (on our side of course) but stat analysis of hundred of game showed that both conditions were present in more than 93% of the game played and each time you give an objective to the bot you take a risk of lowering its overall elo. Any input would be welcome Move 14 https://lichess.org/6eOcmDCO#30 Move 42 https://lichess.org/FrCJ8OY9#82 Move 29 https://lichess.org/4qLj1Xu5 Move 34 https://lichess.org/7XjL1Dey#67 Move 21 https://lichess.org/pxWP4wA6

CamilleQueenOfTheBot

rook sac move 26

https://lichess.org/jp58pVo1#51

Rook sac move 19

https://lichess.org/biPJii64#38

If anyone have idea for improving this bot they are welcome other wise, I think it can just be an add on when playing a bot with handicap can force a 1500/1900 elo to pay special attention to the rook when playing. Here we used the rook, but the horse master had one that was specifically looking for fork opportunity.

For chess 960, a scout that look to exploit undefended pawn could be fun.

rook sac move 26 https://lichess.org/jp58pVo1#51 Rook sac move 19 https://lichess.org/biPJii64#38 If anyone have idea for improving this bot they are welcome other wise, I think it can just be an add on when playing a bot with handicap can force a 1500/1900 elo to pay special attention to the rook when playing. Here we used the rook, but the horse master had one that was specifically looking for fork opportunity. For chess 960, a scout that look to exploit undefended pawn could be fun.