I've been interested in Beeb chess for a long time, and I've been reading this thread. Thanks @Sydney for your efforts to set up these matches of various chess engines against one another--and video the contest!
Back in the day I played White Knight Mk11, and I still remember the little cheery beep when it was your turn to play. Then I upgraded to Colossus 4 on the Beeb. At my peak (sadly, long past) I could beat WK11 on its basic setting (averaging 10s per move of computer thinking time). Colossus was the chess engine I thought about most--and I realised it used two different calculations to evaluate the position: a Material calculation ("Mtrl"), which used standard numerical values for pieces, and a Positional calculation ("Psnl"). Modern engines typically evaluate the position using a combined score in centipawns, which allows them to sacrifice material for strong positional advantage.
I emailed Martin Bryant to ask him about the details of his evaluation function, but he said he couldn't remember, and no longer has access to the source code for Colossus. What I wanted to know is whether Colossus would sacrifice material for positional advantage, and what the trade-off would look like. According to Wikipedia, he retired in 2020, and his website is now archived (though it's accessible via the WayBack Machine) https://en.wikipedia.org/wiki/Martin_Br ... rogrammer).
Colossus was eventually released for the Amiga and PC, as Colossus X. I didn't like the GUI, which I thought was pretty garish. And I suppose the "ultimate" version of Colossus is the UCI-compliant engine which Martin released in 2008 (thoroughly tweaked for some pretty impressive strength upgrades).
What I've had some fun doing lately is pitting Colossus against some other chess programs. Obviously Colossus 4 is pretty strong right out of the box, and I wanted to see what it could come up with on its standard setting. Here's an example of it demolishing Battle Chess Level 6 (the PC version, since I think the Amiga version of BattleChess is a bit stronger). This is a BeebEm uef (savestate) file so that you can replay the game from the start.
I'm interested to see what would happen pitting Colossus 4 on the Master against Colossus 4 on the C64 or the Speccy. But I suppose a realistic comparison would take dozens of games, though under emulation (which I'm using), it's a lot more manageable.
Back in the day I played White Knight Mk11, and I still remember the little cheery beep when it was your turn to play. Then I upgraded to Colossus 4 on the Beeb. At my peak (sadly, long past) I could beat WK11 on its basic setting (averaging 10s per move of computer thinking time). Colossus was the chess engine I thought about most--and I realised it used two different calculations to evaluate the position: a Material calculation ("Mtrl"), which used standard numerical values for pieces, and a Positional calculation ("Psnl"). Modern engines typically evaluate the position using a combined score in centipawns, which allows them to sacrifice material for strong positional advantage.
I emailed Martin Bryant to ask him about the details of his evaluation function, but he said he couldn't remember, and no longer has access to the source code for Colossus. What I wanted to know is whether Colossus would sacrifice material for positional advantage, and what the trade-off would look like. According to Wikipedia, he retired in 2020, and his website is now archived (though it's accessible via the WayBack Machine) https://en.wikipedia.org/wiki/Martin_Br ... rogrammer).
Colossus was eventually released for the Amiga and PC, as Colossus X. I didn't like the GUI, which I thought was pretty garish. And I suppose the "ultimate" version of Colossus is the UCI-compliant engine which Martin released in 2008 (thoroughly tweaked for some pretty impressive strength upgrades).
What I've had some fun doing lately is pitting Colossus against some other chess programs. Obviously Colossus 4 is pretty strong right out of the box, and I wanted to see what it could come up with on its standard setting. Here's an example of it demolishing Battle Chess Level 6 (the PC version, since I think the Amiga version of BattleChess is a bit stronger). This is a BeebEm uef (savestate) file so that you can replay the game from the start.
I'm interested to see what would happen pitting Colossus 4 on the Master against Colossus 4 on the C64 or the Speccy. But I suppose a realistic comparison would take dozens of games, though under emulation (which I'm using), it's a lot more manageable.
Statistics: Posted by aidanthegasman — Sun Jul 07, 2024 1:11 am