MadChess 1.0 Released (C# .NET Engine)

[emadsen]

Hi George,

I've been reading this forum for a number of months as I worked on my engine. I know you're an active tester. So it's nice to hear you say good things about my engine. I'm glad people are interested in weaker engines.

I assume you meant 3.5 .net OR LATER.

Yes, you are correct. I've updated my website and zip files with the phrase "or later" to clarify this.

And you leave me no way to adjust [hashtable size] higher, being UCI.

UCI does allow adjustment of hash size. I know the Shredder GUI has a setting for hash size when configuring a match or tournament. Not sure about the Fritz GUI.

However, I will say my hash implementation is very basic. I did not spend much time on it. I have a sense I'm not using the hash as often as I should. Like you say, hash usage grows slowly. I just guessed at how many positions fit into one MB of memory- it is not well calibrated. I've added a note to my to-do list to improve this in the next version.

Very little impresses me any longer, and nothing excites me- but your engine does. In the 4+2 blitz I played with it ag. Marvin, which won Group 14- and is stronger than Micro-Max- your engine DESTROYED Marvin.

Glad to hear it! I like your enthusiasm George.

I'll download Marvin now and watch a few games...

[geots]

Hi George,

I've been reading this forum for a number of months as I worked on my engine. I know you're an active tester. So it's nice to hear you say good things about my engine. I'm glad people are interested in weaker engines.

I assume you meant 3.5 .net OR LATER.

Yes, you are correct. I've updated my website and zip files with the phrase "or later" to clarify this.

And you leave me no way to adjust [hashtable size] higher, being UCI.

UCI does allow adjustment of hash size. I know the Shredder GUI has a setting for hash size when configuring a match or tournament. Not sure about the Fritz GUI.

However, I will say my hash implementation is very basic. I did not spend much time on it. I have a sense I'm not using the hash as often as I should. Like you say, hash usage grows slowly. I just guessed at how many positions fit into one MB of memory- it is not well calibrated. I've added a note to my to-do list to improve this in the next version.

Very little impresses me any longer, and nothing excites me- but your engine does. In the 4+2 blitz I played with it ag. Marvin, which won Group 14- and is stronger than Micro-Max- your engine DESTROYED Marvin.

Glad to hear it! I like your enthusiasm George.

I'll download Marvin now and watch a few games...

I do not think Marvin is in development any longer- but that doesn't affect its strength. Even if you don't get the same results I did- I would still bet money in a 200 or 250 game match (absolutely no faster than 4m+2s) you would win by a good margin.) Just a bit of advice- as these are similar in average time per move. Run some test games at 4m+2s- but also test it good at 40/3 repeating and makes sure it understands and manages its time well in both. Because when Ingo with IPON tests it- he will use incremental controls. And CCRL and CEGT for the most part use repeating. You want to do well in both. Just a thought, and good luck. I will be testing it a bit for sure.


Best,

george

[Graham Banks]

...... Because when Ingo with IPON tests it- he will use incremental controls

I think that Ingo usually only tests engines that could be considered to be in the top twenty. I'm sure he'll correct me if I'm wrong.

Still a good idea though for an engine to understand both repeating and incremental time controls as you've suggested.

[geots]

...... Because when Ingo with IPON tests it- he will use incremental controls

I think that Ingo usually only tests engines that could be considered to be in the top twenty. I'm sure he'll correct me if I'm wrong.

Still a good idea though for an engine to understand both repeating and incremental time controls as you've suggested.

You are right about Ingo. I realized that when I typed the thread. But with Leo gone, and Olivier gone as well- I really could not come up with another example. And who knows, this time next year MadChess may be there


gts

[geots]

I am quite sure you don't care a bit about what I may be interested or not interested in. But it was sort of my dream to take over for either Leo or Olivier, however there is just no way I can handle either alone. CPUs are not the problem. I got plenty of them, and they still sell 'em.

[Graham Banks]

I am quite sure you don't care a bit about what I may be interested or not interested in. But it was sort of my dream to take over for either Leo or Olivier, however there is just no way I can handle either alone. CPUs are not the problem. I got plenty of them, and they still sell 'em.

I think that what you're doing is great for amateur engine authors and enthusiasts.

The two key things are that you must enjoy what you're doing, and it must also be sustainable if you're intending it to be a long term project.

I know that you are always very thorough in the way that you set up and run your tests.
You might not be able to manage it if you don't have a copy of Chessbase installed (or a similar tool), but if you can, quickly check every game to make sure that nothing has gone wrong (engines accessing their internal books, time losses, issues using the time control correctly, incorrect results, etc).

I follow your testing, just as I follow the testing efforts of others.

[emadsen]

Run some test games at 4m+2s- but also test it good at 40/3 repeating and makes sure it understands and manages its time well in both.

Good point. I've only tested with Fischer time controls, though I've written code for both Fischer and traditional time controls. I'll do more testing with traditional controls.

who knows, this time next year MadChess may be there

I'll set that as a stretch goal. A more realistic goal is to reach master strength (2200 ELO).

[ZirconiumX]

who knows, this time next year MadChess may be there

I'll set that as a stretch goal. A more realistic goal is to reach master strength (2200 ELO).

1111 ELO in 1 year is no mean feat.

Still, I wish you the best of luck.

Matthew:out