CCRL Discussion Board

Hello, Mida 2.3.1 is now avaliable. The main update is the fix of a bug in Move generation, which should be now 100% reliable. Also, there's a small upgrade in Reverse futility pruning. I don't expect, however, significant changes in Elo strength. Thanks for the attention :) https://github.com/Giaco...

Hi Gabor,
sorry for the late response. I will look into it, I changed some code in time management to make it compatible with all OS, so it's possible.
Thanks

Thanks! I also wanted to say I solved the issue of the NNUE file being unnecessairly created every time (that was pointed out in a previous release).
I've updated v2.3 once more with this small fix. Hopefully that's the last issue I find for now

Edit: the fix was surprisingly easy (thanks to the Berserk issue you provided) so I have already updated the release with the fixed code that now works on my end. Thanks for pointing this out, I've been scratching my head around this problem for a while

That’s some good news! I’ll try to fix the issue when i can.
Thanks for the useful info!

Yes, I'm aware of the problem. The issue seems to be complicated, though, because it appeared out of nothing without me changing anything in the actual CLI code. I have experienced the same problem with Berserk 12 (already with 11.1) (maybe it's the same engine you talked about), and I was "rel...

Hello, I'm happy to announce the release of Mida v2.3 ! The updates in the engine itself are small, and should give a +30 ELO gain against v2.2. The greatest update is the makefile, that now makes Mida available on all OS, and also provides optimizations according to the hardware architecture. So, i...

Sorry for the late response. That’s awesome, thanks

No worries, the job you are doing is already great and much appreciated

Hi! I'm Happy to announce the realease of Mida v2.2, which comes with a lot of improvements in search and in transposition entries. Note: Hash size is now an option, and default is set to 64 MB. These are the results I got against v2.1: // Score of v2.2 vs v2.1: 118 - 16 - 66 [0.755] // ... v2.2 pla...

Hello, I'm happy to announce the release of Mida 2.1 . I want to thank Graham Banks for his help thanks to which I can finally provide a .exe that can run on machines withoug GCC installed. I know this release comes shortly after v2.0, but the changes are pretty consistent, and this is reflected on ...

Graham Banks wrote: ↑Sat Sep 16, 2023 10:30 am Won't be tested for 40/15 without the DLL files.

All right, I'll try to add DLL files to v2.1

The net is trained by myself. However, the search doesn't include many new features, so the ELO gain won't be great. I'm working on v2.1 which at the moment sits around 2900 ELO

That's just the implementation I went for. It is also used by CFish https://github.com/syzygy1/Cfish as I saw from the code so I assumed it was OK.

Is something missing? On my PC the .exe compiles and runs fine for v2.0

That’s good to hear. To be honest I didn’t even test it because improving wasn’t the goal. Seeing a reduction in the nps by around 10%, I assumed it would be worse, even though I made some changes to the search. Anyway, the improvements with v2.0 will defenitely be greater

Oh, sorry about that, really bad timing by me :? Anyway, v1.2.1 will probably have a Elo loss compared to the previous ones, since it was more of a code refactoring which resulted in a slight performance drop. You can skip it completely or, if you have already tested it, you can skip it in the futur...

Hello, I'm happy to announce the release of Mida 2.0 This is a big step forward from the previous versions, since NNUE was introduced. The network was trained using leela training data, but very quickly, since I want to focus on the search for the time being. The rating range should be around 2600, ...

All right, thanks

Thanks as always. About the number of games, I don’t think I have the computing power necessary for 60000 games, unless I use some ridiculously small time controls. If that’s the case, can I ask what time controls do you use for such a big number of games?

I'm happy to announce the release of Mida v1.2, which should be aruond 2350-2400. I thank Graham Banks and Gabor Szots for testing the previous version and making me realize it wasn't where I thought it was. I worked hard and introduced some novelties that should bring this version where I hoped. I ...

all right, thanks!

All right, have to say I hoped for better, since I also played some games in Arena between the 2 versions and the result was 13-2 and 5 draws. Anyway, is it a problem using this new version on the CCRL or is it better if the versions have a bigger difference?

I hope around 2350

Good morning, I'm happy to announce that the new version of Mida chess engine is now released! You can find it here https://github.com/GiacomoPorpiglia/Mida . It would be appreciated if this version was now tested on the CCRL. I also notice that the previous version was only tested on blitz format. ...