TalkChess.com

Reading the ICS chat this morning, it seems Minic has issues during some games. Sorry for the inconvenience. It seems to only happen at the very begining of games ? Is this what happened ? Minic not playing the first move ? In fact this is an error at the very end of the previous game. Will be fixed.

Minic 1.32 https://github.com/tryingsomestuff/Minic/releases (official release) Vivien always announces new versions of Minic in the thread titled "A complete 2000 lines of code engine". Yeah sorry, still using my "minic main topic". It is maybe better if I announce update here ? The 1.34 is out now.

Still, awesome improvement. And Minic is doing really well at TCEC, hope to see it promote to L2 :mrgreen: It won't unfortunately ! because of a late minute crash and a good games from opponent. The issue may be fixed by this new release : 1.32 I hope to see it in some bonus match like the FRC tour...

Well a lot of effort and CPU time has been involved in Minic debugging the last few days before QL from TCEC crew and other engine authors, so this is clear to me that L2 engines deserve the same good treatment at least.

Still, awesome improvement. And Minic is doing really well at TCEC, hope to see it promote to L2 :mrgreen: It won't unfortunately ! because of a late minute crash and a good games from opponent. The issue may be fixed by this new release : 1.32 I hope to see it in some bonus match like the FRC tour...

Minic has it quite early then lost it, and found it again stabilized 9 54 3 33587 23 883 0 c5c6 f8e8 a3a4 d3f1 g2f3 f1h3 b6a6 e8b8 b2a3 h3g4 10 46 5 44523 23 890 0 c5c6 d3f1 g2f3 f1h3 b6b4 g7g5 c6c7 f6g4 b4a5 ! 10 32 5 48713 23 902 0 d4d5 f8c8 c5c6 f6d5 b6a6 c8b8 c6c7 b8b2 c3d5 d3e4 g2h3 ! 10 14 8 7...

Thanks for this tourney Graham.

Another one, from Minic iCE game seems pretty sharp https://lichess.org/2MyRcPV7#46 [d] 2rr1k2/5Ppp/1p6/p2B4/n7/4P3/5PPP/R4RK1 w - - 1 24 Here Minic played Bb3, that was probably a bad decision ... Reanalyzing this position at home with more time, Minic goes for Bb7 but never stick to Be6 that is SF...

Using a neural net for anything devolves into some manner of alpha/leela methodology. In this case we are talking about having a policy network as found in the original alpha papers (for the first go engine, I believe) The policy network would generate a weight for how important a particular move (...

A simpler, but still pretty one from Marvin PeSTO game

[d] 1k1r2nr/pp1qn3/3p2p1/3Np1b1/4Pp1p/5P1P/PPPQNBP1/R2R2K1 w - - 6 20

where Bxa7 is a win