cool i can't wait to see how it go's in anti ( would be cool to watch it play against TeamMachina when it's better ) and thx for the update cus i was wondering how it was going
Ziva Zero Progress Updates.
Sort:
@TheUltraTrap right now source code is private, and may become public/open source in the future.
Mar 3, 2021
0
#5
@GTSWPM150 Do you think you will be able to make Ziva Zero an online 4 player chess engine? It would really help out for those who want to analyze their games without having to download the program or do whatever else needs to be done in order to use it.
sick dude, gg, keep it up :0
@JkCheeseChess Yes I believe this is where Ziva could eventually find itself. Once I have a extremely good Ziva (I might partner it with chess.com to make it available for you all via some tensorflow.js (don't worry about all the technicalities xd)). Only time will tell.
@HighEldar thanks for the support.
WOOOO IM EXCITED!!
Mar 4, 2021
0
#10
I recommend scratching the antichess stuff and concentrating on standard mode. Really like that you're doing the omatamix setup as well. good luck
Mar 4, 2021
0
#11
Well antichess is really hard so if he makes it work its a real feat
To keep you updated on the progress of Ziva Zero I will be updating this Forum with current training of the following modes: 4 player chess Teams Standard, 4 player chess Teams Antichess, and 4 player chess Teams Omatamix setup.
Possibilities for FFA training of these modes in the future.
Standard progress:
Ziva shows understanding of promotion, along with okay opening setups, but plays very poorly in the middle game (basically like a rando - should improve as more training data is acquired).
Antichess progress:
Ziva shows remarkable progress in learning openings and middle game play, endgame play is lacking, openings are everywhere. Openings should solidify as more training data is acquired.
Omatamix Setup progress:
Ziva hasn't trained on omatamix position yet. (I can only train 2 setups at a time with my resources, so this will be on the back burner). Making small logic changes to standard then will train this as well (will allow for castling for b-g as well).
Here is a training game Ziva played today (Antichess PGN 4):
[Variant "Teams"]
[RuleVariants "EnPassant Antichess"]
[CurrentMove "86"]
1. d2-d4 .. b10-d10 .. f13-f11 .. Nn10-l9
2. d4-d5 .. Ba9xe13 .. Bf14xBe13 .. Nn5-l4
3. d5-d6 .. b7-c7 .. Be13xm5 .. Nl4xk2
4. Rk1xNk2 .. c7xd6 .. Bm5xBn6 .. Kn7xBn6
5. Rd1xd6 .. Ba6xf11 .. j13-j12 .. Nl9-j8
6. Rd6xm6 .. Bf11xh13 .. Bi14xm10 .. Bn9xBm10
7. Rm6xKn6 .. Bh13xKg14 .. Qh14xBg14 .. Rn4xRn6
8. e2-e4 .. Qa7-b7 .. Qg14xm8 .. Qn8xQm8
9. Bf1xb5 .. Qb7xe4 .. i13-i11 .. Qm8xg2
10. Bb5xRa4 .. Qe4xNe14 .. Rd14xQe14 .. Qg2xf2
11. Qg1xQf2 .. Na10-c11 .. Re14-f14 .. m4-l4
12. Qf2xb6 .. Nc11xd13 .. Rf14-g14 .. Rn6xQb6
13. Ba4-b5 .. b11-d11 .. k13-k11 .. Rb6xBb5
14. h2-h4 .. Ra11-c11 .. i11-i10 .. Nj8xi10
15. Bi1xb8 .. Ka8xBb8 .. Rk14-k12 .. Bm10xRk12
16. h4-h5 .. Kb8-b7 .. Nj14xBk12 .. Rb5xh5
17. i2-i4 .. b4-c4 .. Nk12xm11 .. Ni10xk11
18. Rk2xNk11 .. Kb7-a8 .. Rg14-e14 .. Rn11xNm11
19. Rk11xd11 .. Rc11xRd11 .. Re14-f14 .. Rh5xKh1
20. Nj1-h2 .. Nd13xRf14 .. g13-g12 .. Rh1xNe1
21. Nh2-g4 .. Nf14xg12 .. j12-j11 .. Rm11xj11
22. j2-j4 .. S
Post Last updated: 3/2/2021