Skip to content
Pre-release

@Cscuile Cscuile released this Mar 1, 2019

AntiFish 1.9 is +38 Elo to AF1 based on the tests below. (Thanks to javiros for the test) AF1.9 is trained off of 253,039 AntiFish 1.0 vs SF11Dev with Cerebellum (2-8-19) games with a focus on MSE with 1,600 vs 100k nodes. Cyclical LR is used up to 24,000 steps for Mk313. AF1.9 is trained from AF1.0.

NOTE: After AF2, progress of AF will... slow. All activity after will likely be directory towards either automation of AF training, or Lc1.

Results:

   PLAYER                                  :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)  OppN
   1 AntiFish 1.9 24000 Steps                :    3035     28   150.5     274  54.9      67    84   133    57  48.5     1
   2 AF2_Run5CLR_v2_MSE_Focus-16000          :    3026     27   147.0     274  53.6      76    80   134    60  48.9     1
   3 AF2_Run5CLR_v2_MSE_Focus-49000          :    3012     27   140.5     272  51.7      50    75   131    66  48.2     1
   4 AF2_Run5CLR_v2_MSE_Focus-20000          :    3012     30   141.5     274  51.6      55    70   143    61  52.2     1
   5 AF2_Run5CLR_v2_MSE_Focus-32000          :    3009     28   140.5     274  51.3      52    68   145    61  52.9     1
   6 AF2_Run5CLR_v2_MSE_Focus-28000          :    3008     28   140.0     274  51.1      53    65   150    59  54.7     1
   7 AF2_Run5CLR_v2_MSE_Focus-23000          :    3006     28   139.5     274  50.9      63    70   139    65  50.7     1
   8 AF2_Run5CLR_v2_MSE_Focus-4000           :    3000     28   137.0     274  50.0      50    71   132    71  48.2     1
   9 SFDev2-8-19.exe                         :    3000      7  2192.5    4382  50.0      54  1079  2227  1076  50.8    16
  10 AF2_Run5CLR_v2_MSE_Focus-39000          :    2999     28   136.5     274  49.8      52    67   139    68  50.7     1
  11 AF2_Run5CLR_v2_MSE_Focus-7000           :    2997     28   136.0     274  49.6      50    69   134    71  48.9     1
  12 AntiFish_1.0_Mark_289_20500_Steps.pb    :    2997     29   136.0     274  49.6      64    61   150    63  54.7     1
  13 AF2_Run5CLR_v2_MSE_Focus-35000          :    2990     27   133.0     274  48.5      57    63   140    71  51.1     1
  14 AF2_Run5CLR_v2_MSE_Focus-41000          :    2986     28   131.5     274  48.0      57    62   139    73  50.7     1
  15 AF2_Run5CLR_v2_MSE_Focus-44000          :    2982     28   130.0     274  47.4      57    59   142    73  51.8     1
  16 AF2_Run5CLR_v2_MSE_Focus-15000          :    2978     27   128.5     274  46.9      80    59   139    76  50.7     1
  17 AF2_Run5CLR_v2_MSE_Focus-47000          :    2960     29   121.5     274  44.3     ---    53   137    84  50.0     1
Assets 3

@Cscuile Cscuile released this Feb 10, 2019 · 3 commits to master since this release

First off, I'd like to thank javiros and nameBrandon with the help in testing for both AF and the UTI Project (Universal Testing Initiative). While I am still unsure of how this network scales at higher TCs, this is the strongest network we currently have found. AntiFish 1.0 (Mark 289) has 20,500 Steps, 2 LRs (0.002 and 0.00002) a Policy Focus, and is trained off of 233,620 games.

Since I am now using 2 LRs, far more games are now required which means that progress will slow even further. The networks leading up to AntiFish 1.0 have not been saturated in training because of this as we would risk overfitting if we used more steps.

To Help with the AntiFish Project Please Visit:
AntiFish Contribution Channel Invite Link: https://discord.gg/chtSb2R

Thank you to everyone who contributed to this project!
AntiFish Contributors:
Fulgore
ASilver
Dkappe
joel
Additional Pylon
hv
PaleMale53
Matuiss2
Edosan
❤ u
Corzlyn
AlejoCarlsen
Sonicast
hx
Charis
Bladewood
coolchess123
Plutie
Filipe
javiros
nameBrandon

If I had missed anyone's name, I apologize. Please tell me and I will gladly add you name here.

Assets 3
Pre-release

@Cscuile Cscuile released this Feb 9, 2019 · 3 commits to master since this release

This network performs currently the best at 12.5M vs 3125 nodes against SFDev. Mk 275 is around +15 to +25 Elo to 32930 against SFDev.

Special thanks to javiros for helping me find this release version.

Assets 3
Pre-release

@Cscuile Cscuile released this Feb 6, 2019 · 3 commits to master since this release

This network does better at 3.2M nodes against SF. I am only reuploading this network for the record.

Assets 3
Pre-release

@Cscuile Cscuile released this Feb 6, 2019 · 3 commits to master since this release

I am seeing some major deviations between 3.2M and 12.5M testing. The network released here, Mark 242 does far better at 12.5M nodes against SFDev. Please refer to the Stockfish Discord for more information.

Assets 3
Jan 28, 2019
Update AntiFish_Game_Generation
You can’t perform that action at this time.