By Lothar Jung
Date 2020-01-11 14:26
Edited 2020-01-11 14:40
Jetzt auch unter TC gegen SF10 und T40.
Von 256x20-T40-1541 (SV strongest 20b net) nur 21 Elo entfernt.
**Match:** Gauntlet SF10 vs 61900
**Hardware:** Geforce RTX 2070S & i7-5960X (8 cores/16 threads)
**Lc0 version:** v0.23.2
**Lc0 options:** --backend=cudnn-fp16
**SF9 options:** Threads=13 Hash=512
**SF10 options:** Threads=14 Hash=512
**Speed:** Lc0:13.4 kN/s, SF9:12317 kN/s, SF10:12401 kN/s (from benchmarks)
**Time control:** 15s+0.5s
**Book:** openings-8ply-10k.pgn, random, color reversed
**Tablebases:** 6-man syzygy
**Adjudication:** 6-man TB, -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
**Software:** cutechess-cli
**Comments:** I would be careful about drawing any conclusions about 61900 vs J13B.2-188 from this as they have not played each other directly.
```diff
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L
+ 1 lc0.net.61900 : 4.92 20.77 408.5 750 54.5 68 196 425 129
2 lc0.net.J13B.2-188 : 0.00 ---- 1128.5 2000 56.4 100 563 1131 306
- 3 stockfish_10 : -27.23 12.59 1582.0 3250 48.7 72 617 1930 703
4 lc0.net.61800 : -32.01 20.41 370.0 750 49.3 88 131 478 141
5 lc0.net.61300 : -48.05 20.70 393.5 750 52.5 56 184 419 147
6 lc0.net.61700 : -49.61 15.63 743.5 1500 49.6 87 313 861 326
7 lc0.net.61600 : -61.94 20.85 379.0 750 50.5 61 170 418 162
8 lc0.net.61500 : -65.29 20.80 375.5 750 50.1 52 179 393 178
9 stockfish_9 : -65.76 12.84 3407.5 7000 48.7 72 1471 3873 1656
10 lc0.net.61200 : -70.55 20.31 370.0 750 49.3 61 158 424 168
11 lc0.net.61100 : -73.90 20.74 366.5 750 48.9 52 157 419 174
12 lc0.net.61400 : -74.38 20.60 366.0 750 48.8 70 154 424 172
13 lc0.net.61000 : -80.61 21.19 359.5 750 47.9 --- 154 411 185
White advantage = 54.02 +/- 2.30
Draw rate (equal opponents) = 58.60 % +/- 0.51
```
**Match** Round-Robin
**LC0-version:** v0.23.1
**LC0 options:** default
**Time control:** 100,000 nodes/move for 42767 vs. 66667 nodes/move for 61902
**Hardware:** Titan RTX and Intel®️ Xeon®️ Gold 6142 Processor
**Book:** aloril_opening_suite with reverse
**Tablebases:** 6 man Syzygy
**Adjudication:** movecount=4 score=400
**Software:** Cutechess
**Comment:** Possibly T60 reaches or exceeds level of late T40 nets at MTC with T60 at 1/3 less nodes. Curiously T60 suffers all of the 5 losses as white.
**Results:**
```
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%)
1 61902 : 5.4 13.7 203.0 400 51 78 38 330 32 83
2 LC_42767 : 0.0 ---- 197.0 400 49 --- 32 330 38 83
White advantage = 52.98 +/- 7.24
Draw rate (equal opponents) = 90.19 % +/- 2.20
**Match**: 61915 vs 256x20-T40-1541 (SV strongest 20b net)
**LC0 version:** 0.23.2
**LC0 options:** cudnn-fp16, *1 thread, cpuct=1.45, fpu=0.24, pst=1.85, cpuct-base=5000*
**Time control:** 1Kn/move
**Hardware:** RTX 2070
**Book:** SuperGM_4mvs_500 book, in sequence, reversed color
**Tablebases:** 6-man TB
**Adjudication**: 6-man TB, -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
**Software:**cutechess-cli
**Speed:** Average nps from the pgn file: T40B-1541: 17.5 knps, T60: 9.2 knps
**Comments:** 61915 vs T40-1541: +130 -191 =679 (-21 elo): T60 progresses quite steadily.
```diff
1 thread, vs 256x20-t40-1541 at 1 Knpm, cpuct=1.45, fpu=0.24, pst=1.85, cpuct-base=5000
# PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L
1 lc0.net.J13B.2-136 : 2 11 503.5 1000 50.4 66 134 739 127
- 2 lc0.net.256x20-t40-1541 : 0 ---- 2681.5 5000 53.6 100 970 3423 607
+ 3 lc0.net.61915 : -21 12 469.5 1000 47.0 84 130 679 191
4 lc0.net.61900 : -30 12 457.5 1000 45.8 80 121 673 206
5 lc0.net.61884 : -37 12 447.0 1000 44.7 68 113 668 219
6 lc0.net.61866 : -42 12 441.0 1000 44.1 --- 109 664 227
```
```
Sehr interessant ist zur Zeit das Tunier von Endosani:
Houndini und Komodo sind abgeschlagen und spielen keine Rolle.
Die beiden T60-Netze schlagen sich gut, über 50 %.
Stärkstes T58-Netz spielt erstaunlich und hat SF dev geschlagen.
Der Benchmark ist LS13, SV-Netze und SF10 dev/SF11.
T60 könnte in diese Gruppe hineinwachsen.