Not logged inCSS-Forum
Forum CSS-Online Help Search Login
CSS-Shop Impressum Datenschutz
Up Topic Hauptforen / CSS-Forum / J92-1xx im Vergleich
- - By Lothar Jung Date 2020-09-14 16:24
Hier ein Test:

**LC0-version:** v0.26.2   **J92-100 vs (J92-145, J92-160)**
**LC0 options:** `-nncache=6 000 000 --syzygy-paths=E:\syzygy`
**Time control:** 1600 nodes
**Hardware:** RTX 2070
**Book:** 6ply 1000-openings by Chad, sequential, with reverse colours.
**Tablebases:** syzygy 6man
**Adjudication:**
`-resign movecount=5 score=900
-draw movenumber=50 movecount=6 score=10
-tb E:\syzygy`
**Software:** cutechess-cli
```
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 lc0-v0.26.2.J92-160    :       9      7   512.5    1000  51.3      70  125   775  100  77.5
   2 lc0-v0.26.2.J92-145    :       5      7   507.5    1000  50.8      86  112   791   97  79.1
   3 lc0-v0.26.2.J92-100    :       0      5   980.0    2000  49.0     ---  197  1566  237  78.3

White advantage = 44.77 +/- 3.76
Draw rate (equal opponents) = 81.73 % +/- 0.97
```

Nicht viel, aber immerhin.

Lothar
Parent - - By Reinhold Stibi Date 2020-09-14 16:42
Hallo Lothar,

morgen setze ich beide RTX 2070 in den Computer ein.

Möchte zuerst mit den Grundeinstellungen testen und danach mit Deinen Settings.

Was muss ich dazu einstellen, dass beide Karten erkannt werden ?

Werde zuerst mit dem Netz J92-145 testen.

LG und schöne Tage in Frankreich

Reinhold
Parent - By Lothar Jung Date 2020-09-16 11:03 Upvotes 1
Nach dem letzten Test soll J92-160 das beste Netz sein:

**LC0-version:** v0.26.2   **J92-100 vs (J92-145, J92-160, J92-170)**
**LC0 options:** `-nncache=6 000 000 --syzygy-paths=E:\syzygy`
**Time control:** 1600 nodes
**Hardware:** RTX 2070
**Book:** openings-dd-6ply-500 by Chad, sequential, with reverse colours.
**Tablebases:** syzygy 6man
**Adjudication:**
`-resign movecount=5 score=900
-draw movenumber=50 movecount=6 score=10
-tb E:\syzygy`
**Software:** cutechess-cli
**Comments:** NOTE: I've changed old Chads openings-6ply-1000 to new openings-dd-6ply-500 for J92-170 games
```
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 lc0-v0.26.2.J92-160    :       9      8   512.5    1000  51.3      69  125   775  100  77.5
   2 lc0-v0.26.2.J92-145    :       5      8   507.5    1000  50.8      58  112   791   97  79.1
   3 lc0-v0.26.2.J92-170    :       4      8   505.5    1000  50.5      78  117   777  106  77.7
   4 lc0-v0.26.2.J92-100    :       0      4  1474.5    3000  49.1     ---  303  2343  354  78.1

White advantage = 46.43 +/- 3.00
Draw rate (equal opponents) = 81.75 % +/- 0.80
```
Parent - By Lothar Jung Date 2020-09-17 14:59 Upvotes 1
Jetzt wohl T92-180:

**LC0-version:** v0.26.2   **J92-100 vs (J92-145, J92-160, J92-170, J92-180)**
**LC0 options:** `-nncache=6 000 000 --syzygy-paths=E:\syzygy`
**Time control:** 1600 nodes
**Hardware:** RTX 2070
**Book:** openings-dd-6ply-500 by Chad, sequential, with reverse colours.
**Tablebases:** syzygy 6man
**Adjudication:**
`-resign movecount=5 score=900
-draw movenumber=50 movecount=6 score=10
-tb E:\syzygy`
**Software:** cutechess-cli
**Comments:** Starting J92-170 test, I've changed old Chads openings-6ply-1000 to his new openings-dd-6ply-500
```
   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 lc0-v0.26.2.J92-180    :      12      9   516.5    1000  51.6      65  131   771   98  77.1
   2 lc0-v0.26.2.J92-160    :       9      9   512.5    1000  51.3      69  125   775  100  77.5
   3 lc0-v0.26.2.J92-145    :       5      9   507.5    1000  50.8      58  112   791   97  79.1
   4 lc0-v0.26.2.J92-170    :       4      9   505.5    1000  50.5      77  117   777  106  77.7
   5 lc0-v0.26.2.J92-100    :       0      4  1958.0    4000  49.0     ---  401  3114  485  77.8

White advantage = 47.81 +/- 2.72
Draw rate (equal opponents) = 81.72 % +/- 0.68
```
Parent - - By Lothar Jung Date 2020-09-21 08:12
Hallo Reinhold,

J92-190 ist verdammt nah dran.

J92-205 wird gerade getestet.

**Match**: J92-100, J92-180, J92-190 vs stockfish 12 in fixed nodes
**LC0 version:** 0.26.2, built with cuda11 + cudnn7
**LC0 options:** cudnn-fp16,  1 thread, default for the rest
**SF options:** 1 thread, 1GB Hash.
**Time control:**  Leela: 1 kn/move, SF 12: 400Kn/move
**Hardware:** RTX 2070, i7-7700
**Book:** Balsa 2724 openings
**Tablebases:** 6-man TB for both.
**Adjudication**: 6-man TB, -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
**Software:**cutechess-cli
**Comments:** J92-180 and J92-190 are clearly better than J92-100. The two are very close with slightly edge for J92-190. Next: J92-205.

```diff
Leela: 1 thread, 1Kn/move
SF 12: 1 thread, 400Kn/move

Balsa 2724

   # PLAYER                :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W      D     L
-  1 Stockfish_12.400k     :       0   ----  8372.0   16344  51.2      89  3298  10148  2898
+  2 lc0.net.J92-190.1k    :      -4      6  2696.0    5448  49.5      73   996   3400  1052
+  3 lc0.net.J92-180.1k    :      -6      6  2676.5    5448  49.1      99   992   3369  1087
+  4 lc0.net.J92-100.1k    :     -16      5  2599.5    5448  47.7     ---   910   3379  1159

White advantage = 36.14 +/- 1.72
Draw rate (equal opponents) = 63.05 % +/- 0.37
```

Lothar
Parent - - By Horst Sikorsky Date 2020-09-21 11:03
LCO bewertet sich schon seit ewigen viel zu niedrig im Gewinnsinn und Verluststellung.
Merkt er nichts mehr 
Parent - - By Thomas Lagershausen Date 2020-09-21 14:49
Du solltest eine zweite Karriere als Schachtrainer starten.

Erst dann verstehen die Computer Schach.
Parent - - By Horst Sikorsky Date 2020-09-21 16:18
vor 25 Jahren habe ich in Landsberg Saalekreis 3-4 Monate Grundschüler trainiert.
Sie haben in Halle/ Saale im 4er-Mannschaftsturnier als (Neulinge) den 3.Platz von 10 belegt.
Thema Verfehlt.
Parent - - By Thomas Lagershausen Date 2020-09-22 08:47
Sonst noch irgendwelche WM-Kandidaten die Du trainiert hast?
Parent - By Horst Sikorsky Date 2020-09-22 12:37
  
Up Topic Hauptforen / CSS-Forum / J92-1xx im Vergleich

Powered by mwForum 2.29.3 © 1999-2014 Markus Wichitill