DOTS3

[dorsz]

Dynamic Openings Test Suite (DOTS) is a set of selected positions for running engine matches. The purpose of this set is not to measure Elo differences but to provide dynamic and interesting games.

DOTS3 has 100 positions: 50 from DOTS2 + 50 new.

https://www.mediafire.com/file/0ogyy3n9 ... 1.zip/file

[Chess_Dragon]

Code: Select all

Exhibition, for example the +-'s are higher than the Elo's.
DOTS3 random reversed, 48 game RR.
G/6'+1" ; 2Gh Sy.6 ; 2700X 2060
                            Elo   +   -    Games   Score     Data     Draws
  1 Blue Marlin 14.3-avx2 :  44  94  92   19.0/32  59.4%  +12 -6 =14  43.8%
  2 Swordfish 14.3-avx2   :   0  87  87   16.0/32  50.0%  +8  -8 =16  50.0%
  3 Lc0-0.28.0_610069     : -44  108 110  13.0/32  40.6%  +9 -15 =8   25.0%

Already worse Lc0 lost very quickly with its 25...g5??. Blue Marlin drew the reversed.

Blue Marlin won the ending, in the reversed it got counterplay and drew.

[Chess_Dragon]

Thank you again for your DOTS3 !!

Do you mind if I mention it on Discord ?

Code: Select all

[Event "G/75''+0.8'' 2Gh Sy.6 | DOTS3 each"]
[Site "R7 2700X 16threads | RTX 2060"]
[Date "2021.09.23"]
[Round "32"]
[White "Stockfish_21092119"]
[Black "Lc0-0.28.0_610069"]
[Result "1-0"]
[FEN "rnbqkb1r/pp3ppp/4pn2/1BpP4/3P4/8/PPPN1PPP/R1BQK1NR b KQkq - 0 1"]
[GameDuration "00:01:26"]
[GameEndTime "2021-09-23T01:20:51.537"]
[GameStartTime "2021-09-23T01:19:25.467"]
[PlyCount "41"]
[SetUp "1"]
[Termination "adjudication"]
[TimeControl "75+0.8"]

1... Bd7 {-0.58/15 1.3s} 2. Bxd7+ {+1.70/26 2.8s} Qxd7 {-0.60/15 1.2s}
3. dxe6 {+1.75/27 1.3s} Qxe6+ {-0.61/14 1.1s} 4. Ne2 {+1.75/27 1.4s}
Bd6 {-0.63/14 2.4s} 5. Nf3 {+1.68/28 4.0s} O-O {-0.66/11 4.7s}
6. O-O {+1.70/28 1.2s} Nc6 {-0.68/11 0.80s} 7. Re1 {+1.66/28 1.9s}
Rfe8 {-0.74/12 4.8s} 8. Bg5 {+1.71/28 1.8s} Qf5 {-0.72/12 0.95s}
9. c3 {+1.73/30 2.1s} cxd4 {-0.82/13 2.7s} 10. Bxf6 {+1.74/30 1.6s}
Qxf6 {-0.84/13 0.80s} 11. Nexd4 {+1.74/29 2.0s} Rxe1+ {-0.84/12 1.4s}
12. Qxe1 {+1.72/28 1.6s} Nxd4 {-0.84/11 1.3s} 13. Nxd4 {+1.77/28 2.6s}
Rd8 {-0.84/10 6.2s} 14. Rd1 {+1.79/28 2.3s} Bf8 {-0.83/10 1.5s}
15. g3 {+1.73/28 6.3s} Qa6 {-0.81/10 1.2s} 16. a3 {+1.81/26 1.1s}
Qb6 {-0.75/10 2.0s} 17. Rd2 {+1.80/27 2.2s} g6 {-0.73/9 1.2s}
18. Qe3 {+1.82/28 2.0s} Bg7 {-0.65/9 2.9s} 19. Nf5 {+5.49/30 1.7s}
Rxd2 {-5.66/17 3.0s} 20. Qe8+ {+7.38/23 1.6s} Bf8 {-7.22/1 0s}
21. Nh6+ {+7.46/23 1.6s} Kg7 {-7.80/1 0s, White wins by adjudication} 1-0

[dorsz]

Why do you want to mention it on Discord?

[Chess_Dragon]

I was thinking of posting a Sf vs Lc0 there.

But, if I do I can use a different book.

---------------------------------------------------

( I like the way DOTS3 gets to the critical, typically:
same opening:
stronger 1-0 weaker
weaker 1/2-1/2 stronger
in a low number of moves often. )

[dorsz]

No problem for me, just wanted to know. Of course you can use it.
---------------------------------------------------
As you are the main user of this book, I need your advice. Since releasing DOTS3 I am working on DOTS4. What is better, to release it with 150 positions or to wait until the number of positions is 200? At the moment DOTS4beta has 121 positions (99 from DOTS3, 1 deleted because of too many quick draws, 22 new).

[Chess_Dragon]

Wow! DOTS4 sounds interesting already!

For me: I use the DOTS with the latest engines, so:
(DOTS4 ?) 150 with the latest engines/nets at that time,
(DOTS5 ?) 200 with the newer latest engines/nets (I would not have any 'duplication', my engines/nets for the 200 would be newer than the ones I used with the 150)
so, for me, I would prefer the 150.

Thank you for asking!

[dorsz]

As you wish - new version every 50 positions added.

[Chess_Dragon]

Thanks! I have to catch up on a few things then I will try your DOTS and 'fish' again.

I am not going to do what I was thinking of for Discord.
It was too similar to the below, there is already a lot of data about it

Code: Select all

Lc0 T60 http://zz4032.bplaced.net/lc0/charts.html
http://training.lczero.org/networks/1 
     | vs Stockfish NCM
     | https://nextchessmove.com/dev-builds
                 
608000 2021-03-12 3053 +00 | Sf 20210310-1132 373.13 +00.00
608500 2021-04-10 3059 +06 | Sf 20210406-1623 378.63 +05.50
609000 2021-05-06 3067 +14 | Sf 20210507-0745 381.31 +08.18
609500 2021-06-21 3069 +16 | Sf 20210621-2116 421.00 +47.87
610000 2021-08-27 3081 +28 | Sf 20210827-1941 440.35 +67.22

[dorsz]

I sometimes look for NCM Stockfish tests. You can find similar tests here too: https://www.sp-cc.de/stockfish-regression.htm

(My new fishes 14.4 are going to be released tomorrow).