Bikjump V3.0 is now available for download

[Bart Weststrate]

You should call 3.0 Bigjump. Score SM against 2.1 until now: 189-20-21 [0.867] so the difference from 2.1 to 3.0 is gigantic against the beancounter.

[Bart Weststrate]

3rd match 60/20+1:

Score of SM_2182b vs Bikjump21: 411 - 42 - 47 [0.869] 500
... SM_2182b playing White: 207 - 19 - 24 [0.876] 250
... SM_2182b playing Black: 204 - 23 - 23 [0.862] 250
... White vs Black: 230 - 223 - 47 [0.507] 500
Elo difference: 328.7 +/- 40.6, LOS: 100.0 %, DrawRatio: 9.4 %

So difference Bikjump 3.0 vs 2.1 against the beancounter Bikjump 3.0 +- 188 elo stronger than 2.1.

Shortest win:

[pgn][Event "?"]
[Site "?"]
[Date "2026.04.11"]
[Round "224"]
[White "SM_2182b"]
[Black "Bikjump21"]
[Result "1-0"]
[ECO "B00"]
[GameDuration "00:00:44"]
[GameEndTime "2026-04-11T13:18:29.801 W. Europe Daylight Time"]
[GameStartTime "2026-04-11T13:17:45.650 W. Europe Daylight Time"]
[Opening "King's pawn"]
[PlyCount "42"]
[Termination "adjudication"]
[TimeControl "60/20+1"]
[Variation "Nimzowitsch defense"]

1.e4
{book}
1...Nc6
{book}
2.d4
{book}
2...e5
{book}
3.dxe5
{book}
3...Nxe5
{book}
4.f4
{book}
4...Nc6
{book}
5.Bc4
{+0.10/10 0.64s}
5...Qe7
{-0.07/8 1.0s}
6.Nc3
{+0.25/13 0.64s}
6...f5
{+0.05/9 1.3s}
7.Nf3
{+1.14/11 0.75s}
7...fxe4
{-0.12/9 1.3s}
8.Ng5
{+1.66/12 1.0s}
8...Nh6
{-0.55/10 2.4s}
9.Nd5
{+2.02/13 2.3s}
9...Qd8
{-0.79/9 0.75s}
10.Nxe4
{+2.29/12 1.3s}
10...Be7
{-0.39/8 0.84s}
11.Qh5+
{+3.57/12 0.94s}
11...Kf8
{-1.77/10 3.2s}
12.f5
{+3.44/12 0.74s}
12...Nf7
{-2.53/9 3.2s}
13.f6
{+4.18/12 1.1s}
13...Bxf6
{-2.44/9 1.3s}
14.Ndxf6
{+5.93/13 2.0s}
14...Nce5
{-3.75/9 1.3s}
15.Bxf7
{+6.99/13 1.1s}
15...Nxf7
{-6.22/10 1.3s}
16.Nxh7+
{+10.80/14 2.0s}
16...Ke8
{-6.22/11 1.3s}
17.O-O
{+14.60/14 0.64s}
17...Qe7
{-12.10/9 1.3s}
18.Bg5
{+15.51/14 1.3s}
18...d5
{-12.61/8 0.90s}
19.Bxe7
{+16.73/14 1.2s}
19...Be6
{-15.64/9 1.3s}
20.Nc5
{+17.87/13 0.93s}
20...Rxh7
{-17.82/10 1.2s}
21.Qxh7
{+20.83/13 0.88s}
21...Kxe7
{-18.97/9 0.82s, White wins by adjudication}
1-0[/pgn]

[abik]

So difference Bikjump 3.0 vs 2.1 against the beancounter Bikjump 3.0 +- 188 elo stronger than 2.1.

SWEET! Thanks for running the tests, Bart, much appreciated!

The improvements from v2.0.1 to v3.0 were merely focused on faster move generation (visible when running perft) and improved pruning techniques (visible with higher search depths; ironically but obviously this hides the reported move generation speed improvements). For version v3.1, I plan to improve the evaluation function by putting all the knowledge of the Dutch "JeugdSchaak" series in my engine . I never became a strong player myself, but hopefully the accumulated knowledge into BikJump will account for something. Stay tuned!

[abik]

FWIW, these were my self play results, so with the usual caveat of rating engines against older versions, I was hopeful on at least +100 ELO.

Code: Select all

--------------------------------------------------
Match   : ./bikjump201 vs. ./bikjump300
Games   : 400, movetime: 100ms
Openings: 200 (selected) 200 (available)
--------------------------------------------------
FINAL: eng1 wins=83 eng2 wins=285 draws=32
Score: eng2 301.0/400 = 75.2%
Approx Elo diff eng1 -> eng2: +193