Bumblebee 1.0

[zd3nik]

https://github.com/zd3nik/BumblebeeReleases/releases

We all know the world doesn't need another chess engine. This one is a milestone for me simply because it is, or will be, the culmination of all the tinkering I've done with chess engines over the years.

So this will be my final engine. And as I have time I'll add all the finishing touches I have never bothered with before. Things like opening book support, endgame tablebase support, and SMP.

This initial release is just the base so it's pretty bare bones right now. But in the testing I've done so far it plays considerably stronger than the other engines I've released.

The source code for Bumblebee will remain private for the time being. But I suspect I'll make it open source at some point.

[Graham Banks]

........so far it plays considerably stronger than the other engines I've released.

How much stronger roughly?

[zd3nik]

I haven't done ELO calculations (I generally don't). But in the last blitz Round Robin I had it play the results against Bitfoot and Clubfoot were:

Bumblebee vs Bitfoot: 12 wins, 3 losses, 5 draws
Bumblebee vs Clubfoot: 18 wins, 1 loss, 1 draw

That was at 5 min 2 sec inc time control.

I'm about to start a new RR with a shorter time control, more engines, and 200 rounds instead of 20. It'll take several days to run though. I'll post the results when I have them if you'd like.

[Graham Banks]

I haven't done ELO calculations (I generally don't). But in the last blitz Round Robin I had it play the results against Bitfoot and Clubfoot were:

Bumblebee vs Bitfoot: 12 wins, 3 losses, 5 draws
Bumblebee vs Clubfoot: 18 wins, 1 loss, 1 draw

That was at 5 min 2 sec inc time control.

I'm about to start a new RR with a shorter time control, more engines, and 200 rounds instead of 20. It'll take several days to run though. I'll post the results when I have them if you'd like.

Sounds good.

[Roger Brown]

We all know the world doesn't need another chess engine.

Hello Shawn,

The world has not yet reached its limit for chess engine consumption.

Nowhere near that point in fact.

So feel free when bitten by the creative bug to publish away...

Later.

[Sylwy]

We all know the world doesn't need another chess engine.

WOW ! Sounds like: the world doesn't need another girl !
Well , this world definitely isn't for me..................

[zd3nik]

I haven't done ELO calculations (I generally don't). But in the last blitz Round Robin I had it play the results against Bitfoot and Clubfoot were:

Bumblebee vs Bitfoot: 12 wins, 3 losses, 5 draws
Bumblebee vs Clubfoot: 18 wins, 1 loss, 1 draw

That was at 5 min 2 sec inc time control.

I'm about to start a new RR with a shorter time control, more engines, and 200 rounds instead of 20. It'll take several days to run though. I'll post the results when I have them if you'd like.

Sounds good.

The tournament finished last night. BTW, I ran a gauntlet instead of a round robin.

It doesn't do as well in the shorter time control (2 min + 0.5 sec inc) as it did in the 5/2 time control, but it's still 80 ordo points above Bitfoot. I'm going to run a smaller tourney at a longer time control to verify.

Code: Select all

$ ./ordo -p Bumblebee.pgn 

File: Bumblebee.pgn
Loading data (2000 games x dot): 

.|

Total games                2400
 - White wins              1103
 - Draws                    335
 - Black wins               962
 - Truncated/Discarded        0
Unique head to head        1.00%
Reference rating      2300.0 (average of the pool)

Loose Anchors = none
Relative Anchors = none

Convergence rating calculation (cycle #1)

phase iteration    deviation    resolution
  0       6     37.465080152      60.15952
  1       1     37.465080152      54.58141
  2       2     35.097970644      45.64427
  3       9     27.207464309      30.28760
  4       7     18.894960131      22.25534
  5       9     11.788625136      14.82163
  6      15      5.630040860       7.71102
  7      31      0.923373521       1.39904
  8     118      0.000004578       0.00001
  9      46      0.000000098       0.00000
done

White Advantage = 0.0
Draw Rate (eq.) = 50.0 %

Post-Convergence rating estimation
done

   # PLAYER                 : RATING    POINTS  PLAYED     (%)
   1 Spike 1.4              : 2909.6     194.0     200    97.0
   2 Rybka 2.3.2a 64 bit    : 2869.2     192.5     200    96.2
   3 Ruffian 1.0.5          : 2517.2     155.0     200    77.5
   4 Hermann 2.8 64 bit     : 2488.4     149.0     200    74.5
   5 SOS 5.1 for Arena      : 2463.8     143.5     200    71.8
   6 AnMon 5.75             : 2414.8     131.5     200    65.8
   7 Bumblebee              : 2300.5    1200.5    2400    50.0
   8 Bitfoot                : 2220.2      77.5     200    38.8
   9 Shield                 : 2098.5      48.0     200    24.0
  10 Clunk                  : 2083.7      45.0     200    22.5
  11 KingSlayer             : 2009.9      32.0     200    16.0
  12 Clubfoot               : 1989.5      29.0     200    14.5
  13 FairyMax               : 1534.8       2.5     200     1.2

[Graham Banks]

Nice gain.

[zd3nik]

Hm. Apparently the last couple tweaks I made before releasing were bad - even though they showed an improvement in the quick tests I usually do to verify changes.

So far bumblebee is not doing well in the new 5/2 tourney. Even losing to bitfoot.