To the Copy and Paste Generation of Programmers

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
bob
Posts: 20404
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: To the Copy and Paste Generation of Programmers

Post by bob » Thu Mar 31, 2011 6:58 pm

Dann Corbit wrote:
gerold wrote:
Michel wrote:
Besides the houdini author, who else has managed it?
Examples are the authors of Rybka and Toga.

Another example is Marco and Joona improving Glaurung.

Very likely there are more examples which are not known.
True. Just take a look at the top 20 engines on rating lists.
Most have taken idea's, code etc. from the Ivanhoe family.
Where did all these new engines come from and with big improvements
in just a few months.
How is it that you know programs on the top 20 list have taken code from the Ivanhoe family?

Toga did not improve 200 Elo over fruit from other people's code.
If fact, almost all of the improvement of Toga is from SMP:

Code: Select all

Fruit 2.3.1	2888	+14	-13	50.0%	-0.4	46.1%	1721
Toga II 1.4.1SE	2930	+12	-11	49.8%	+3.7	43.7%	2451
The strongest Toga is only 42 Elo stronger than Fruit when run on a single CPU.

Marco and Joona started with Stockfish and not Glaurung and they did not impove it 200 Elo with code from other programs.

These are the top 20 program from the CCRL list, after removing duplicates:

Code: Select all

no Program Elo + - Games Score Av.Op. Draws 
1 Houdini 1.5a x64 6CPU 3294 23 23 600 73.0% 3121 36.0% 
3 Deep Rybka 4 x64 6CPU 3245 23 23 500 71.2% 3087 43.6% 
6 Stockfish 1.8 x64 6CPU 3209 22 22 500 65.6% 3096 48.8% 
14 Critter 0.90 x64 6CPU 3162 26 26 350 53.6% 3138 48.3% 
19 Naum 4.2 x64 6CPU 3148 25 25 350 49.9% 3149 52.3% 
35 Spike 1.4 4CPU 3065 20 20 650 50.8% 3059 43.2% 
36 Deep Fritz 12 6CPU 3062 26 26 348 43.5% 3107 49.7% 
37 Deep Shredder 12 x64 4CPU 3061 8 8 4329 52.6% 3043 42.7% 
38 Zappa Mexico II x64 6CPU 3050 24 24 400 42.4% 3103 51.2% 
40 Protector 1.4.0 x64 4CPU 3047 23 23 500 48.3% 3059 42.6% 
42 Hiarcs 13.2 MP 4CPU 3042 15 15 1165 41.9% 3098 45.7% 
46 Komodo 1.3 x64 3022 14 14 1370 47.4% 3040 44.3% 
48 Spark 1.0 x64 4CPU 3021 14 14 1300 43.7% 3065 44.5% 
55 Deep Junior 12 x64 4CPU 3004 15 15 1250 46.0% 3032 40.4% 
63 Fruit 090705 x64 4CPU 2980 17 17 917 42.4% 3033 41.3% 
64 Deep Sjeng ct 2010 2977 13 13 1553 46.7% 3000 43.5% 
65 Gull 1.1 x64 2975 15 15 1122 50.4% 2973 42.8% 
67 Thinker 5.4Di x64 4CPU 2972 11 11 2140 51.2% 2964 42.4% 
68 Deep Onno 1.2.70 x64 4CPU 2971 18 18 807 45.0% 3006 46.3% 
I guess two or three may have some code from other open source programs. Of course, that is a very bad thing. I suspect that almost alll of them have ideas from these programs. Of course, that is a very good thing and exactly how things ought to work.
I suspect your "I guess ..." is much farther from the truth than you might suspect...

This is a massive problem, not a tiny one. Rybka is an example where _everybody_ said it was not a derivative. That balloon has pretty well burst.

Post Reply