Strelka 5.5 vs Houdini & Critter

[Dr.Wael Deeb]

It's like Spanish soap opera, isn't it

Nah,definitely more exciting

[Dr.Wael Deeb]

Ok, so to be perfectly correct, Strelka 5.5 is the strongest engine under my test conditions. Good enough?
I don't see why Strelka has to be the Houdini clone. Houdini 1.5 is just slightly modified Robbolito. I disaassembled its search and found only minor changes compared to the Robbolito 0.085
If Strelka author also started from the Robbolito base, he will end up with the engine very similar to Houdini.

That's an interesting observation, but Strelka 5.5 seems to be much closer to Houdini 1.5 in strength than either are to Robbolito. Since you have disassembled Houdini search, can you comment on those differences you found that might account for the huge strength increase from Robbolito to Houdini 1.5? Critter author Richard Vida did so about a year ago, but the differences he cited were not enough to account for the large strength increase, in my opinion. Perhaps his list was incomplete, or else the gains came more from eval, I don't know.

Aha....now we're trying to look into the opponent's anatomy which makes him so powerful ,eh

God,I love computer chess and I'll always do

[Houdini]

Since you have disassembled Houdini search, can you comment on those differences you found that might account for the huge strength increase from Robbolito to Houdini 1.5? Critter author Richard Vida did so about a year ago, but the differences he cited were not enough to account for the large strength increase, in my opinion. Perhaps his list was incomplete, or else the gains came more from eval, I don't know.

Houdini's strength derives from magic.
Resistance is futile.

Robert

[LudiBuda]

Same conditions as the first test, but this time I used Strelka executable patched by R.Vida in order to disable internal opening book.

Strelka 5.5 - Houdini 58 - 70 with 112 draws

Very sneaky Mr.Osipov. You got me fooled

[ernest]

in order to disable internal opening book.

I don't understand: how can Strelka use his internal opening book, if the match games start with the book you choose (and you can check those openings...)

[LudiBuda]

Problem with my test is that I used well known test suites (silver, noomen, nunn).
I hope testing groups are not doing that

[IWB]

Hello Alex,

Do you have a link to any statement by Richard Vida for that. (I have missed that )

30000 position in the engines seems to be a nice little trick, but it is another good reason not to show the opening positions

Bye
Ingo

[LudiBuda]

Link to the patched Strelka is in:
http://talkchess.com/forum/viewtopic.ph ... 38&t=43424

His statements are in the same thread.

Test groups should make opening books and games secret I guess if they want to avoid cheating. Computer chess really became the wild west recently
I can only imagine what a 3-5 move deep internal opening book could do for Chess960 rating. Probably 50-100 elo gain no sweat for some positions.

[Mike S.]

Here for comparison another result:

1:30m+0,5s on N455/1,66 GHz --> a rather slow cpu
ponder off, 64 MB hash each
singlethread, 32 Bit all
S.Canbaz' Top10 opening set
Arena 2.0.1, Windows 7 32 Bit

Code: Select all

   Engine          Score                    St
1: Strelka 5.5     39.5/60 ···················· 
2: Strelka 5.1     10.0/20 011===010=0101===1=0 
3: Houdini 1.5a-2d 7.5/20  =00=010=01====0=0=== 
4: Ivanhoe 946hm-G 3.0/20  00=00000=0000=0=0010 

I am aware that the result against Ivanhoe is strange, but that is what they have produced. I have looked into the game's eval/depth profiles and all seemed ok, technically. No losses on time.

(Houdini 1.5a was customized with the internal 2.0b defaults, with the HoudiConfig tool.)

[IWB]

Link to the patched Strelka is in:
http://talkchess.com/forum/viewtopic.ph ... 38&t=43424

His statements are in the same thread.

Test groups should make opening books and games secret I guess if they want to avoid cheating. Computer chess really became the wild west recently
I can only imagine what a 3-5 move deep internal opening book could do for Chess960 rating. Probably 50-100 elo gain no sweat for some positions.

Thanks, I missed that as the other thread is about an irrelevant engine and too messy. Nonetheless it cought my attention here!

Actually my assumtion of the worst become true - thx I did not reviel the IPON openings. I know that this might not have helped in this case but at least I don't make it too easy for the cheaters!

Thanks again (and thanks to Richard Vida!) for showing this
Ingo