Crafty 23.2 released.

[Aser Huerga]

OK, thanks Bob.
I'm testing the 64-bit version now, I'll post the results in a few days.

[Werner]

OK, thanks Bob.
I'm testing the 64-bit version now, I'll post the results in a few days.

Hi Aser,
just tested a few Crafty engines here inside Arena 2.01:

None of them using that field TBHits inside Arena - but of course you see in the line: Tbs are used - and you see it of course if you setup a 5men position which has a long mate.

[Aser Huerga]

OK, thanks Bob.
I'm testing the 64-bit version now, I'll post the results in a few days.

Hi Aser,
just tested a few Crafty engines here inside Arena 2.01:

None of them using that field TBHits inside Arena - but of course you see in the line: Tbs are used - and you see it of course if you setup a 5men position which has a long mate.

Yes, I've just checked it. Since is my first Crafty testings I prefer to ask to be sure things are right.

Thanks!

[jwes]

Try putting "noise 1" in your crafty.rc file. Crafty does not display any pv if it does not search enough nodes.

[swami]

Crafty 23.2

by Robert Hyatt, USA..

Strategic Test Suite Conditions:

Core2Quad 32 bits, Q6600, 2 GB RAM, 2.4GHZ
10 seconds per position
900 positions
Engine uses 128 Mb Hash.
Single CPU
Arena GUI


Overall Performance:

• 
  Total Score: 598/900
  [...]
  Average: 66.44%
  [...]
  Grade: B+
  [...]
  Total Rated Time: 57.92/150 minutes [3475 Seconds/9000 Seconds]
  [...]
  Percentage Time Used: 38.61%

Subject-wise Scores:

STS (v1.0) - Undermining:
78/100, Grade: A

STS (v2.1) - Open Files and Diagonals:
67/100, Grade: B+

STS (v3.0) - Knight Outposts/Centralization/Repositioning:
78/100, Grade: A

STS (v4.1) - Square Vacancy:
69/100, Grade: B+

STS (v5.0) - Bishop vs Knight:
65/100, Grade: B+

STS (v6.0) - Re-Capturing:
62/100, Grade: B

STS (v7.0) - Offer of Simplification:
60/100, Grade: B

STS (v8.1) - Advancement of f/g/h Pawns:
58 /100, Grade: C+

STS (v9.0) - Advancement of a/b/c Pawns:
61/100, Grade: B

Best Wishes,
Swami

[F. Bluemers]

OK, thanks Bob.
I'm testing the 64-bit version now, I'll post the results in a few days.

Hi Aser,
just tested a few Crafty engines here inside Arena 2.01:

None of them using that field TBHits inside Arena - but of course you see in the line: Tbs are used - and you see it of course if you setup a 5men position which has a long mate.

Yes, I've just checked it. Since is my first Crafty testings I prefer to ask to be sure things are right.

Thanks!

Don't worry
What arena puts in there (tbhits) is likely to come from uci engines reporting their tb hits with the uci protocol "info tbhits"

[swami]

Partial - Credit Scores Calculation:
Using STS Stat 2.5

Crafty 23.2

by Robert Hyatt, USA.

Overall Performance:

• 
  Total Score: 6888/9000
  [...]
  Average : 76.53%
  [...]
  Grade: A

Subject-wise Scores:

STS (v1.0) - Undermining:
839/1000, Grade: A+

STS (v2.1) - Open Files and Diagonals:
732/1000, Grade: A-

STS (v3.0) - Knight Outposts/Centralization/Repositioning:
833/1000, Grade: A+

STS (v4.1) - Square Vacancy:
790/1000, Grade: A

STS (v5.0) - Bishop vs Knight:
755/1000, Grade: A

STS (v6.0) - Re-Capturing:
814/100, Grade: A+

STS (v7.0) - Offer of Simplification:
706/1000, Grade: A-

STS (v8.1) - Advancement of f/g/h Pawns:
703 /1000, Grade: A-

STS (v9.0) - Advancement of a/b/c Pawns:
716/1000, Grade: A-

Best Wishes,
Swami

[Robert Flesher]

20 elo is actually pretty impressive, imo. bob, do you have a guesstimate as to what crafty's human elo would be? i know come CC fans don't even think in such terms but i'd think as an author it would be one of the most interesting aspects to me. i guess computer chess has in some ways (certainly in strength) surpassed human chess though.

Too many different rating pools. USCF? CCF? FIDE? Each would be different. And since there are not many games between Crafty and FIDE players (except for blitz where humans have real problems anyway) it is difficult to produce a decent guess...

Some of us remember when you believed, and quite openly stated engines were not GM strength (not even close). However, that was then and this is now. Clearly Crafty is of strong GM strength, and seems alot harder to beat than many of the ELO giants. Personally, I like the direction you have taken your engine. True to its roots, and a S.O.B to beat in blitz.

That was 10-15 years ago, when it was correct. Hardware has changed a bunch since then.

I was just remembering the past Bob, nothing more. But, also stating Crafty is now clearly GM strength.

[Jouni]

No big difference with tablebases: 68/100 (time 45:52). Stockfish 1.6.3 got 86/100 (time 23:30) !!

Jouni

[bob]

That's not the case with xboard. And, in fact, unless you put the crafty.rc file in the wrong place (it should be in the -fd directory) I don't see how crafty could not use that file. It certainly does here. But that's xboard and not winboard.

My rc file is in the -fd directory like it has been for the past 15 yrs

I am trying to figure out what the problem is and email HGM about it, because no matter the settings in my rc file, Winboard is taking them over.

If I use mt=2, and Winboard is set to use a _maximum_ of 4 cores, it defaults to Winboard's settings and uses 4 cores.

That is an expected behavior. Crafty now is responsive to winboard commands, so when you set winboard to 4 cores, winboard sends "cores 4" and crafty adjusts to it.

Actually, no. If you set max threads in the .craftyrc/crafty.rc file, it will not use the "cores" command from xboard/winboard. If you set either hash or hashp it will also not use (and actually disabled) the "memory" command.

What I do is first initialize via the .craftyrc file. If I find a mt or hash/hashp command, I make a note. Then when I start talking to xboard, I tell it "feature memory=0" or "feature cores=0" if the .craftyrc file set those values. So it really looks like Peter's problem is related to Crafty never even seeing the .craftyrc/crafty.rc file at all.

Miguel

Just like if I set no hash or hashp settings, it defaults to Winboard's 1024MB limit I have set. There is no way to unset those values. Clearing them does nothing.

Using version 4.2.7b, I don't have this issue.

Peter