Page 1 of 2

IvanHoe T53 testing

Posted: Sun Jul 18, 2010 3:33 pm
by robbolito
Best engines-4 2010
Site/ Country: Q9550-XP64, United States
Level: Blitz 10/0
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

Code: Select all

                                         1              2              3              
1   IvanHoe 9.53b x64 KLO    3190   +62  ************** 0½1½11½0½1½½½½ 1½½½½½½½½½½11½  16.5/28
2   Houdini 1.03a x64 4_CPU  3190   -12  1½0½00½1½0½½½½ ************** ½½½½0101½1½½½½  13.5/28
3   Deep Rybka 4 x64         3190   -49  0½½½½½½½½½½00½ ½½½½1010½0½½½½ **************  12.0/28


Re: IvanHoe T53 testing

Posted: Sun Jul 18, 2010 3:58 pm
by benstoker
robbolito wrote:Best engines-4 2010
Site/ Country: Q9550-XP64, United States
Level: Blitz 10/0
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

Code: Select all

                                         1              2              3              
1   IvanHoe 9.53b x64 KLO    3190   +62  ************** 0½1½11½0½1½½½½ 1½½½½½½½½½½11½  16.5/28
2   Houdini 1.03a x64 4_CPU  3190   -12  1½0½00½1½0½½½½ ************** ½½½½0101½1½½½½  13.5/28
3   Deep Rybka 4 x64         3190   -49  0½½½½½½½½½½00½ ½½½½1010½0½½½½ **************  12.0/28

53 implements "magic bitboards". And it lowers the nps a bit. But, your results obviously suggest improvement. To my limited knowledge "magic bitboards" are all about speeding up search. Do they do something else? If the nps goes down because of "magic bitboards", then how can the ELO go up for 53?

Re: IvanHoe T53 testing

Posted: Sun Jul 18, 2010 5:55 pm
by robbolito
Hi benstoker,
this was the first test just for fun but results were very interesting.Even KLO's version T54 had a good score on 6 cores in a tournament that I have posted previously after he had implemented certain changes in the compile.
Amazing thing is how close these engines are in strength but the nice thing is that Robert is doing great job making Houdini excellent engine and how community effort at IvanHoe in cooperation with authors is producing engine that is stronger with every new release.And these programs are free so Vas is already in trouble if he can not follow these new engines that also include marvelous Stockfish.

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 2:26 am
by notyetagm
robbolito wrote:Hi benstoker,
this was the first test just for fun but results were very interesting.Even KLO's version T54 had a good score on 6 cores in a tournament that I have posted previously after he had implemented certain changes in the compile.
Amazing thing is how close these engines are in strength but the nice thing is that Robert is doing great job making Houdini excellent engine and how community effort at IvanHoe in cooperation with authors is producing engine that is stronger with every new release.And these programs are free so Vas is already in trouble if he can not follow these new engines that also include marvelous Stockfish.
Yes.

Who in there right mind is going to pay for Rybka 4 when Stockfish 1.8 is close in strength to Rybka 4 and free, while Houdini 1.03 and Ivanhoe 9.5x are probably slightly stronger than Rybka 4 and free to boot?

Vas needs a new business model. Perhaps he should release the cluster version of Rybka immediately as Rybka 5, a free upgrade for those who just bought Rybka 4. :-)

The days of Vas not upgrading the commercial Rybka for 2 years(!) at a time are O-V-E-R.

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 2:31 am
by notyetagm
robbolito wrote:Best engines-4 2010
Site/ Country: Q9550-XP64, United States
Level: Blitz 10/0
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

Code: Select all

                                         1              2              3              
1   IvanHoe 9.53b x64 KLO    3190   +62  ************** 0½1½11½0½1½½½½ 1½½½½½½½½½½11½  16.5/28
2   Houdini 1.03a x64 4_CPU  3190   -12  1½0½00½1½0½½½½ ************** ½½½½0101½1½½½½  13.5/28
3   Deep Rybka 4 x64         3190   -49  0½½½½½½½½½½00½ ½½½½1010½0½½½½ **************  12.0/28

Where can I find IvanHoe 9.53b ?

It's not at http://www.chesslogik.com/ivanhoe.htm.

Thanks

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 2:35 am
by notyetagm
robbolito wrote:Best engines-4 2010
Site/ Country: Q9550-XP64, United States
Level: Blitz 10/0
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

Code: Select all

                                         1              2              3              
1   IvanHoe 9.53b x64 KLO    3190   +62  ************** 0½1½11½0½1½½½½ 1½½½½½½½½½½11½  16.5/28
2   Houdini 1.03a x64 4_CPU  3190   -12  1½0½00½1½0½½½½ ************** ½½½½0101½1½½½½  13.5/28
3   Deep Rybka 4 x64         3190   -49  0½½½½½½½½½½00½ ½½½½1010½0½½½½ **************  12.0/28

I know, I know, I know that this is a short sample, and only 10' 0" to boot, but this result says

Houdini 1.03 is 37 elo stronger than Rybka 4 and

OMFG

Ivanhoe 9.53b is 111(!!) elo stronger than Rybka 4!

YIKES!

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 2:40 am
by notyetagm
benstoker wrote:
robbolito wrote:Best engines-4 2010
Site/ Country: Q9550-XP64, United States
Level: Blitz 10/0
Hardware: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz4x 3500 MHz with 4,096 MB Memory
Operating system: Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)

Code: Select all

                                         1              2              3              
1   IvanHoe 9.53b x64 KLO    3190   +62  ************** 0½1½11½0½1½½½½ 1½½½½½½½½½½11½  16.5/28
2   Houdini 1.03a x64 4_CPU  3190   -12  1½0½00½1½0½½½½ ************** ½½½½0101½1½½½½  13.5/28
3   Deep Rybka 4 x64         3190   -49  0½½½½½½½½½½00½ ½½½½1010½0½½½½ **************  12.0/28

53 implements "magic bitboards". And it lowers the nps a bit. But, your results obviously suggest improvement. To my limited knowledge "magic bitboards" are all about speeding up search. Do they do something else? If the nps goes down because of "magic bitboards", then how can the ELO go up for 53?
Anyone care to explain what "magic bitboards" are?

Thanks

PS Nevermind -> http://chessprogramming.wikispaces.com/Magic+Bitboards

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 3:48 pm
by Charles
I dont think there are enough games to conclude that Rybka 4 is weaker than either houdini and Ivanhoe. So far in my testing they seem about equal with Rybka 4 getting a slight edge in some cases. I also suspect that Rybka 4 will be much more dominating in longer time controls ...

Additionally, I think Rybka 4 has a very interesting style and extremely good command of rook pawn endings -- it is superior to ivanhoe with triple bases in this regard. Also Rybka 4 tends to walk its king into the enemy camp very quickly -- many games have been won agains the ippo family this way. So even if Rybka 4 turns out to be slightly weaker than the others at short time control it is still a very intriguing engine to have.


By the way - my copy of ivanhoe53 is very weak -- the default 32 bit ... is there a stronger version ?

Re: IvanHoe T53 testing

Posted: Mon Jul 19, 2010 6:00 pm
by robbolito
Interim report.
I have been intrigued by this new version T53.x64 compiled by KLO.
Since there is always complain: not enough games ,so I have started 100 games match.So far after 52 games the score is 10-1 and 41 draws for Ivanhoe T53 KLO.
The fact is, that these engines are very close in strength. but the new version T53 at least on Q9550 has been stronger than Rybka 4.Those people who know can determine ELO difference with this kind of score.
Final score will be done tomorrow morning.

Re: IvanHoe T53 testing

Posted: Tue Jul 20, 2010 11:42 am
by notyetagm
robbolito wrote:Interim report.
I have been intrigued by this new version T53.x64 compiled by KLO.
Since there is always complain: not enough games ,so I have started 100 games match.So far after 52 games the score is 10-1 and 41 draws for Ivanhoe T53 KLO.
The fact is, that these engines are very close in strength. but the new version T53 at least on Q9550 has been stronger than Rybka 4.Those people who know can determine ELO difference with this kind of score.
Final score will be done tomorrow morning.
The Chess Elo Rating Difference Calculator at http://www.3dkingdoms.com/chess/elo.htm

says that +10 -1 =41 is 30.5/52, a winning percentage of 58.65% and an elo difference of +61.

So the preliminary results of your test show Ivanhoe T53 is a whopping +61 elo stronger than Rybka 4!

YIKES!