Rybka-GM Benjamin Pawn Odds Match

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
User avatar
Steve B
Site Admin
Posts: 10140
Joined: Sun Jul 29, 2007 10:02 am
Location: New York City USofA
Contact:

Post by Steve B »

spacious_mind wrote:
Yes Nick
as you just mentioned it was the Revelation which runs on a Xscale PXA255 processor-500 Mhz-32MB Ram 16 MB Flashrom

it should also be mentioned that the Revelation has not yet been released for sale and the computer used by Ruud was a prototype

Best
Steve
Thanks Steve.
Are you # 1 again on the order list ? :D

But Of Course Regards
Steve
:wink:
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Uri:
I think this example might be better suited since there is more game history:

152 Ruffian 2.1.0 2649 = Your Rating

Match Score: (Same Tournament)
Res Ruffian 0.5 - 2.5 Elite 68060 V11 (2296)

142 Deep Sjeng 1.6 2CPU 2666 = Your Rating

Match Score: (Same Tournament)
Mephisto Montreux (2279) 3.5 - 1.5 Res Deep Sjeng 1.8

Bear in mind that Ruffian 2.1 and Deep Sjeng 1.8 are running at 200Mhz with a bunch of other bells and whistles.

I dont even see a 100 ELO difference between these. These are NOT flukes or lucky games.

Best Regards
Nick
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

spacious_mind wrote:Uri:
I think this example might be better suited since there is more game history:

152 Ruffian 2.1.0 2649 = Your Rating

Match Score: (Same Tournament)
Res Ruffian 0.5 - 2.5 Elite 68060 V11 (2296)

142 Deep Sjeng 1.6 2CPU 2666 = Your Rating

Match Score: (Same Tournament)
Mephisto Montreux (2279) 3.5 - 1.5 Res Deep Sjeng 1.8

Bear in mind that Ruffian 2.1 and Deep Sjeng 1.8 are running at 200Mhz with a bunch of other bells and whistles.

I dont even see a 100 ELO difference between these. These are NOT flukes or lucky games.

Best Regards
I think that something is wrong with these results.
one possibility is that the programmers of ruffian and deep sjeng simply did not care to optimize their programs for the slow hardware
that is used and I do not mean different search algorithm but speed optimizations with the same algorithm.

Maybe it is possible to make ruffian or Sjeng 10 times faster on old hardware not by changing the algorithm but only by optimization for slow hardware that is not important for commercial programmers.

Another possibility is that the opponents are not normal programs but dedicated computers that were build only for chess and in that case I suspect that the mhz is simply misleading because hardware that is designed for chess has a speed advantage relative to software that run on computers with the same mhz that were not designed for chess.

I know nothing about
Elite 68060 V11 or Montreux and I wonder if they are chess software or dedicated chess computers that are combination of software and hardware like hydra.

Uri
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Uri Blass wrote:
spacious_mind wrote:Uri:
I think this example might be better suited since there is more game history:

152 Ruffian 2.1.0 2649 = Your Rating

Match Score: (Same Tournament)
Res Ruffian 0.5 - 2.5 Elite 68060 V11 (2296)

142 Deep Sjeng 1.6 2CPU 2666 = Your Rating

Match Score: (Same Tournament)
Mephisto Montreux (2279) 3.5 - 1.5 Res Deep Sjeng 1.8

Bear in mind that Ruffian 2.1 and Deep Sjeng 1.8 are running at 200Mhz with a bunch of other bells and whistles.

I dont even see a 100 ELO difference between these. These are NOT flukes or lucky games.

Best Regards
I think that something is wrong with these results.
one possibility is that the programmers of ruffian and deep sjeng simply did not care to optimize their programs for the slow hardware
that is used and I do not mean different search algorithm but speed optimizations with the same algorithm.

Maybe it is possible to make ruffian or Sjeng 10 times faster on old hardware not by changing the algorithm but only by optimization for slow hardware that is not important for commercial programmers.

Another possibility is that the opponents are not normal programs but dedicated computers that were build only for chess and in that case I suspect that the mhz is simply misleading because hardware that is designed for chess has a speed advantage relative to software that run on computers with the same mhz that were not designed for chess.

I know nothing about
Elite 68060 V11 or Montreux and I wonder if they are chess software or dedicated chess computers that are combination of software and hardware like hydra.

Uri
Perhaps you are right perhaps you are wrong, without asking the programmers directly I will trust that Ruud Martin gave me Fruit 2.1, Toga 2 1.2, Ruffian 2.1, Deep Sjeng 1.8 etc in my Resurrection.

But like every collector of dedicated chess computers I am also curious about what I have so I do spend time playing around with it. So since I also like engines I am also curious about where my dedicated computers fit in a ranking. Therefore I am playing some tests to bridge the gap between dedicated and engines.

Humor me for a moment and look at this logic and if I am wrong somewhere show me where so that I can correct it.

1) If I take say 22 games at a start rating of 2400 and run them through ELO
Stat I would get approximately +134/-136
2) Generally everyone talks about +50-+75 ELO per doubling of CPU speed, so for my test lets do +62.5 ELO
3) My Laptop Info is: HP AMD Turion 64 Mobile Technology ML-321.79 GHz, 1.00 GB RAM
4) I want to test against Chess Engines on my Laptop but I want the Hash 32 MB to be about the same and no tablebases and at 30sec/move setting.
5) You show Toga II 1.2 with an XXX CPU as 89 Toga II 1.2 2761, I say on my laptop Toga II 1.2 engine using my play setting is about ELO 2650 (Still matches your ELO within the ELO Stat tolerances)
6) Now I know that I have to deduct about 250 ELO points (268 at 62.5 per doubling) to get the approx level to Test Resurrection Toga II 1.2 at 200 Mhz.
7) So who do I test against? Easy lets look at Division III 3rd Division edition 14 WBEC Ridderkerk for the opponents.

So after 22 Games giving Toga II 1.2 a start rating of ELO 2400 here is the current ELO Stat stats:


Program Elo + - Games Score Av.Op. Draws

1 Resurrection Toga II 1.2 : 2375 134 136 22 43.2 % 2423 22.7 %


Individual statistics:

1 Resurrection Toga II 1.2 : 2375 22 (+ 7,= 5,- 10), 43.2 %

Anechka 0.08 : 4 (+ 3,= 0,- 1), 75.0 %
Alaric 704 : 4 (+ 1,= 0,- 3), 25.0 %
Snitch 1.6.2 : 4 (+ 1,= 2,- 1), 50.0 %
Phalanx22 : 4 (+ 1,= 1,- 2), 37.5 %
Delphil 1.6c : 4 (+ 0,= 2,- 2), 25.0 %
Bruja v1.9 : 2 (+ 1,= 0,- 1), 50.0 %

As you can see sofar the results match pretty closely to what I estimated before I started. Therefore I cannot find any reasons to dispute Ruud Martin on the software he has provided. I plan to complete this test someday with all the Resurrection engines when I have the time.

There is nothing here sofar that tells me the software in a Resurrection is different.

PS. Mephisto Montreaux and Fidelite Elite V11 are both dedicated board computers !!


In case you are interested in our lowly dedicated matches here are the games:

http://www.geocities.com/spacious_mind/tests/toga.txt

You will have to rename the text file to .pgn since geocities will not take pgn files.

best regards
Nick
User avatar
Steve B
Site Admin
Posts: 10140
Joined: Sun Jul 29, 2007 10:02 am
Location: New York City USofA
Contact:

Post by Steve B »

spacious_mind wrote:
As you can see sofar the results match pretty closely to what I estimated before I started. Therefore I cannot find any reasons to dispute Ruud Martin on the software he has provided. I plan to complete this test someday with all the Resurrection engines when I have the time.

There is nothing here sofar that tells me the software in a Resurrection is different.

Ruud did the compilations himself to port the Engines in the Resurrection to run on the Strongarm SA1110 processor-203 Mhz ..with the engine authors permission of course

so only he really knows what differences if any the programs have compared to the PC Engines
for the not yet Released Resurrection II and Revelation computers i believe the Rybka compilation was preformed by Rajhlich himself(but i am not 100% certain of that)

personally i dont see what difference it makes if the program is or is not the same exact as the original pc engine

collectors are not trying to compare the engines in the Resurrection units with their Pc counterparts...only against other dedicated computers

and i dont think it is very interesting to say that Rybka running on an eight core opteron (or whatever)would make one move and in the Resurrection it only chooses another move

but i will admit i am guilty of not being interested in Pc engines so i guess others might find this sort of comparison of interest

Dedicated to Dedicated Regards
Steve
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

1)The best Toga in the list is
68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
2)The list was done based on hardware of PIV 2 ghz so I do not see how can you reduce so much to get 200 mhz.

I think that
200 mhz is only 10 times slower than 2 ghz.

268 elo at 62.5 per doubling mean that your 200 mhz is more than 16 times slower than your laptop and I wonder if this is correct for correct programming.

Maybe this is correct for the compilation of toga by Ruud but I wonder if there are no significant speed tricks that it is possible to get in case that toga was written for that hardware in the first place.

Uri
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Uri Blass wrote:1)The best Toga in the list is
68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
2)The list was done based on hardware of PIV 2 ghz so I do not see how can you reduce so much to get 200 mhz.

I think that
200 mhz is only 10 times slower than 2 ghz.

268 elo at 62.5 per doubling mean that your 200 mhz is more than 16 times slower than your laptop and I wonder if this is correct for correct programming.

Maybe this is correct for the compilation of toga by Ruud but I wonder if there are no significant speed tricks that it is possible to get in case that toga was written for that hardware in the first place.

Uri
Hi Uri:
You are correct I have one too many 62.5's in my estimation when I quickly created it yeasterday for you but still it really does not matter, even with the extra 62.5 the end result is still the same. If I run the 22 games through Chessbase I get ELO 2397, if I use FIDE Rating - Einstiegsberechnung I get 2408 whatever number I use Toga II 1.2 (not Toga II 1.2a which you are trying to sell to me) plays at 200Mhz plays at approx 2400 ELO.

Now if you want to show that I am wrong, then how about a match, do any of your friends have a 200Mhz, if so lets play because I am also after the truth here, is your theory right at 400 or mine at 150-200. If I am wrong then good at least I am a step closer to the truth.

Best regards

Nick
Nick
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

spacious_mind wrote:
Uri Blass wrote:1)The best Toga in the list is
68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
2)The list was done based on hardware of PIV 2 ghz so I do not see how can you reduce so much to get 200 mhz.

I think that
200 mhz is only 10 times slower than 2 ghz.

268 elo at 62.5 per doubling mean that your 200 mhz is more than 16 times slower than your laptop and I wonder if this is correct for correct programming.

Maybe this is correct for the compilation of toga by Ruud but I wonder if there are no significant speed tricks that it is possible to get in case that toga was written for that hardware in the first place.

Uri
Hi Uri:
You are correct I have one too many 62.5's in my estimation when I quickly created it yeasterday for you but still it really does not matter, even with the extra 62.5 the end result is still the same. If I run the 22 games through Chessbase I get ELO 2397, if I use FIDE Rating - Einstiegsberechnung I get 2408 whatever number I use Toga II 1.2 (not Toga II 1.2a which you are trying to sell to me) plays at 200Mhz plays at approx 2400 ELO.

Now if you want to show that I am wrong, then how about a match, do any of your friends have a 200Mhz, if so lets play because I am also after the truth here, is your theory right at 400 or mine at 150-200. If I am wrong then good at least I am a step closer to the truth.

Best regards

Nick
I do not try to sell you toga1.2.1a
You use an early version of toga for comparison of software(Toga1.2) when the public version is toga1.2.1a for a long time.

You cannot use Toga1.2 to find the improvement in software

Edit:I can add that testers do not take 1.2 seriously

1.2.1a has 5183 games when 1.2 has only 131 games
I guess that there is a reason for that

68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
89 Toga II 1.2 2761 48 49 131 41.2 % 2823 35.1 %


Uri
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

spacious_mind wrote:Uri:
I think this example might be better suited since there is more game history:

152 Ruffian 2.1.0 2649 = Your Rating

Match Score: (Same Tournament)
Res Ruffian 0.5 - 2.5 Elite 68060 V11 (2296)

142 Deep Sjeng 1.6 2CPU 2666 = Your Rating

Match Score: (Same Tournament)
Mephisto Montreux (2279) 3.5 - 1.5 Res Deep Sjeng 1.8

Bear in mind that Ruffian 2.1 and Deep Sjeng 1.8 are running at 200Mhz with a bunch of other bells and whistles.

I dont even see a 100 ELO difference between these. These are NOT flukes or lucky games.

Best Regards

I have a comment about this example.

Based on my knowledge Deep Sjeng and Ruffian use bitboards that are 64 bit numbers.

BitBoards are fast on fast computers but I believe that they are slow on old machines.
I will be surprised if Martin worked on the data structure of Deep sjeng to use the same algorithm by faster methods than BitBoards.

My point is that it is not fair to claim that there was a small improvement in software when today no programmer works to optimize his program for the old machines.

I will not be surprised if ruffian and Sjeng lost 200 elo from unefficient implementation that made them 10 times slower.

Uri
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

Uri Blass wrote:
spacious_mind wrote:
Uri Blass wrote:1)The best Toga in the list is
68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
2)The list was done based on hardware of PIV 2 ghz so I do not see how can you reduce so much to get 200 mhz.

I think that
200 mhz is only 10 times slower than 2 ghz.

268 elo at 62.5 per doubling mean that your 200 mhz is more than 16 times slower than your laptop and I wonder if this is correct for correct programming.

Maybe this is correct for the compilation of toga by Ruud but I wonder if there are no significant speed tricks that it is possible to get in case that toga was written for that hardware in the first place.

Uri
Hi Uri:
You are correct I have one too many 62.5's in my estimation when I quickly created it yeasterday for you but still it really does not matter, even with the extra 62.5 the end result is still the same. If I run the 22 games through Chessbase I get ELO 2397, if I use FIDE Rating - Einstiegsberechnung I get 2408 whatever number I use Toga II 1.2 (not Toga II 1.2a which you are trying to sell to me) plays at 200Mhz plays at approx 2400 ELO.

Now if you want to show that I am wrong, then how about a match, do any of your friends have a 200Mhz, if so lets play because I am also after the truth here, is your theory right at 400 or mine at 150-200. If I am wrong then good at least I am a step closer to the truth.

Best regards

Nick
I do not try to sell you toga1.2.1a
You use an early version of toga for comparison of software(Toga1.2) when the public version is toga1.2.1a for a long time.

You cannot use Toga1.2 to find the improvement in software

Edit:I can add that testers do not take 1.2 seriously

1.2.1a has 5183 games when 1.2 has only 131 games
I guess that there is a reason for that

68 Toga II 1.2.1a 2800 8 8 5183 53.0 % 2779 35.6 %
89 Toga II 1.2 2761 48 49 131 41.2 % 2823 35.1 %


Uri
I can add that I do not understand how do you get
2408 fide rating for Toga II 1.2 on 200 mhz

from the ssdf I get

89 Hiarcs 7.32 64MB P200 MMX 2472 21 -21 1134 52% 2459
source
http://ssdf.bosjo.net/rlwww071.txt

Hiarcs7.32 is clearly weaker than togaII1.2

from the CEGT blitz I get
44 Toga II 1.2.1a 2802 10 10 3364 50.6 % 2798 32.2 % (sorry but there is no toga1.2 for you and even if you reduce it to 2770 there is 150 elo difference)
205 Hiarcs 7.32 2620 22 22 684 54.8 % 2587 32.2 %

source:
http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

This mean that toga is more than 100 elo better than hiarcs7.32 even on the same hardware(40/4 on 2 ghz should be equivalent to slow time control on slower hardware)

This means that Toga should get at least 2600 fide rating on 200 mhz.

Uri
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

[quote="Uri Blass]
I have a comment about this example.

Based on my knowledge Deep Sjeng and Ruffian use bitboards that are 64 bit numbers.

BitBoards are fast on fast computers but I believe that they are slow on old machines.
I will be surprised if Martin worked on the data structure of Deep sjeng to use the same algorithm by faster methods than BitBoards.

My point is that it is not fair to claim that there was a small improvement in software when today no programmer works to optimize his program for the old machines.

I will not be surprised if ruffian and Sjeng lost 200 elo from unefficient implementation that made them 10 times slower.

Uri
Uri:
It only just now struck me that your name is a name I know so let me start by saying that I have the highest respect for you and all other engine programmers, you can do something which I cannot do which is create a chess engine. Your engine has also given me a lot of pleasure :D

When you talk about fairness, my defence is for dedicated chess computers and the softwares of the past. When you say it is unfair about taking a modern engine and lowering it down to 200Mhz because of bitbases, then you are maybe also being unfair to the programmers of the past who did not have that technology available to them back then otherwise they also would have taken advantage of it, so for me it would be fair to say that the only way to compare fairly is to take modern software back to the past and sofar with all the examples I have shown from Rybka/Toga/Ruffian/Deep Sjeng I have not seen that great ELO gap that you believe there is when I have taken modern software backwards.

I do accept that you have strong theories of why, and this interests me also. So I ask you do you think that you could configure your engine to fit in a Resurrection? StrongARM, 204 MHz, 32 Bit
Speicher: 12 MB RAM

If you believe that you can configure your software accordingly and have it working at say ELO 2500-2600, this would be of great interest I am sure for the dedicated chess computer fans and this gives you at the same time the opportunity to test your theories of making a strong engine at a low CPU :D

Also to bring this thread back to Humans vs Machines I am convinced that GM's will come back to these kind of matches someday readily if they were sitting in front of a wooden board computer opponent that plays at ELO 2600 - 2800 :D

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Uri Blass wrote: I can add that I do not understand how do you get
2408 fide rating for Toga II 1.2 on 200 mhz

from the ssdf I get

89 Hiarcs 7.32 64MB P200 MMX 2472 21 -21 1134 52% 2459
source
http://ssdf.bosjo.net/rlwww071.txt

Hiarcs7.32 is clearly weaker than togaII1.2

from the CEGT blitz I get
44 Toga II 1.2.1a 2802 10 10 3364 50.6 % 2798 32.2 % (sorry but there is no toga1.2 for you and even if you reduce it to 2770 there is 150 elo difference)
205 Hiarcs 7.32 2620 22 22 684 54.8 % 2587 32.2 %

source:
http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

This mean that toga is more than 100 elo better than hiarcs7.32 even on the same hardware(40/4 on 2 ghz should be equivalent to slow time control on slower hardware)

This means that Toga should get at least 2600 fide rating on 200 mhz.

Uri
As I said from the beginning the Toga II 1.2 I am talking about is the engine inside Resurrection. I wanted out of curiosity to find out at what level the Resurrection engines can play if they played against a computer opponent. I knew that Div 1 and Div 2 engines might be too strong for it since my laptop runs at 1.8 Ghz so I played against Division 3. First with the champion Aleric 4 games and then 2nd place etc and so on.

I have played 22 games so far at 30 sec/move. This is the time standard generally preferred by dedicated computer fans because it is faster and more fun because we have to move the pieces manually on the dedicated machines so it is not a 100% exact science but pretty close most of the time since both machines get the same time benefit. Therefore most comparisons by dedicaters is made at 30 sec / move.

Since I know the published ELO of Aleric for example from the published site where I took the information from I know all the ELO's of the opponents of Toga II 1.2.

I played these 22 matches using gui in Frtiz or in Winboard depending on the engine. I saved all the games in Chessbase. I gave Toga a start ELO of 2400 (not that it would make any difference what I used) After 22 games Chessbase Tournament stat tells me the ELO is 2397.

To check on this ELO 2397 I went here:

FIDE Rating - Einstiegsberechnung

http://www.schachcomputer.info/html/elo_listen.html

Here I manually entered the results of the 22 games and the ELO calculation was ELO 2408. (assuming I did not make any mistakes entering all 22 results)

So as you can see the statement that I can make is that my Resurrection Toga II 1.2 can play at @ 2400 ELO against Computer Chess Engines rated at ridderkerk (i know he plays 2/40 but this is a good reference base for me) that use 32 Mb Hash and no endgame tablebases on my laptop which runs at 1.8 Ghz.

Now this is interesting for me to know because I now have the possibility if I wanted to play an engine tournament to enter Toga in this tournament with other engines at 30/move rated say between 2350-2450 ELO and RES Toga would play competitively. I also have fruit 2.1, toga 1.0, ruffian 2.1, deep sjeng 1.8 and soon rybka to test the same way for my home enjoyment.

I hope this explains the numbers better.

Best regards
Nick
Uri Blass
Member
Posts: 82
Joined: Sun Aug 12, 2007 1:40 pm

Post by Uri Blass »

spacious_mind wrote:
[quote="Uri Blass]
I have a comment about this example.

Based on my knowledge Deep Sjeng and Ruffian use bitboards that are 64 bit numbers.

BitBoards are fast on fast computers but I believe that they are slow on old machines.
I will be surprised if Martin worked on the data structure of Deep sjeng to use the same algorithm by faster methods than BitBoards.

My point is that it is not fair to claim that there was a small improvement in software when today no programmer works to optimize his program for the old machines.

I will not be surprised if ruffian and Sjeng lost 200 elo from unefficient implementation that made them 10 times slower.

Uri
Uri:
It only just now struck me that your name is a name I know so let me start by saying that I have the highest respect for you and all other engine programmers, you can do something which I cannot do which is create a chess engine. Your engine has also given me a lot of pleasure :D

When you talk about fairness, my defence is for dedicated chess computers and the softwares of the past. When you say it is unfair about taking a modern engine and lowering it down to 200Mhz because of bitbases, then you are maybe also being unfair to the programmers of the past who did not have that technology available to them back then otherwise they also would have taken advantage of it, so for me it would be fair to say that the only way to compare fairly is to take modern software back to the past and sofar with all the examples I have shown from Rybka/Toga/Ruffian/Deep Sjeng I have not seen that great ELO gap that you believe there is when I have taken modern software backwards.

I do accept that you have strong theories of why, and this interests me also. So I ask you do you think that you could configure your engine to fit in a Resurrection? StrongARM, 204 MHz, 32 Bit
Speicher: 12 MB RAM

If you believe that you can configure your software accordingly and have it working at say ELO 2500-2600, this would be of great interest I am sure for the dedicated chess computer fans and this gives you at the same time the opportunity to test your theories of making a strong engine at a low CPU :D

Also to bring this thread back to Humans vs Machines I am convinced that GM's will come back to these kind of matches someday readily if they were sitting in front of a wooden board computer opponent that plays at ELO 2600 - 2800 :D

Best regards
I do not claim that it is fair to compare performance on the new hardware
when programs of old programs could not optimize their program for the new hardware.

My point is that the programmers of today do not use a data structure
that is fast for the slow computers because they do not care about performance of their program in slow hardware so comparing performance on slow hardware is also unfair when we want to learn about the improvement.

Another question is if the rating in Leo's site is correct.

The rating at the top seems to be logical but difference of 70 elo between computers may be translated to different of only 50 elo against humans so it is possible that at lower level they are too low.

Uri
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Uri Blass wrote:
I do not claim that it is fair to compare performance on the new hardware
when programs of old programs could not optimize their program for the new hardware.

My point is that the programmers of today do not use a data structure
that is fast for the slow computers because they do not care about performance of their program in slow hardware so comparing performance on slow hardware is also unfair when we want to learn about the improvement.

Another question is if the rating in Leo's site is correct.

The rating at the top seems to be logical but difference of 70 elo between computers may be translated to different of only 50 elo against humans so it is possible that at lower level they are too low.

Uri
Hi Uri
What is your suggestion for a fair comparison of old and new?

I have an old Pentium 320 which I use for DOS only to play my Tasc Chess Machine cards and old DOS chess software. I am tempted to load Windows 3.11 on it and see if I can get Winboard working on it and then try some more tests to see what the difference is between it, Resurrection at 203 MHz and my laptop at 1.8Ghz. I might even start looking to see if I can find an old 66Mhz.

This topic intrigues me.

best regards
Nick
Post Reply