Compltee Chess system in Amiga: LOT faster than in DOS

[Fernando]

It reaches 14-15 in middle game as a windows chess programs of the beginnings of this century. Amazing...

[spacious_mind]

It reaches 14-15 in middle game as a windows chess programs of the beginnings of this century. Amazing...

Not beginning of this century. 1993 = 25 years old!

[Fernando]

It reaches 14-15 in middle game as a windows chess programs of the beginnings of this century. Amazing...

Not beginning of this century. 1993 = 25 years old!

My dear friend, I said "as windows programs of the beginning of XX century", NOT that appertained to that epoch.
I wonder now IF really CCC goes so far OR it computes ply -in Amiga- in a different way.

Fern

[spacious_mind]

It reaches 14-15 in middle game as a windows chess programs of the beginnings of this century. Amazing...

Not beginning of this century. 1993 = 25 years old!

My dear friend, I said "as windows programs of the beginning of XX century", NOT that appertained to that epoch.
I wonder now IF really CCC goes so far OR it computes ply -in Amiga- in a different way.

Fern

It was configured to be either very selective, but that is hard to imagine because all the other Whittington's on same amiga only search about 50% as fast and reach about 8 ply on my computer. If I set up 30 seconds average move and play 1. f3 and follow this with 2. g3 CCC does moves after 2.g3 at 16 ply. Others Whittington's at 8 ply.

[Fernando]

It reaches 14-15 in middle game as a windows chess programs of the beginnings of this century. Amazing...

Not beginning of this century. 1993 = 25 years old!

My dear friend, I said "as windows programs of the beginning of XX century", NOT that appertained to that epoch.
I wonder now IF really CCC goes so far OR it computes ply -in Amiga- in a different way.

Fern

It was configured to be either very selective, but that is hard to imagine because all the other Whittington's on same amiga only search about 50% as fast and reach about 8 ply on my computer. If I set up 30 seconds average move and play 1. f3 and follow this with 2. g3 CCC does moves after 2.g3 at 16 ply. Others Whittington's at 8 ply.

So really CC goes deeper, but not better. Perhaps too much selective. 2150 looks to me as the best.

[Martin Hertz]

2150 looks to me as the best.

I've played hundreds of games with CP2150 DOS at lichess to see the rating against human players.
The conditions: Real DOS @ 3.3 GHz, book on, ponder on, 30 sec/average, what gives 6-7 plies in
the midgame. The lichess classic rating under these conditions varies within 2300-2400 Elo.
The enormous speed helps a lot, because CP2150 does often find better moves in depth 6 or 7.
CP2150 DOS is surely significantly less selective than the Amiga version, taking the speed and
reached ply into consideration. The CP2150 is not that strong, but one of the most interesting
Shannon B programs to play with for humans, due to the special playing style and all the mistakes
and blunders of the B strategy. Maybe I will repeat the experiment on lichess with the speed of a
8088 @ 4.77 MHz (around 7000 times slower I guess), to see the Elo rating at ply 2-3.

[spacious_mind]

2150 looks to me as the best.

I've played hundreds of games with CP2150 DOS at lichess to see the rating against human players.
The conditions: Real DOS @ 3.3 GHz, book on, ponder on, 30 sec/average, what gives 6-7 plies in
the midgame. The lichess classic rating under these conditions varies within 2300-2400 Elo.
The enormous speed helps a lot, because CP2150 does often find better moves in depth 6 or 7.
CP2150 DOS is surely significantly less selective than the Amiga version, taking the speed and
reached ply into consideration. The CP2150 is not that strong, but one of the most interesting
Shannon B programs to play with for humans, due to the special playing style and all the mistakes
and blunders of the B strategy. Maybe I will repeat the experiment on lichess with the speed of a
8088 @ 4.77 MHz (around 7000 times slower I guess), to see the Elo rating at ply 2-3.

Hi Martin,

It is interesting that you are getting 2300-2400 results for CP2150. I have done some research using Mike Watter's site information for CCR and CCN reports.

CP2150 is not really represented well but Chess Champion 2175 is. I don't know if you have CC2175 for PC? In any case CC2175 is approximately as strong as CP2150 in all formats whether it be PC, Amiga or Atari ST.

CCR in 1995 report showed CC2175 as follows:




For a 486-66/DX2 or 486-50 PC, CCR showed CC2175 at 2170 USCF ELO which is almost exactly the same as what CC2175 calls itself.

What Larry Kaufmann did in order to obtain the USCF ELO is to add 100 points to CCN's ELO.


Here is CCN's ELO for 1995:




As can be seen CCN had 2070 Europen (FIDE) rating 100 points less for For a 486-66/DX2 or 486-50 PC.

For rating consistency the following chart was used.




This chart with adjustments between European (FIDE) rating and USCF is consistenly shown by both CCN and CCR in their reports. OCT/Nov 1995 CCN report and 1995 CCR report are good reports to compare against each other as with adjustments they are pretty much the same. CCN showed a speed doubling of approx. 60 ELO.

CCN also showed the approximate performance increase between 486-50-66 MHz as 79 ELO.




My Division 4 Tournament is also a pretty good baseline for comparisons. Below is the final Division 4 Table using USCF rating:




Since I had no idea what the real ratings are I used USCF 2175 for CC2175 and USCF 2150 for CP 2150 as their start rating. I played DOS programs in DOSBox using 69,300 CPU cycles which should exactly match a Pentium 90. Within this tournament DOS CC2175 finished with 2067 USCF and CP2150 finished with 2087 USCF.


In this tournament I often showed two charts with different start values for each program. The reason being is that I was also interested in tracking the what Larry Kaufmann called European rating (FIDE rating).




With European rating PC CC2175 started with 2112 ELO and finished with 2011 ELO. PC 2150 started with 2088 ELO and finished with 2028 ELO playing at P90.


I am sure people who followed this tournament probably took the ratings that I was showing with a great pinch of salt (humoring old Nick) as they seemed much higher than what people are used to seeing from other rating lists for these programs.


In fact my start and end ratings were probably too low because based on CCN and CCR I should probably have added an additioal 79 ELO to the start rating to cover a Pentium 90 rating. So therefore if I were to add 79 ELO to CC2175's final rating of 2011, then I would get ELO 2090 which is almost scary as this would be within 20 ELO to what both CCN and CCR estimated 23 Years ago as the strength of CC2175.


You mentioned CP2150 plays at around 2300 - 2400 ELO at Lichess. If you look at the above chart Amiga CP2150 performed at 3302 ELO. Logically speaking 79 ELO could be added to that which would make it 2381 ELO.


I am still not really conviced that CP2150 is better than CC2175 too few games have been played to know that for sure. I would lean on them being approximately very similar in strength.


Martin, If you do have time and interest I would recommend that you test CC2175 at Lichess using DOSBox set at 69,300 CPU cycles to emulate a Pentium 90 or use DOSBox to emulate a 486-66/50. Because CC2175 is the only Chris Whittington program that you could accurately compare against CCN and CCR reports and match performance against Atari ST and Amiga.

I would do it but just dont have time at the moment as I am too busy fixing my old website since my old computer crashed and I am pretty much having to rebuild every page.


Best regards

[spacious_mind]

I went and looked at Division 3 as well:




I think it is inconclusive which version is better here CP2150, CC2175 and Checkmate Aggressive finished with the same points and Checkmate Aggressive was rated highest. It does seem that PC Complete Chess System is differently configured to Amiga Complete Chess System. It seems that Amiga's high speed search does not make it better. Probably worse.


Best regards

[Fernando]

It would be nice, my dear friends, if you publish here the estimate strength of all those babies NOW, with current processors running with 3-4Gb in the CPU or more and 3-4 or more Gib in RAM.

I fell that with CM2000, 2100, CP 2150, Sargon etc I am playing above 2250 creatures in all cases and pehaps more than 2300 in Sargon III case IF in Amiga.

Waiting current data regards
Fern

[Martin Hertz]

Thanks Nick for all the data. It's true that CP2150 reaches around 2350 Elo on LiChess, but the FIDE Elo would be surely lower.
I've tested CC2175, CCSystem and CP2150 some times ago against human players and there was no significant difference. The stronger
players told me, that the playing style of CP2150 is the most entertaining and human like. Playing CP2150 on very high speed agianst
some other DOS programs shows that CP2150 and surely CC2175 and CCSystem are not that strong:

Code: Select all

                         MC5   Gid   CG3   WCh   Soc   MC1   CST   Fr2   Rex   Psi   Col   CPl   Tur

 1. M-Chess Pro 5.0       *    2.0   2.5   2.5   4.0   3.0   3.0   3.5   2.5   4.0   3.5   3.5   4.0     38.0/48   2530

 2. Gideon Pro 1.0       2.0    *    2.0   2.5   2.0   2.5   2.5   3.0   3.5   4.0   4.0   4.0   4.0     36.0/48   2490

 3. Chess Genius 3.0     1.5   2.0    *    2.0   3.0   3.0   3.0   3.0   4.0   3.0   3.0   4.0   4.0     35.5/48   2480

 4. WChess 1.05          1.5   1.5   2.0    *    3.5   2.0   1.5   2.5   3.0   3.0   3.0   4.0   4.0     31.5/48   2410

 5. Socrates 3.0         0.0   2.0   1.0   0.5    *    2.5   2.5   3.5   4.0   3.0   3.0   3.5   4.0     29.5/48   2380

 6. M-Chess 1.70         1.0   1.5   1.0   2.0   1.5    *    3.0   1.5   2.0   4.0   4.0   3.0   4.0     28.5/48   2370

 7. Chess System Tal     1.0   1.5   1.0   2.5   1.5   1.0    *    2.5   2.0   4.0   4.0   3.5   3.5     28.0/48   2360

 8. Fritz 2.51           0.5   1.0   1.0   1.5   0.5   2.5   1.5    *    3.0   3.0   3.5   3.5   4.0     25.5/48   2320

 9. RexChess 2.30        1.5   0.5   0.0   1.0   0.0   2.0   2.0   1.0    *    4.0   3.0   4.0   3.0     22.0/48   2270

10. Psion 2.13           0.0   0.0   1.0   1.0   1.0   0.0   0.0   1.0   0.0    *    3.0   3.5   4.0     14.5/48   2150

11. Colossus X           0.5   0.0   1.0   1.0   1.0   0.0   0.0   0.5   1.0   1.0    *    2.0   3.5     11.5/48   2100

12. Chess Player 2150    0.5   0.0   0.0   0.0   0.5   1.0   0.5   0.5   0.0   0.5   2.0    *    3.0      8.5/48   2030

13. Turbo Chess          0.0   0.0   0.0   0.0   0.0   0.0   0.5   0.0   1.0   0.0   0.5   1.0    *       3.0/48   1830

[Cyberchess]

This brings back memories....

Back around late 1990 Steven Schwartz – the founder of I.C.D. – sold me a copy of Mchess for $100. I had known Steven Schwartz for many years and had purchased several chess computers from I.C.D., though also desired a strong program that would take advantage of the greater computational power of my DOS based 386 PC. This early version didn’t even have mouse input capability, but featured some very nice looking EGA graphics and was substantially stronger than CM2k, Sargon III and Rex. Soon thereafter I would head back to I.C.D. for a free upgrade to a version with mouse support, and I eventually purchased the Mchess Professional package when it became available. These came on special copy protected 3&1/2” discs that kept track of your install count. This required that you first uninstall the program before installing it on a different system. If you were unable to achieve this due to a hardware failure, there was a number to call for a reset code.

Marty Hirsch, the NASA programmer that wrote Mchess, had his heyday in 1995 when he won the World Microcomputer Chess Championship.
Mchess defeated several Grandmasters then – a milestone achievement for computer chess.

GM Zsuzsa Polgar

Mchess Pro

We dubbed the program “The eMulator”.

Zero Gravity Chess Regards,
John

[spacious_mind]

Thanks Nick for all the data. It's true that CP2150 reaches around 2350 Elo on LiChess, but the FIDE Elo would be surely lower.
I've tested CC2175, CCSystem and CP2150 some times ago against human players and there was no significant difference. The stronger
players told me, that the playing style of CP2150 is the most entertaining and human like. Playing CP2150 on very high speed agianst
some other DOS programs shows that CP2150 and surely CC2175 and CCSystem are not that strong:

Code: Select all

                         MC5   Gid   CG3   WCh   Soc   MC1   CST   Fr2   Rex   Psi   Col   CPl   Tur

 1. M-Chess Pro 5.0       *    2.0   2.5   2.5   4.0   3.0   3.0   3.5   2.5   4.0   3.5   3.5   4.0     38.0/48   2530

 2. Gideon Pro 1.0       2.0    *    2.0   2.5   2.0   2.5   2.5   3.0   3.5   4.0   4.0   4.0   4.0     36.0/48   2490

 3. Chess Genius 3.0     1.5   2.0    *    2.0   3.0   3.0   3.0   3.0   4.0   3.0   3.0   4.0   4.0     35.5/48   2480

 4. WChess 1.05          1.5   1.5   2.0    *    3.5   2.0   1.5   2.5   3.0   3.0   3.0   4.0   4.0     31.5/48   2410

 5. Socrates 3.0         0.0   2.0   1.0   0.5    *    2.5   2.5   3.5   4.0   3.0   3.0   3.5   4.0     29.5/48   2380

 6. M-Chess 1.70         1.0   1.5   1.0   2.0   1.5    *    3.0   1.5   2.0   4.0   4.0   3.0   4.0     28.5/48   2370

 7. Chess System Tal     1.0   1.5   1.0   2.5   1.5   1.0    *    2.5   2.0   4.0   4.0   3.5   3.5     28.0/48   2360

 8. Fritz 2.51           0.5   1.0   1.0   1.5   0.5   2.5   1.5    *    3.0   3.0   3.5   3.5   4.0     25.5/48   2320

 9. RexChess 2.30        1.5   0.5   0.0   1.0   0.0   2.0   2.0   1.0    *    4.0   3.0   4.0   3.0     22.0/48   2270

10. Psion 2.13           0.0   0.0   1.0   1.0   1.0   0.0   0.0   1.0   0.0    *    3.0   3.5   4.0     14.5/48   2150

11. Colossus X           0.5   0.0   1.0   1.0   1.0   0.0   0.0   0.5   1.0   1.0    *    2.0   3.5     11.5/48   2100

12. Chess Player 2150    0.5   0.0   0.0   0.0   0.5   1.0   0.5   0.5   0.0   0.5   2.0    *    3.0      8.5/48   2030

13. Turbo Chess          0.0   0.0   0.0   0.0   0.0   0.0   0.5   0.0   1.0   0.0   0.5   1.0    *       3.0/48   1830

Hi Martin,

Thanks for the chart. Let me also show you some charts:




Left is CC2175 P90 Emulation and right is Amiga 68060 CC2175 Emulation. In the above the search was done for exactly 1 minute. Both the P90 and Amiga 68060 seemed to reach exactly the same search depth, but as you can see after e7e5 they varied the move choice. Yet if you look in Division 4 Amiga CC2175 finished 2.5 points and almost 200 ELO better. Yet the above chart shows similar search depth. The reason is quite obvious really. For the Amiga you have to write a 16 bit program. For DOS it was 8 Bit. Totally differently written programs. Searching same depth Amiga program is much stronger.



Whoever put the 8 Bit program together for PC did an even worse job with CP2150. CP2150 on PC is slower whereas CP2150 on Amiga searches the same depth as CC2175. ELO difference between the two = 3.5 points better in 12 games and nearly 300 ELO. This is really substantial. Also notice that PC version is the only one of the Whittington's including CST shown below that wants to reply with d7-d5? Everyone else is e7-e5! Are they really the same programs????




With Sargon 3, not only is the Sargon 3 8 Bit a worse program it is also 2 ply slower in 1 minute. There is no question that the program is worse, it finished 4th from last in Division 4 whereas Sargon 3 is fighting at the top of Division 2 in second place.



CST at P90 searches 9 deep 1 ply less than Amiga Sargon 3, yet if you look at your table and compare Division 2 Amiga Sargon 3 beat in Division 2 CST and is ahead of WChess 1.05 which you show as finishing 4th.

I don't know what ELO's you used or if you adjusted them, so it interesting to know what depth CST and WChess 1.05 are searching in your tournament? I am guessing you are using the old published ratings for these computers in your tournament but you are running the programs at 3 GHz if I understand corretly?

If you are achieving with PC CP2150 2300 in matches, then I would be even more convinced that CP2150 Amiga at 3 GHz will be more likely be 2800 ELO and who knows where Amiga Sargon 3 would be, its improvement is astonishing maybe even 3000 at 3 GHz PC speed if that were possible.

These are all 16 Bit programs and it seems to me that they perform much better than old same named 8 Bit programs.



Above you can see Atari ST at 32 MHz. You can see that both are much slower than PC or Amiga. What I will do when I have time is play Atari 32 Mhz CP2150 and CC2175 against the PC versions at P90 speed and see what happens. It would not surprise me if the slower Atari search beats the faster PC P90 search.


Best regards

[Martin Hertz]

Yes, the Whittington programs are not well suited to compare Amiga vs PC, bacause they are too different and apparently not well ported to the PC.
You've asked for the speed at 3.3 GHz, so here the results after the move f3: WChess needs 4 sec for finishing depth 11 and CST needs 9 sec for
finishing depth 10. I'll make further tests with CC2175 on lichess and will report, but it can take a while.

[spacious_mind]

Yes, the Whittington programs are not well suited to compare Amiga vs PC, bacause they are too different and apparently not well ported to the PC.
You've asked for the speed at 3.3 GHz, so here the results after the move f3: WChess needs 4 sec for finishing depth 11 and CST needs 9 sec for
finishing depth 10. I'll make further tests with CC2175 on lichess and will report, but it can take a while.

Hi Martin


Thanks, wow substantial improvement for CST who takes 60 seconds for Ply 9/1 on 69,300 cpu cycles (P90). Perhaps when you get a chance you can post what it gets for 60 seconds I would estimate maybe 12/13 ply?

Best regards

[Fernando]

Yes, the Whittington programs are not well suited to compare Amiga vs PC, bacause they are too different and apparently not well ported to the PC.
You've asked for the speed at 3.3 GHz, so here the results after the move f3: WChess needs 4 sec for finishing depth 11 and CST needs 9 sec for
finishing depth 10. I'll make further tests with CC2175 on lichess and will report, but it can take a while.

Hi Martin


Thanks, wow substantial improvement for CST who takes 60 seconds for Ply 9/1 on 69,300 cpu cycles (P90). Perhaps when you get a chance you can post what it gets for 60 seconds I would estimate maybe 12/13 ply?

Best regards

You put the CSTAL I here. Why? You say nothing about it.