Excellence/Par Excellence

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.
Post Reply
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Excellence/Par Excellence

Post by Reinfeld »

I realize this topic is a tar pit, but let me get past a couple of obvious points. I understand that dedicated-machine ratings are suspect, inflated, deflated, distorted by various intricate factors. So it's not a perfect system. I get that.

I'm trying to understand the true difference between these two machines. What I read here

http://electronicchess.free.fr

tells me that:

1. Excellence was first. There are versions with 3MHz and 4MHz.
2. Par Excellence came next. One version, 5MHz.

Supposedly, Par Excellence is stronger.

(Stupid side question: how does one tell the MHz level, anyway?)

Additional factors:
3. Designer 2000/Display contains the Excellence program. (with no mods?)
4. Designer 2100/Display/Chesster contains the Par Excellence program at 6 MHz. (with no mods?)

Ergo,

5. Designer 2000 is a clone of Excellence at 3 and 4MHz; (with no mods?)
6. Designer 2100 is a clone of Par Excellence at 6MHz(?) (with no mods?)

If true:

- Excellence and Designer 2000 ratings (equal MHz) should behave like clones.
- Par Excellence ratings should be higher than Excellence
- Designer 2100 should be stronger than Par Excellence

Given all this, consider these numbers;

I. Current INDIVIDUAL entries on wiki-elo-liste (same numbers as SSDF above):

Designer 2100 Display (6102 6MHz) - no rating
Par Excellence (6502 5 MHz) 1835
Excellence (6502 4 MHz) - 1801
Designer 2000 Display (6502 3MHz) - 1760

http://www.schach-computer.info/wiki/in ... igner_2100
http://www.schach-computer.info/wiki/in ... Excellence
http://www.schach-computer.info/wiki/in ... Excellence
http://www.schach-computer.info/wiki/in ... 00_Display

At this point, all appears to be right with the world.


II.
FULL LIST posting on:
http://www.schach-computer.info/wiki/in ... -Elo-Liste

Designer 2100/Display (6102 6MHz) - 1873
Excellence (6502 4 MHz) - 1851
Designer 2000/Display (6502, no MHz listed) - 1828
Par Excellence (6502 5mhz) - 1822
Excellence (6502 3mhz) - 1789

These last numbers appear to reflect the active rating list on the same site, which issued its latest numbers in 2010 as far as I can tell. And now we're at the root of my question:

How is it that Excellence/Designer 2000 appear to outdo Par Excellence?

I can see that the differences are minimal. I'm trying to understand whether these underlying programs (by the Spracklens) differ in any way. If not, why?

Thoughts?

- R.
Mike Watters
Member
Posts: 429
Joined: Fri Sep 26, 2008 12:31 pm
Location: Milton Keynes
Contact:

Post by Mike Watters »

Reinfeld

I think the answer to your question "How is it that Excellence/Designer 2000 appear to outdo Par Excellence?" lies in the relatively low numbers of games played -

Par Excellence - 69
Designer 2100 - 85
Designer 2000 - 24
Excellence 4 MHz - 115
Excellence 3 MHz - 54

Eric Hallsworth's Selective Search includes a very large number of results produced worldwide dating from the mid 1980s. The current (2011) Selective Search Elo ratings for those chess computers are -

Par Excellence/Designer 2100 - 1829
Excellence 4 MHz - 1783
Excellence 3 MHz/Designer 2000 - 1754

Back in 2005, the last time Eric published the figures, the number of games played was 2461, 1637, 1607 respectively. So the ratings will almost certainly be much more accurate (approx +- 10 Elo points)

The Wiki Active Elo list also gives a +- confidence range on their Elo ratings (+- 75 for the Par Ex) and you can also see exactly which machines each played and the results. Wiki is great for the large number of machines tested and the ongoing testing efforts, but for rating accuracy for the oldies I would be using Selective Search.

All the best
Mike
IanO
Member
Posts: 162
Joined: Sat Aug 15, 2009 11:43 pm
Location: Portland, OR

Post by IanO »

Mike has it right, the Active list has far fewer games in its rating pool for many of its devices. You should examine the detailed results for a device to see how well it was tested. For example, for the anomalous 4 MHz Excellence:

Code: Select all

139 Excellence 4 MHz          : 1851  115 (+ 53,= 27,- 35), 57.8 %

GK 2000                       :  10 (+  0,=  3,-  7), 15.0 %
Mondial II                    :   6 (+  4,=  1,-  1), 75.0 %
Advanced Starchess            :  20 (+ 15,=  4,-  1), 85.0 %
Kishon Chesster               :   4 (+  0,=  1,-  3), 12.5 %
Europa Europa A Marco Polo    :   4 (+  3,=  0,-  1), 75.0 %
MM II + HG240                 :  10 (+  3,=  4,-  3), 50.0 %
Rebell 5.0                    :  20 (+  5,=  7,-  8), 42.5 %
Grandmaster / Igor            :  10 (+  5,=  3,-  2), 65.0 %
D 6 MHz                       :   4 (+  2,=  1,-  1), 62.5 %
Dominator 2.04                :   4 (+  0,=  0,-  4),  0.0 %
Challenge / Regency           :  14 (+  8,=  3,-  3), 67.9 %
Chess Station                 :   3 (+  3,=  0,-  0), 100.0 %
Mark V                        :   2 (+  2,=  0,-  0), 100.0 %
Designer 2000 / Display       :   4 (+  3,=  0,-  1), 75.0 %
Many of these matches are only three games long (which is actually pretty good for the Active list; many matches only consist of a single game!) In contrast, the SSDF requires at least 20 games between opponents before a match is entered into their rating pool.

There is a further explanation which applies to some of the devices. Many early machines were tuned to the tournament time controls used in competitions and early rating lists like the SSDF: 40 moves in 2 hours (3 min/move). Playing at faster time controls like the Active list's G/30 or 30 seconds per move may fall into a machine's "sour spot".

Also puzzling to me is that the slower 26 MHz Sapphire/Diamond have higher Active ratings than the faster 32 MHz Sapphire/Diamond II. And that Lang's London update sometimes tests weaker than the previous Genius or Vancouver versions.

Ian
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

Ian's post begs new questions:

1. If Designer/Display 2000 is the Excellence program (assuming 4 MHz), aren't we talking about clones? I realize it's a chore, but it would be interesting to see a clone test of these models, similar to what Spacious Mind published on the GK 2100 models.

2. If older machines were tuned to tournament time controls, what is the true measure of strength? My inclination would be to run games among the models at 3 min/move, to avoid the so-called sour spots.

3. SSDF standards (20 games) strike me as much more fitting. One would assume that such a match between Excellence and Designer 2000/display would, over time, lead to even results. Right?

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

Post by spacious_mind »

Mike Watters wrote:Reinfeld

I think the answer to your question "How is it that Excellence/Designer 2000 appear to outdo Par Excellence?" lies in the relatively low numbers of games played -

Par Excellence - 69
Designer 2100 - 85
Designer 2000 - 24
Excellence 4 MHz - 115
Excellence 3 MHz - 54

Eric Hallsworth's Selective Search includes a very large number of results produced worldwide dating from the mid 1980s. The current (2011) Selective Search Elo ratings for those chess computers are -

Par Excellence/Designer 2100 - 1829
Excellence 4 MHz - 1783
Excellence 3 MHz/Designer 2000 - 1754

Back in 2005, the last time Eric published the figures, the number of games played was 2461, 1637, 1607 respectively. So the ratings will almost certainly be much more accurate (approx +- 10 Elo points)

The Wiki Active Elo list also gives a +- confidence range on their Elo ratings (+- 75 for the Par Ex) and you can also see exactly which machines each played and the results. Wiki is great for the large number of machines tested and the ongoing testing efforts, but for rating accuracy for the oldies I would be using Selective Search.

All the best
Mike
Hi Mike,

I agree with your comments about the low number of games, but I do not necessary agree that the Designer 2000 Display is weaker than a Par Excellence due to its lower MHz.

I have played both quite a lot and although the Designer 2000 Display only has 3 MHz compared to the Par Excellence at 5 MHz, I have a good feeling that the Designer 2000 may be stronger than the Par Excellence. I at least have found it to play more consistently better.

I suspect there may be differences in the Programs. You also have a Designer 2100 without Display which has different level settings to a Designer with Display and also it has only 5 MHz compared to the 6 MHz of the Designer 2100 Display. Therefore these two machines are different.

Here is a tournament in which I played the Designer 2000 Display. An Elite Budapest 8MHz was also present in this tournament.

http://www.schach-computer.info/wiki/in ... rnier_2009

I am not really much of a fan of bundling computers for ELO purposes unless they have proven that they repeat all the moves of one another.

Besides with Fidelity who can say that every model always kept the same insides, or did they sometimes use what was available in stock? :)

Best regards,

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

Post by spacious_mind »

OOOps I forgot to mention...

Designer 2100 Display Info Ratiing for active play (30 seconds per move) shows a difference of +/- 66

The difference between Par Excellence and 2100 Display shows as 51 ELO. This is becoming very close to confirming at least in active play (30 seconds per move) that Par Excellence and 2100 Display are not the same.

What makes the situation worse is that the 3 versions of Phantom and also probably the 2100 (no display) are also bundled into the ELO rating shown at Wiki and Selective Search which in my mind makes the rating very unreliable unless you want to know the average rating of 6 chess computers:

Designer 2100
Desinger 2100 Display
Par Excellence
Phantom
Chesster Phantom
Eyeball Chesster Phantom

= Wiki Rating X or Selective Rating Y

Even the average rating is skewed since you don't know which machine played which game or how many games. Therefore even my average rating statement is probably incorrect.

Best regards,

Nick
Nick
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

First - I am a troglodyte, repeating what must be an elementary side question: how does one tell the MHz level?

Second, for what it's worth, my oldest machines are Excellence and Designer 2000, and I've always had an easier time beating Designer 2000 (leaving both at default level.)

Third, Mike and Nick combined paint this pretty picture:

POTENTIAL CLONE 1:
Excellence (3 MHz)
- Designer 2000 (3MHz only?)
- Designer 2000 display (3MHz only?)

POTENTIAL CLONE 2:
Excellence (4MHz)
- Designer 2000 (4MHz?)
- Designer 2000 display (4MHz only?)

POTENTIAL CLONE 3:
Par Excellence (5MHz)
- Designer 2100 (5MHz)
- Designer 2100 Display (6MHz) - debated

Mike says low numbers of games skew ratings. Higher numbers reflect truer trends. Must agree. He cites Hallsworth's ratings like this:

Par Excellence/Designer 2100 - 1829 (not clear whether Display is bundled here)
Excellence 4 MHz - 1783
Excellence 3 MHz/Designer 2000 - 1754 (not clear that Display is bundled.)

All this suggests that standard assumptions apply.

Nick argues that Designer 2100 Display is stronger than Par Excellence, and perhaps (gasp) an entirely different beast. Prove the clone before you bundle ratings, he says, which also seems reasonable. Also suggests (gasp) that there are 3 versions of 2100 No Display (is there more to this? what would the "versions" be?)

Leave Phantom out of this for the moment.

Based on the tree above, I know where the clones SHOULD be:

CLONE THE FIRST
Excellence (3MHz)
Designer 2000 (3MHz)
Designer 2000 Display (3MHz)

Note:
- Nick finds D2D (3MHz) stronger than Par Excellence (5 MHz).

CLONE THE SECOND - UNCLEAR
Excellence (4MHz)
Designer 2000 (4MHz) - Don't know if this is right

CLONE THE THIRD
Par Excellence (5 MHz)
Designer 2100 (5 MHz)

SEPARATE MACHINE?
Designer 2100 Display

The clone-verification standard Spacious Mind used on the GK 2100 seemed reasonable to me: 90+ percent, as a working boundary.

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

Post by spacious_mind »

Reinfeld wrote:First - I am a troglodyte, repeating what must be an elementary side question: how does one tell the MHz level?

Second, for what it's worth, my oldest machines are Excellence and Designer 2000, and I've always had an easier time beating Designer 2000 (leaving both at default level.)

Third, Mike and Nick combined paint this pretty picture:

POTENTIAL CLONE 1:
Excellence (3 MHz)
- Designer 2000 (3MHz only?)
- Designer 2000 display (3MHz only?)

POTENTIAL CLONE 2:
Excellence (4MHz)
- Designer 2000 (4MHz?)
- Designer 2000 display (4MHz only?)

POTENTIAL CLONE 3:
Par Excellence (5MHz)
- Designer 2100 (5MHz)
- Designer 2100 Display (6MHz) - debated

Mike says low numbers of games skew ratings. Higher numbers reflect truer trends. Must agree. He cites Hallsworth's ratings like this:

Par Excellence/Designer 2100 - 1829 (not clear whether Display is bundled here)
Excellence 4 MHz - 1783
Excellence 3 MHz/Designer 2000 - 1754 (not clear that Display is bundled.)

All this suggests that standard assumptions apply.

Nick argues that Designer 2100 Display is stronger than Par Excellence, and perhaps (gasp) an entirely different beast. Prove the clone before you bundle ratings, he says, which also seems reasonable. Also suggests (gasp) that there are 3 versions of 2100 No Display (is there more to this? what would the "versions" be?)

Leave Phantom out of this for the moment.

Based on the tree above, I know where the clones SHOULD be:

CLONE THE FIRST
Excellence (3MHz)
Designer 2000 (3MHz)
Designer 2000 Display (3MHz)

Note:
- Nick finds D2D (3MHz) stronger than Par Excellence (5 MHz).

CLONE THE SECOND - UNCLEAR
Excellence (4MHz)
Designer 2000 (4MHz) - Don't know if this is right

CLONE THE THIRD
Par Excellence (5 MHz)
Designer 2100 (5 MHz)

SEPARATE MACHINE?
Designer 2100 Display

The clone-verification standard Spacious Mind used on the GK 2100 seemed reasonable to me: 90+ percent, as a working boundary.

- R.
Hi Reinfeld,

I don't think I stated that there was a third Designer 2100 version :) Let us stick with two.

Also I am not aware of a Designer with 4 MHz.

The only ones that I would consider the same (but I have not looked into this) would be Designer 2000 and Designer 2000 Display.

I would treat them all as unique until they have been tested further. Having the same MHZ does not mean you have the same software program running in differently named models with the same CPU.

This assumption was generally made in the past by testers by opening the machine looking at the ROM chip and the clock quartz. But this only gives you an indication that there may be a same program being sold here.

It does not take into account any possibilities of program modifications. For that you would need to do other Tests.

In addition it almost unlikely that two programs with different MHZ, even if they are the same can have the same ELO rating. The faster clock speed of one will ensure that in the end it is theone with a higher ELO rating.

Best regards,

Nick
Nick
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

Nick -

I am going to try a clone test with Excellence, Designer 2000, Designer Display 2000, and Designer Display 2100, following the protocol on the GK 2100 clone tests listed on the Spacious Mind site. (That is you, yes?)

But first, I have a question:

1. How did you choose the game model for the test? What was your thinking?

- R.
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

P.S.

Nick said:
The only ones that I would consider the same (but I have not looked into this) would be Designer 2000 and Designer 2000 Display.
At the very least, I think it's possible to establish this one for certain.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Reinfeld wrote:Nick -

I am going to try a clone test with Excellence, Designer 2000, Designer Display 2000, and Designer Display 2100, following the protocol on the GK 2100 clone tests listed on the Spacious Mind site. (That is you, yes?)

But first, I have a question:

1. How did you choose the game model for the test? What was your thinking?

- R.

Hi Reinfeld,

If you want to try the Test, then just play the two Designers against each other if possible with Ponder Off and if there is a Random setting then turn that off as well. Afterwards switch colors and try and let the computers repeat the game.

You can also play a second game with Ponder ON and Random ON and see what difference this makes.


Just take back any move that is not the same and note it. You will have to force the correct move each time they play a different move.

Thanks

Nick
Nick
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Back to this issue (Attention, Spacious Mind!)

Post by Reinfeld »

Seems like the following excerpt underscores Spacious Mind's hypothesis that Designer 2100 Display and Par Excellence are different machines.

Source is here:
http://www.schach-computer.info/wiki/in ... Challenger

Money quote:
Although the Fidelity Chesster Challenger is based on the Fidelity Designer 2100 there are a number of important differences. Firstly the clock speed is 5 MHz rather than 6 MHz. The opening book is 16,000 ply and designed for variety compared with the 28,000 ply opening book of the Fidelity Designer 2100, and the narrower optimised 16,000 ply opening book of the Fidelity Par Excellence.
So it's Chesster contrasted with Designer 2100, but surely the related differences are obvious:

Par Excellence/Chesster: 5 MHz
Designer 2100: 6 MHz

Par Excellence/Chesster: 16K ply opening book (built for variety)
Designer 2100: 28K ply opening book

I think it was Nick who said MHz alone didn't necessarily dictate strength. But what about MHz plus a broader opening book? Is that worth 50 rating points?

Inevitably you get into a moral discussion of where random factors begin and end: Is it just the same beast, overlaid with a bigger opening book and jacked-up power? If you add those parameters, can you still call it the same thing?

- R.

P.S.: Observation-
The site above says Chesster is a Designer 2100 variant. The site below says Chesster is a Par Excellence variant:

http://electronicchess.free.fr/heroes.html#chesster
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
yoyo_chessboard
Member
Posts: 126
Joined: Mon Oct 17, 2016 5:44 pm
Location: South of France - cote d'azur

par excellence bugs ?

Post by yoyo_chessboard »

hi

does the par excellence program has known bugs ?

why that strange question ?
when putting side by side par excellence program and designer 2000 program (in assembly code), there are very few differences.
it is as if the designer 2000 was a par excellence with some bugs corrected.
(i expect that there are not any regressions or bugs added).
there are less than ten minors codes changes.

of course the two boards don't cruise at same speed (5 mhz vs 3 mhz) but that isn't the question.

regards.
User avatar
mclane
Senior Member
Posts: 1600
Joined: Sun Jul 29, 2007 9:04 am
Location: Luenen, germany, US of europe
Contact:

Post by mclane »

One can differentiate excellence and par ex very good by the different playing style.

Excellence plays good with pawn structures and plays positional, but lacks tactics, while par ex is opposite.
It plays horrible with pawns and produces double and triple pawns, often isolated double or triple pawns !!!
But overall par ex is stronger in tactics.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
User avatar
Fernando
Admiral of the Fleet
Posts: 3059
Joined: Tue Jul 31, 2007 4:35 pm
Location: Santiago de Chile

Pragmatic Approach

Post by Fernando »

It could also be adopted a pragmatic. no mathematical approach and say that in average all those machines are very similar, all of them moves inside the same territory, with minor differences as much 50 elo points are not that big a difference anyway. All of them are spracklen programs of the same age and so all of them play a similar kind of game.
I consider a better approach to locate machines in segments or "territories" inside which differences comes from chaos, kind of test, how we human played that day, humor of the program some other days, etc.

Now If we are anyway interested in detailed differences, probably there is NO machine same as other, no clones, because even a change of the brand of the processor makes a difference in perfomance. I am not processor savvy, but I know so far that the internal architecture in the treatment of data, threads, etc, makes a difference.
Festina Lente
Post Reply