Matching Computers Against Each Other

[afos99]

Hi all

This may be a novice question, but what is the best way for me to compete my chess machines against each other.

I don't want a long forensic study but something which is fun, not too long-winded but meaningful.

How do I do it and what should I look for as an outcome?

Thanks

Dave

[Monsieur Plastique]

Well a good way to start is to obtain the ratings of your machines from a reliable source. It is not so much important that the actual number is right - more that it comes from a consistent source where a large number of games has been played against a variety of opponents not too dissimilar in overall strength.

I would recommend the SSDF list as the first choice, then there are lists from schachcomputer.info as examples.

You can then decide - depending on how many machines you own and what relative strengths they have - to play individual matches, round robin tournaments, knockout tournaments, swiss style, etc. I think for a beginner a match between two machines of fairly similar rating - alternating colours of 10 rounds - is a good way to start.

Some rules that I have always followed: Play the games to mate, play the move immediately the machine comes up with it on both boards to minimise the effects of pondering and try to avoid repeated opening lines (since without a random option games may end up repeating themselves). Also, be aware that rating differences between computers tend to produce exaggerated outcomes relative to human competition. By that I mean a computer rated 100 points below another may not achieve as good a score statistically speaking to the stronger machine than a human might against another human.

Finally, some programs simply have a lot of "trouble" against other particular programs despite the fact they may be similarly rated. It is the same with human versus computers. For example, I have always had terrible trouble against the strong Kittinger and Schroder programs but have less trouble against Lang and dedicated Morsch programs (this is in relative terms -they are all still significantly stronger than me, but a Sapphire II, for example, feels much tougher to me than a 32 bit Lang does since I have trouble against aggressive programs but less trouble against positional ones). So don't be discouraged if you pit two similar machines against each other and one of them goes down by 8 to 2. That is just the way it is and you may well find the machine that lost could well do better against a third machine than the one it went down to the first time.

I have always done computer versus computer games at 40 moves in 2 hours with no exceptions but obviously many people don't have the time and are not as patient. At the very minimum, I'd be going for a minute per move for any serious computer tournaments.

There is obviously a lot of detail to it all but that is a useful initial overview. A lot of the excitement can be running a swiss style tournament or round robin with the machines seeded by official rating. You then might get an upset or too - it is always exciting to see a stronger machine get into trouble and there is a lot of fun seeing if a computer does or doesn't fall for an horizon effect. Blunders and brilliancies are all part of the fun.

[klute]

I have always done computer versus computer games at 40 moves in 2 hours with no exceptions

Yep - the only way to do it if you're truly serious!

I've just checked my database and I've personally played 2,342 computer vs computer games at 40/2 so far, not counting games I lost the scores to 30 years ago.

I've found that the most machines I can handle at once is eight, although six at a time is more sustainable.

[afos99]

Thanks for the very detailed info. One follow up point about settings / options.

Should I set up machines with tournament book and selective search / brute force? Or out of box with just 40/2 level?

Thanks

Dave

[klute]

Should I set up machines with tournament book and selective search / brute force? Or out of box with just 40/2 level?

There's no absolute "right or wrong" on this point, but personally and where the options exist, I've always used tournament books and selective search.

If I "had my time again" I might have relaxed my tournament book settings and gone with a complete book just for more variety and entertainment.

There's also the decision (if the option exists for a given machine) whether to go with "best" or "random" - I've always gone with "best" settings.

Ultimately, the main point to me is that whatever settings you choose, you should always stick with them and can reasonably justify them.

[Steve B]

I've just checked my database and I've personally played 2,342 computer vs computer games at 40/2 so far, not counting games I lost the scores to 30 years ago.

i have been meaning to ask you Cameron
i know Jon prefers the portable plastic computers (well ANYTHING in plastic really)but i am curious to know if you ever owned a mid-sized to large wooden chess computer?
something along the lines of a Fidelity EAG, Mephisto Exclusive or Novag Super Expert?
i know you once owned the Fidelity Elegance and Elite Champion but these are on the small size.. i am wondering if you ever had something a bit larger and in wood?
with all of those 40/2 rated games you played over the years ..one hopes you gave the larger wooden computers a good run for their money as well as the portables and smaller plastic desktops

Diversity Is Good Regards
Steve

[klute]

with all of those 40/2 rated games you played over the years ..one hopes you gave the larger wooden computers a good run for their money as well as the portables and smaller plastic desktops

Well I think these are my most recent 40/2 statistics and the only Schachcomputer in there you would probably consider "Big Holz" is the Mephisto MM IV in an Exclusive board.

Really the only other "Big Holz" I have is a Chess Master Diamond, but it's a chunky and funky Schachcomputer that just doesn't "do" 40/2 well no matter how much you fiddle with it.

The biggest "Big Holz" I ever owned was unquestionably an AVE ARB 2.5 many years ago - I wish I'd never parted with it now but back then there were these incredible new models called "Super Constellation" and "Turbostar 432" that I needed funds for...

I just remembered I also had a Fidelity Decorator Challenger, which was so huge I had to operate it on the carpet...

Program Elo + - Games Score Av.Op. Draws

1 Emerald : 2058 222 193 15 83.3 % 1778 20.0 %
2 Cosmos : 1953 41 40 320 78.6 % 1727 15.9 %
3 Citrine : 1949 54 53 163 77.0 % 1739 21.5 %
4 Amber : 1921 40 39 302 74.5 % 1735 17.9 %
5 Mephisto MM IV : 1904 44 43 246 72.2 % 1739 15.0 %
6 Miami : 1891 37 37 332 71.7 % 1730 17.5 %
7 Designer Display 2100 : 1839 41 41 242 61.0 % 1762 16.9 %
8 Excel 68000 : 1820 367 367 5 50.0 % 1820 20.0 %
9 Jade : 1815 42 42 202 60.4 % 1741 26.7 %
10 Grandmaster : 1773 36 36 300 53.8 % 1746 17.0 %
11 Super Constellation : 1773 81 81 59 50.0 % 1773 18.6 %
12 Turbostar Kasparov : 1771 40 40 249 52.4 % 1754 13.3 %
13 Elegance 3MHz : 1770 42 42 220 55.0 % 1735 19.1 %
14 Simultano : 1754 71 71 82 48.2 % 1766 13.4 %
15 Mondial II : 1734 207 191 12 70.8 % 1580 25.0 %
16 Turbostar 432 : 1727 186 200 15 36.7 % 1822 6.7 %
17 Conchess 2MHz : 1673 39 40 254 39.0 % 1751 18.9 %
18 Primo : 1645 38 38 271 33.0 % 1768 22.5 %
19 Europa A : 1645 35 36 290 33.8 % 1762 26.2 %
20 Conchess -82 : 1614 334 409 6 25.0 % 1805 16.7 %
21 Advanced Star Chess : 1585 37 38 330 27.9 % 1750 17.0 %
22 Chess 2001 : 1583 47 48 230 25.2 % 1772 12.2 %
23 Express 16K : 1521 41 42 338 22.2 % 1739 10.7 %
24 Sensor XL : 1493 367 367 5 50.0 % 1493 20.0 %
25 Star Opal : 1491 71 74 128 16.8 % 1769 11.7 %
26 SciSys Mark V : 1327 322 154 7 7.1 % 1772 14.3 %
27 Opal Plus : 1281 190 232 22 15.9 % 1570 4.5 %
28 Counter Gambit : 1278 236 103 37 5.4 % 1775 0.0 %

[spacious_mind]

That is very impressive!

[Steve B]

Thanks Cameron
your rating list is fascinating
Please consider posting your entire rating list...

Some Day Regards
Steve