Dave C wrote:Hi Nick,
Excellent updates! I was a little surprised that the London did not score a little higher than the Vancouver TM. I mistakenly assumed each update from Mr. Lang would result in a slight (or more) increase in strength. Is there some other aspect to playing that the London excels at that is not evident in these tests? Also interesting to see how hard it is to make significant ratings increases from a strictly software prospective. A few points here and a few points there...
Very informative and interesting results....thanks.
Dave
Hi David,
London is the most passive of all the Lang's. Supposedly it improves at longer time settings. But I am not sure if that is 100% proven yet, since no one really plays enough games at tournament level to absolutely be able to convincingly prove it. There really is not that much difference in strength between Vancouver and London. Lyon also has its day as you can see from the 1992 CCR rating where Lyon shows higher than Vancouver. Even at Schachcomputer.Info for a long time if you have been observing the Active List over the past 10 years the Lang's with Lyon, London and Vancouver flip and flop. In the past Lyon was rated higher than Vancouver and now it is Vancouver over Lyon, same with London.
Unlike humans, when you play games, the opening book moves are randomly picked by the computer out of its tournament book. With that in mind it is possible to have a sequence of a dozen games where a computer picks really bad openings for itself and its rating drops and vice versa its rating increases. So for dedicated computers, all rating lists are a little fluid and best used as good guides, but not as absolutes.
Same with the above tests. With only 5 tested games TM Vancouver completes the tests better than TM London. The next 5 different tests could flip this around.
I would say that Mephisto Genius is the strongest overall, of all the 68030 computers.
Below is a Revelation Tournament that was completed a couple of years ago where each computer played 2 games against everyone, 62 games in total for each computer.
As you can see Revelation Vancouver finished higher than any of the TM's with the exception of TM Vancouver, yet it scores 122 ELO points less in my tests and 121 ELO points less than TM London.
Here are the results from the same tournament with the TM's playing each other twice:
TM Lyon finished clearly first over TM London and TM Vancouver.
Below is another comparison by adding both Mephisto Genius and Mephisto Genius London:
Now Mephisto Genius finishes first by a clear margin.
Here is another list from the same Tournament with all the Revelation Lang's added as well:
And Mephisto Genius wins it by 1/2 point over Mephisto Genius London, Revelation Vancouver and TM Vancouver.
Take away the Lang results from the overall Tournament score and they performed as follows against other computers, which again other than Genius showing best again, does shows different standings for the rest.
1. Mephisto Genius - 44.5 - 17.5 = 27 points out of 36.
2. Mephisto TM Vancouver - 42.5 - 17 = 25.5 points out of 36.
3. Mephisto TM London - 39 - 15.5 = 23.5 points out of 36.
4. Mephisto Genius London - 40 - 17 = 23 points out of 36.
4. Revelation Vancouver - 40 - 17 = 23 points out of 36.
6. Revelation Lyon - 37 - 14.5 = 22.5 points out of 36.
7. Mephisto TM Lyon - 36 - 14.5 = 21.5 points out of 36.
7. Revelation London - 36 - 14.5 = 21.5 points out of 36.
7. TM Portorose - 34.5 - 13 = 21.5 points out of 36.
10. Revelation Portorose - 26.5 - 11 = 15.5 points out of 36.
Is Mephisto Genius really the strongest? Probably based on all the above scenarios. Regarding the rest of them it is still very fuzzy between Vancouver, London and Lyon.
Best regards
PS. ... Resurrection Ruffian finished a lowly 8th in the mammoth tournament yet in the Schachcomputer.Info list it shows up as 2nd. As I said things change all the time and everything is just a guide in the end.