Mephisto MMV 5 Mhz vs. Heracles

Designed for posting all types of tournaments and Games (e.g. Man vs. Machine, Computer vs. Computer and basement matches.)

Moderators: Harvey Williamson, Watchman

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
TracySMiller
Full Member
Posts: 865
Joined: Sun Mar 13, 2016 2:24 am
Location: Kingsport, TN

Mephisto MMV 5 Mhz vs. Heracles

Post by TracySMiller »

I have been having so much fun with my new toy, the Mephisto MMV! I can hardly concentrate on my other tournament. It's a pleasure to play a game with it, with it's nice gentle beeps and silky smooth auto response board. I can hang with it on Level 0. After that, it's tough for someone like me.

Anyway, I decided to match it up against another chess program. I found the wonderful Heracles, created by Frenchman Franck Tempet. Here is his program's homepage: http://chessengine.free.fr/wordpress/. It looks like he lost interest and hasn't updated it in 10 years, but the webpage still works, and you can still download his programs. It's interesting that he tested his program against Colossus X running on a CPC 464 (Heracles won all six test games)!

The last version of Heracles is rated 1977 over at the CCRL ratings list. I thought this might be a fun opponent for Mephisto MMV to play. I set up Heracles to run on the equivalent hardware settings the CCRL uses, so the rating would be accurate (Ponder OFF, time settings appropriate, etc). The Mephisto MMV 5Mhz has a rating of 1878 at SSDF.

Heracles won 5.0-1.0, so I suppose it over performed, unless the 1977 CCRL rating is low, or the SSDF 1878 rating is high. Hard to say with just six games, I know. Have any of you forum veterans who have an established chess rating and own the MMV played lots of games against it? What are your results?
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Re: Mephisto MMV 5 Mhz vs. Heracles

Post by spacious_mind »

TracySMiller wrote:I have been having so much fun with my new toy, the Mephisto MMV! I can hardly concentrate on my other tournament. It's a pleasure to play a game with it, with it's nice gentle beeps and silky smooth auto response board. I can hang with it on Level 0. After that, it's tough for someone like me.

Anyway, I decided to match it up against another chess program. I found the wonderful Heracles, created by Frenchman Franck Tempet. Here is his program's homepage: http://chessengine.free.fr/wordpress/. It looks like he lost interest and hasn't updated it in 10 years, but the webpage still works, and you can still download his programs. It's interesting that he tested his program against Colossus X running on a CPC 464 (Heracles won all six test games)!

The last version of Heracles is rated 1977 over at the CCRL ratings list. I thought this might be a fun opponent for Mephisto MMV to play. I set up Heracles to run on the equivalent hardware settings the CCRL uses, so the rating would be accurate (Ponder OFF, time settings appropriate, etc). The Mephisto MMV 5Mhz has a rating of 1878 at SSDF.

Heracles won 5.0-1.0, so I suppose it over performed, unless the 1977 CCRL rating is low, or the SSDF 1878 rating is high. Hard to say with just six games, I know. Have any of you forum veterans who have an established chess rating and own the MMV played lots of games against it? What are your results?
The problem with engines rating lists like CCRL is that they are based solely on engines vs engines. Therefore over time the weaker engines were knocked down because of many losses over strong engines to a level where their rating is no longer realistic and can only be used to compare engine against engine and not engine against human or engine against dedicated computer.

They have tried to improve it in recent years but you will find that probably a ziggurat even performs better than a 1796 dedicated computer or human with same rating. The cause is the rating systems that calculate only wins and losses and not quality of game itself.

Best regards
Nick
TracySMiller
Full Member
Posts: 865
Joined: Sun Mar 13, 2016 2:24 am
Location: Kingsport, TN

Post by TracySMiller »

That's a good insight. So your feeling is that the "real" ratings of the lower-rated engines at CCRL should be higher. Do you think the higher rated programs' ratings are close? They seem to correlate fairly well to the SSDF, at least in the longer time control.
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

TracySMiller wrote:That's a good insight. So your feeling is that the "real" ratings of the lower-rated engines at CCRL should be higher. Do you think the higher rated programs' ratings are close? They seem to correlate fairly well to the SSDF, at least in the longer time control.
No the higher ratings are miraculously set at around 3400 which results in the lower ratings at the bottom once you have a few hundred programs and thousands upon thousands of games.

To more accurately reflect the human to engine you would end up setting a higher bar at the top end so ratings would then end up being around 3700 or so.

Try it play the weakest engine on the list with your PC against MM 6 that is rated at around 2021 ELO at Schachcomputer.info. Ziggurat is 1796 ELO. Play them against each other. If the ratings were to match then MM6 should win around 9-3 in a 12 game match. But it won't probably it will be lucky if it loses 9-3.

Regards
Nick
TracySMiller
Full Member
Posts: 865
Joined: Sun Mar 13, 2016 2:24 am
Location: Kingsport, TN

Post by TracySMiller »

Why did they set a ceiling of 3400? It sounds like from your analysis, the whole CCRL list needs to be shifted upwards 300 points. If that means Stockfish, Komodo, and Houdini are closer to 3700 compared to human reality, then so be it. I am inclined to believe you, that the Dedicated Machines' ratings are closer to being similar to what humans show than the CCRL engine ratings are. I certainly believe that the Mephisto MMV is a lot closer to 1878 (as opposed to 1578), and Heracles is a LOT stronger than 1977, probably closer to 2277.
TracySMiller
Full Member
Posts: 865
Joined: Sun Mar 13, 2016 2:24 am
Location: Kingsport, TN

Post by TracySMiller »

Check out this post from over at Chess.com that adds some interesting information about chess computers and their ratings (versus humans): https://www.chess.com/forum/view/genera ... st_comment
Post Reply