Strange result

IvenGO · Post by **IvenGO** » Sat Mar 17, 2012 5:08 am

Hi all!

It began when I was thinking about best settings for Mephisto Polgar to get most challenging play against. So I asked Ricardo about and he gave me an interesting link with tourney where MP owned 1st place beating all its opponents ( see http://www.anacadigital.com/index.html > Micros dedicados > I Torneo 2001 30' ).
So I started a serie of small matches 30min/game with 2x2 1.d4 d5 and 1.e4 e5 openings and got the followed:

RS2250XL - Zircon II = 1.5 : 2.5
Polgar v1.5 - Zircon II = 2.5 : 1.5 > ok, this one was played good by MP, but then...

Polgar v1.5 - RS2250XL = 0.5 : 3.5 > that was a total crush, every game MP lost by tactics and passive play. Then I had set Random=ON and...

Polgar v1.5 - RS2250XL = 0.5 : 3.5 > the same again! Even in the positions with a big space advantage of MP RS found a way to open the lines and gain a strong initiative.

Any thoughts or suggestions about? Could it be caused by the fact that I used ELO-ACTIVE 2000 level for MP instead of simple BLITZ ?

Larry · Post by **Larry** » Sat Mar 17, 2012 10:26 am

Hi Iven, the results versus RS2250XL might have something to do
with the RS being rated about 70 elo higher than Polgar.
regards,
Larry

IvenGO · Post by **IvenGO** » Sat Mar 17, 2012 11:19 am

Yes, I know about ELO difference, but at that spanish tourney MP had beaten MM-VI (3.5 : 0.5) which is similar to RS; also I read here on forum Ricardo's test matches MP vs RS where MP won! So I wonder much watching a completely other picture of their competition...

ricard60 · Post by **ricard60** » Sun Mar 18, 2012 3:34 am

Hi IvenGo,

In the match between RS and MP i did not use elo active 2000 i just use the 30 sec/move level. MP when it is first turn on it has the 10/sec move level i put it in the 30/sec move level. Some machines does not have an even elo when you go from one time control level to other type of control level. In the spanish forum we are starting to make tournaments at 15 min the whole game (sudden death) and there is a lot of surprises look at this link:

http://meca.mundoforo.com/42-vt1466.html?start=615

The best performance goes to the mephisto supermondial II it gained 72 elo points playing at 15 min blitz. If you scroll down the page you can see a table of all the machines that participate and the elo that had gain or lost at that level. The starting elo comes from the active elo list of schachcomputer.info forum.

IvenGO · Post by **IvenGO** » Tue Jun 26, 2012 4:11 am

Milano shows its higher skill comparing to Polgar (see 1st post):

won vs TurboKing D+ 3,5 : 0,5 { Polgar lost 1 : 3 } > saved double-rook ending having pawn for TK's knight [!]

won vs RS2250XL 3,5 : 0,5 { Polgar lost 0,5 : 3,5 } > saved rook ending having pawn for RS's bishop [!!!]

lost vs Zircon II 1 : 3 { Polgar won 2,5 : 1,5 } = very strangely played both 3rd and 4th spanish games and made tactical mistakes [

]

So I wonder why some people say that they are same strength...

ricard60 · Post by **ricard60** » Tue Jun 26, 2012 10:44 pm

Hi Ivengo,

What settings are you using for polgar?. Polgar can change its playing strength by changing its settings. Probably it can win between 20 and 30 elo points. What time controls did you use?

IvenGO · Post by **IvenGO** » Wed Jun 27, 2012 2:11 am

I set blitz 30min/game generally; also for Polgar SSD=4 and Ran=ON
The other thing is that Milano's book looks bug-fixed/improved.

ricard60 · Post by **ricard60** » Wed Jun 27, 2012 11:06 am

Ran=ON not allways choose the best move. Try it with Ran=OFF

IvenGO · Post by **IvenGO** » Wed Jun 27, 2012 1:22 pm

For Polgar I tried both with the same score vs RS and Z-II.

For Milano: when I set Ran=ON I find that its playing more active; at least against me this setting is tougher to beat.

IvenGO · Post by **IvenGO** » Fri Jun 29, 2012 12:50 pm

Two more games between Milano and RS2250 were played yesterday: French Defence > Classical Variation; in both RS2250 had a strong initiative in middle transformed into one pawn advantage in rook ending, but Milano won both (!!!) by creating passed pawn and moving it forward with active support of its king... so current score is 5.5 : 0.5 in Milano's favor!

ricard60 · Post by **ricard60** » Sat Jun 30, 2012 3:54 am

Hi IvenGo

Are you playing RS2250XL in brute force or selective search?

IvenGO · Post by **IvenGO** » Sat Jun 30, 2012 9:08 am

Selective and Random=ON. Maybe Moscow is an anomaly area for them?

ricard60 · Post by **ricard60** » Sat Jun 30, 2012 11:47 pm

To get a real elo of a machine have to make a lot of games and against different type of dedicated machines. For example in this link :

http://www.zanchetta.net/

and click on tournaments you can see a 6 game match between milano and polgar they ended 3.5 vs 2.5 for polgar. So have to make more games to come up with a real elo.

IvenGO · Post by **IvenGO** » Fri Nov 16, 2012 1:51 pm

After a long break Im back on the track:

Milano vs GK2100 30min/game: 0.5 - 5.5

comparing to result Milano vs RS2250 5.5 - 0.5

IvenGO · Post by **IvenGO** » Sun Nov 25, 2012 10:16 am

RS2250XL vs GK2100 30min/game: 2.5 - 3.5

Also I had performed a fun clone-test by playing 1.f3 d5 2. Kf2 and here RS replied e5 / GK replied e6 (level 30min/game).