Milano Pro / Mephisto Master Chess nodes per second

[ricard60]

The mephisto master chess has a weak endgame but it is very stong on middlegame, it's tactic is what gives it a better elo than citrine. Now i have played 11 games between citrine and masterchess and it goes 4,5 vs 6,5 for master chess; what i see in those 2 games that got ahead is that because of its tactics; master chess could won 2 pawns in the middlegame and got to an endgame with this 2 full pawns and citrine resigned. The other game masterchess got a rook and a pawn for a knight and after 10 more moves citrine also resigned. You can also see the test 2630 y 2450 where masterchess has a better performance than citrine and this is because the positions that are shown mostly are realted to middlegame.

strong tactics regards
Ricardo

[ricard60]

I do not have the games on pgn but for the rest 12 games i will conect citrine to a pc so i can store them.

no memory regards
Ricardo

[mackgra]

Monsieur Plastique,

Running the position you give on the first page of this thread i can confirm the Mephisto Atlanta's move choice at different levels.

The Atlanta considers Rxh5 during the first 7-8 seconds of its search then switches to Rd5+ after this. It has to search to 10 ply (which takes about 7 sec) before it considers the alternative Rd5+ as being a better move.
On the average time 30 seconds level it doesn't complete the full 10 ply search and so chooses Rd5+ as its move.
When set to average time 1 min level it completes a full 10 ply search and chooses a6 as the best move. Therefore it needs to be on an average responce time of 10 seconds or more in order to see Rxh5? as being a bad move. Hope this is of help to you

Regards

Graham

[Monsieur Plastique]

Hi Graham,

I really appreciated you taking the time to test Atlanta for me. I hope you are recovering from your virus!

I strongly suspected Atlanta would find the "right" move. Yesterday evening I was looking at the games played by Atlanta in the Hague tournament of 1997. I was analysing all the games it lost (all to extremely strong 2300+ players, so no disrespect at all to what is a very strong high Expert to low Candidate master machine) and checking the crucial moves where the tide of the game suddenly turned.

I then checked these critical positions (about 6 in all) against Fritz 3. I found 100% correlation between Fritz 3 and Atlanta in terms of responses. I also tested various positions throughout these games and again they precisely match what Fritz 3 plays once out of the opening.

So now I am happy in my own mind there seems to be practically no changes at all - PC versus dedicated - that effect choice of moves between Fritz 3 and Atlanta, though of course there might be ocassions they make different moves, but this would be more due to hardware differences and differing hash table sizes rather than the program itself I think.

As for Milano Pro, I do have to say that on the basis of the analysis of that machine I have performed over this last week, I am struggling to find any significant correlation not only between it and Fritz 3, but by inference it and Atlanta.

I know it has been stated previously that Milano Pro / Master Chess is an Atlanta without hash tables, but on the basis of my own evidence I would have to disagree with this. There are just too many move deviations between Fritz 3 and Milano Pro, and moves that do match could simply be moves that many good programs would play in those same positions.

It is also clear that Atlanta has more chess knowledge - at the very least in the endgame - compared to Milano Pro. In the above position we have been testing, hash tables are not going to save the day on their own - they are just gong to make the machine reach a stable evaluation faster. Milano Pro clearly has no clue about how to play that position (unless perhaps given an amount of thinking time far beyond what would be available under tournament conditions), whereas Atlanta and Fritz 3 clearly do (incidentally, Fritz 1 also has no clue how to play that position either).

It is quite possible that perhaps Milano Pro is a machine based on an early development version of Fritz 3 before algorythms were tuned or modified and knowledge added, but as far as I am concerned, nothiwthstanding it bears a rather tenuous and dubious relationship to Fritz 3 or Atlanta in terms of the way it actually plays chess.

That is not to say it is a bad machine, since it currently has an SSDF rating just shy of 2000. But I have to say I am a bit disappointed looking at it's games and am left wondering why Saitek did not make the Atlanta their top end machine and ditch the Milano Pro program instead. I mean, how much more would it have cost to do this? Another $20 - $50? Surely anyone serious enough about buying a strong, premium machine like this would not care about spending more money to get the better program that at the very least plays a significantly stronger endgame.

[Monsieur Plastique]

The mephisto master chess has a weak endgame but it is very stong on middlegame, it's tactic is what gives it a better elo than citrine.

Interesting. Thanks for that. This is exactly what I find with the Fritz program for Nintendo DS too. It's attacking play and relatively strong tactical ability at the high levels are what enables it to scrape through with a 2000+ rating. But if it comes down to an endgame, it is fairly similar in strength to something like the Mephisto MMV or Expert Travel.

Interestingly though, the Fritz for Nintendo DS is extremely weak at fast chess, probably because these versions of Fritz from 9 onwards had much more chess knowledge and the node count was significantly reduced compared to earlier versions.

One has to now wonder: given that one can get a very mature version of Fritz for the Nintendo console, I am wondering what the attraction is in relation to the Mephisto Master Chess (unless you are a collector or want specific time controls that work correctly, etc). I had thought Master Chess might save the day with a better endgame, but it is seemingly no better than the one on the Nintendo.

I would also speculate that is an perhaps another reason why I believe we will never ever see a new, strong chess program from Saitek. The programs have now just gotten too larger and too slow to be used in a dedicaed machine where the manufacturer is trying to keep costs down.

[Steve B]

as to the position
i walked away from the computer after setting the position up and i returned about a half hour later
the MM5 was showing Re4+ and an eval of -.24
easy enough id say
probably a good idea to start a bit lower down on the Schroeder food chain
lets try this with the Rebel 5.0...

Speaking of the Rebel 5.0(1825 Elo)...

here is a nice game it played against the Conchess T6 Amsterdam ( 1802 Elo )
T6 had the L16 book module locked and loaded
Time Control 1 Minute per Move Avg..
the T6 was in book until move 8
End game play by the Rebel looked pretty good ...

[Date "2010.3.7"]
[White "T6 Amsterdam"]
[Black "Rebel 5.0"]
[TimeControl "1 min /Avg."]
[Result "0-1*"]

1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Nc3 Nf6 5.d3 d6 6.Bg5 h6 7.Bxf6 Qxf6 8.O-O
Be6 9.Bxe6 Qxe6 10.Na4 O-O-O 11.Nxc5 dxc5 12.Qd2 c4 13.b4 cxd3 14.cxd3
Nd4 15.Nxd4 Rxd4 16.Rfd1 Rhd8 17.Qc3 Qd6 18.b5 Rxe4 19.Qb3 Rf4 20.g3 Rf3
21.Rac1 Qd5 22.Qc3 Rd7 23.Qa3 Qxb5 24.Qxa7 Qa6 25.Qxa6 bxa6 26.Rc6 e4
27.d4 Rd3 28.Re1 Kb7 29.Rc4 R3xd4 30.Rxe4 c5 31.Kf1 Kc6 32.Rexd4 Rxd4
33.Rxd4 cxd4 34.Ke2 Kd5 35.Kd3


[fen]8/5pp1/p6p/3k4/3p4/3K2P1/P4P1P/8 b - - 0 35[/fen]

up a pawn.. Rebel knows the concept of "King Opposition" like an Endgame Master
Capablanca Himself would be proud

35..g6 36.f4 f6 37.g4 a5 38.h4 a4 39.a3 g5
40.fxg5 fxg5 41.hxg5 hxg5 42.Kd2 Ke4 43.Ke2 Kf4 44.Kd3 Kxg4 45.Ke4 Kh3
46.Kxd4 g4 47.Kc5 0-1


just to show that the T6 is no slouch..
here is a game where it eats the Novag Jade II for lunch and with the black pieces no less..
the JII is rated more then 100 points higher then the T6..tipping the scales at 1953
in this game the T6 was in book until move 12...

Time Control 30 Sec Avg..
[Date "2010.3.6"]
[White "Jade II"]
[Black "T6 Amsterdam"]
[TimeControl "30 Sec.Avg."]
[Result "0-1"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Nxe4 6.d4 b5 7.Bb3 d5 8.dxe5 Be6 9.c3 Bc5 10.Nbd2 O-O 11.Bc2 Bf5 12.Nb3 Bg6 13.Nfd4 Nxd4 14.cxd4 Bb6 15.f3 Ng5 16.Bxg6 hxg6 17.Be3 Ne6 18.Qd3 Qh4 19.Rac1 Rae8 20.f4 Re7 21.Nc5 Bxc5 22.g3 Qh3 23.dxc5 d4 24.Bd2 Rc8 25.Qa3 Ra8 26.c6 Ree8 27.Bb4 Qf5 28.Bd2 g5 29.fxg5 Qxe5 30.h4 Qe2 31.Rf2 Qe4 32.Qf3 Qxf3 33.Rxf3 Ra7 34.Re1 g6 35.Kg2 Raa8 36.g4 Nc5 37.h5 Rxe1 38.Bxe1 Ne6 39.Bd2 Rb8 40.Rf6 Nf8 41.Ba5 Rc8 42.Rf3 Ne6 43.Bd2 gxh5 44.gxh5 Rd8 45.Ra3 Rd5 46.Rxa6 Nxg5 47.Ra8+ Kh7 48.Re8 Rc5 49.Rd8 Ne6 50.Rd7 Kg8 51.Ba5 Rxc6 52.h6 Rc2+ 53.Kf3 Rxb2 54.Bxc7 Rxa2 55.Bd6 Ra1 56.Bb4 Rh1 57.Bd2 Rh2 58.Bc1 Rh3+ 59.Kg4 Rh1 60.Bd2 Rd1 61.Bb4 Rb1 62.Bd2 b4 63.Kf5 b3 *

FINAL POSITION


[fen]6k1/3R1p2/4n2P/5K2/3p4/1p6/3B4/1r6 w - - 0 64[/fen]

i Resigned for the Jade II here.
its hopeless

Rebel Yell Regards
Steve

[Steve B]

Of Course Rebel's endgame play can have the occasional bad hair day

this position was reached in another game vs.the T6 with Rebel in charge of the white pieces(TC -1Min Avg.)

[fen]6n1/8/4P3/2pK3P/1k6/8/4B3/8 w - - 0 58[/fen]
Rebel up a pawn has a clear win with Kd6
instead it manages to play one of the only moves that draws
h6??

Rebel was showing a +2.26 Score because it loves the idea of under promoting in the next few moves.
play ensued..
58..Nxh6 59.e7 Ng8 60.e8=N

[fen]4N1n1/8/8/2pK4/1k6/8/4B3/8 w - - 0 60[/fen]
promoting to a Q would have given black 60..Nf6+ forking the Q and K
this made the Rebel very happy ...sporting a + 2.68 score
the only small problem of course is that the game is now a hopeless draw

Rebel Without a Cause Regards
Steve

P.S.
FULL PGN
[Date "2010.3.7"]
[White "Rebel 5.0"]
[Black "T6 Amsterdam"]
[TimeControl "1 Min. Avg"]
[Result "1/2-1/2"]

1.c4 c5 2.Nc3 Nf6 3.Nf3 Nc6 4.g3 g6 5.Bg2
Bg7 6.O-O e5 7.d3 O-O 8.Bg5 d6 9.Qb3 Rb8 10.Nd5 Bg4 11.Rac1 Bxf3
12.Nxf6+ Bxf6 13.Bxf6 Nd4 14.Bxd8 Nxb3 15.Bxf3 Nxc1 16.Bc7
Nxa2 17.Bxb8 Rxb8 18.Ra1 Nb4 19.Rxa7 b5 20.b3 bxc4 21.bxc4 Kg7 22.Rd7
Rb6 23.Kg2 Kf6 24.h4 h5 25.Ra7 Ra6 26.Rxa6 Nxa6 27.Be4 Nb4 28.f4 exf4
29.gxf4 Nc2 30.Bb7 Ne3+ 31.Kh3 Ke7 32.Bf3 Ke6 33.Bc6 Ke7 34.Be4 f5
35.Bc6 Kd8 36.Kg3 Ng4 37.Kf3 Nf6 38.Ba4 Kc7 39.e3 Kb6 40.e4 fxe4+
41.dxe4 Ka5 42.Bb5 Ng8 43.Be8 Ne7 44.Ke2 Kb4 45.Kd3 Kb3 46.Bf7 Kb2 47.e5
dxe5 48.fxe5 Kb3 49.Be6 Nc6 50.Ke4 Nd4 51.Bf7 Nf5 52.Bxg6 Ne7 53.Bxh5
Kxc4 54.Be2+ Kb4 55.h5 Ng8 56.Kd5 Nh6 57.e6 Ng8 58.h6 Nxh6 59.e7 Ng8
60.e8=N c4 61.Nd6 c3 62.Bd1 c2 63.Bxc2 Ne7+ 1/2-1/2

[Monsieur Plastique]

i Resigned for the Jade II here.
its hopeless

This particular Jade II is simply out of practice and lonely. She will play much better after a month at my place and with Hilary as a new playmate.

AA Mignon Regards.

[Steve B]

i Resigned for the Jade II here.
its hopeless

This particular Jade II is simply out of practice and lonely. She will play much better after a month at my place and with Hilary as a new playmate.

AA Mignon Regards.

the above game might still be in the Jade II's Memory when it arrives in the Land Of OZ
oh no ..wait
thats only if the batteries are left in it

Acid Trip Regards
Steve

[Monsieur Plastique]

It just ocurred to me that given how exceptionally thorough Australian customs are, I expect they will open the package under the suspicion that the corrosion in the battery compartment are drugs of some sort!

All I can hope is that the poor little Beagles don't try to eat it or lick it

Maybe you should put a warning on the box: Not fit for animal consumption, let alone human!

[Monsieur Plastique]

The Mephisto Expert has several levels Elo of strength as the PC Fritz 1 and simulates well the human game players of Club.

Now standard with 32 bits is Fritz 5.32 but Fritz is too strong for club players, therefore having a machine like the Saitek President, GK-2100, Mephisto Expert .... All machines of Frans Morsch 32K are perfect for club and players are better focused than the Mephisto Milano Pro / Master for example that are based on Fritz3

Be that as it may, I just purchased a Saitek Expert Travel machine last week and I have to say I am incredibly disappointed with this machine. Maybe my expectations have changed over time having used PC engines as sparring partners and for analysis, but thus far it has displayed nothing but pathetically incomprehensible positional play, weak tactical ability and a propensity for incomprehensible blunders.

Yes, I am aware that there is a "fun level" bug with this machine, but I was careful to reset the machine using the ACL button on the underside of the machine and to not even so much as accidentally "scroll through" any of the fun levels. So I would reset the machine, press the level button then press the C5 square to go straight to the 40 moves in 2 hours level.

Here is a game it played against an unnamed opponent where it was sytematically crushed barely out of the opening and subsequently paralysed - just like I would expect to see if Kasparov was playing a simul against a 1600 player:

[Event "40 moves in 2 hours"]
[Site "?"]
[Date "2010.03.11"]
[Round "?"]
[White "Saitek Expert Travel level C5"]
[Black "Average Player"]
[Result "0-1"]
[ECO "C40"]
[PlyCount "90"]

1. e4 e5 2. Nf3 f5 3. Nxe5 Qf6 4. d4 d6 5. Nc4 fxe4 6. Ne3 Nc6 7. Bb5 a6 8. Ba4
b5 9. d5 Ne5 10. Bb3 Qg6 11. a4 b4 12. Bd2 a5 13. Qc1 Ba6 14. Ba2 Nf6 15. Bb3
Nh5 16. g3 Qf6 17. Bxb4 Qf3 18. Kd2 Qxf2+ 19. Kc3 Qf6 20. Kd2 Nf3+ 21. Kd1 Be2+
22. Kxe2 Nd4+ 23. Ke1 Qf3 24. Bc4 Qxh1+ 25. Nf1 Nf3+ 26. Kd1 axb4 27. Bb5+ Kd8
28. c3 b3 29. Be2 Be7 30. Qe3 Nxh2 31. Nbd2 Nxf1 32. Nxf1 Rf8 33. Bc4 Nxg3 34.
Qxg3 Bh4 35. Qxh4+ Qxh4 36. Ra3 Qg4+ 37. Ke1 Qg1 38. Rxb3 e3 39. Kd1 Rxf1+ 40.
Bxf1 Qxf1+ 41. Kc2 e2 42. Rb8+ Rxb8 43. b4 e1=Q 44. Kb3 Qb1+ 45. Ka3 Qfc1# 0-1


And here is an unfathomable display of one of the most horrific blunders I have ever seen in computer chess:


[fen]4k2B/5p1p/1b1p2p1/3q4/8/2R3P1/P1R1PP1P/6K1 w - - 0 34[/fen]


In the above position, Saitek Expert Travel on level C5 (40 moves in 2 hours) is white and it is white's turn to move. After several minutes thought, it comes up with the totally inexplicable 34. Bf6??, losing instantly. I stared at this position for a long time trying to understand what on earth motivated the computer to play this terrible move. I came to the conclusion that it is either a horrible flaw in the selective search or that it is a bug.

So as I say, I am not impressed at all thus far with a machine that is rated at over 2000 ELO on the Selective Search rating list. It's not even impressive for it's 1879 rating on the SSDF list. The "fun level" bug is extremely annoying, because if the user so much as even "scrolls through" a fun level without actually playing the machine at that level, then it still plays in fun mode. So if the owner ever accidentally allows "fun" to appear in the display at all, they have to reset the machine using the ACL button or remove and reinsert the batteries. Inexcusable for a machine that has been around in various guises for a over 15 years now.

It has me wondering if my unit is faulty. As I say, I made sure it was not the fun level bug at work by clearing the machine with the ACL button, but honestly....Bf6?? That is the worst blunder I have seen from any computer rated over 1600, let alone one that is supposed to have an Expert Rating. 16 years ago this very same program scored an Action Chess USCF rating of over 2300. Unbelievable.

Please tell me that Master Chess does not play this horrible move. Please!

Incidentally, Fritz 1 does not play this losing move, nor does it even consider it at any point during it's search. After seeing this pathetic display of chess playing from the Expert Travel, there would have to be a very significant step up in quality of play for me to consider adding a Master Chess to my collection.

[Steve B]

Please tell me that Master Chess does not play this horrible move. Please!

It Dosen't
on 40/2 level ..The Master considers first a4 but after a few seconds it switches to Rc4 which it plays after about 2 minutes sporting a hefty +1.7

i let it think for 30 minutes and it never considers Bf6

2102 Elo Regards
Steve

[IA]

Hello Monsieur Plastique, we should not attach too much importance to these games, these machines have performed very human and sometimes do play we see when we face any human, such errors and even worse I've seen machines like the Mephisto Master not I hate this machine and has a Bug, Novag Sapphire II, Saitek Cougar than a Tower gift for a Bug, Mephisto Milano ... .. starting at the end almost all machines make plays very loose, two machines are very strong in tournament games, one is the Mephisto Advanced Travel Chess / Miami of only 16K and has an Elo rating of 2019 Elo WIKI Turnier , while The Mephisto Expert is a Elo WIKI Turnier 2006 Elo this figure is spectacular for Mephisto Advanced , another very solid machine that just makes mistakes is important Mephisto Berlin 68000 has a very balanced game and an Elo rating of 2150 Elo WIKI Tournier ....

The Mephisto Expert is a game machine with a very human and you have the impression at any time be confronting a chess computer for his amazing moves, despite these annoying errors that have almost all the machines remains an excellent choice for players of Club,Besides having a tactical trainer is Self-Adaptive Weak Levels (Squares H1,H2....8 ) where you can select the playing strength from the lowest 1600 ELO to 2000 ELO H1 to H8 level, this is the same as there are levels of play in the Palm Hiarcs, the opening book is very refined and adjusted to render the machine as possible without getting into complications, tactical game is great but even better positional play is incredibly, this machine is very special and should not be compared with computer programs nor with PC and PDA , the behavior of this machine is like a human, the Mephisto Expert has spent several test and has not been detected as a machine as it plays like humans Players Club, is undetectable as machine and anyone can see everywhere, That to me is more than enough to choose it as a training machine, otherwise I would have bought the Mephisto Master but seeing play seems like a very boring machine despite its strength, so congratulations, you know the magnificent machine that has purchased despite the small errors that has passed I too and so on are not only all that happens to you counted, that does not happen with Mephisto Mephisto but has an inferior game, I am very happy with this great machine despite having some problems and if I had to do without my PDA Hiarcs would play only with the Mephisto Expert.

Do not forget that one of the leading specialists in the world chess machines Eric Hallsworth of Selective Search Magazine recommends it as the best portable machine today:

http://www.elhchess.demon.co.uk/expert.htm

Regards...

[Steve B]

Please tell me that Master Chess does not play this horrible move. Please!

It Dosen't
on 40/2 level ..The Master considers first a4 but after a few seconds it switches to Rc4 which it plays after about 2 minutes sporting a hefty +1.7

i let it think for 30 minutes and it never considers Bf6

2102 Elo Regards
Steve

It Might also interest you to learn that the Saitek Travel Champion 2100 at 40/2 will play a4 clocking in at +1.5
While the Cosmos plays ..instantly....Bf6
as it does on an all but the Adaptive levels where it plays e4
this is highly Symptomatic of the dreaded bug
it must be a new strain of the H8 virus

H8N1 Regards
Steve

[Monsieur Plastique]

Hi Steve,

Yes, when I replayed the position it kept playing Bf6 instantly. I was wondering if somehow the machine was tricked into thinking it was in it's openings book (but it did not play the move instantly in the actual game itself!). I had not even thought of that H8 bug, but now that the wiki is closed I can't use it as a reference to refresh my memory on what that bug is all about. But at least the Master does not play it (of course it is a different type of processor).

I guess I will have to seriously think about getting the Master Chess now, before they disappear off the face of the earth. Even though it "breaks the rule" I have given that it is not strictly portable, it seems small enough to sit on the bed, presumably it is fairly lightweight and it does have bright LEDs on the board edges. Afterall, if Alexandra The Not-So-Great can work fine on my bed, I don't think the Mephisto Master Chess is much bigger. Plus there is the attraction of all that black, textured plastique...


Hi IA,

I have seen a lot of Cosmos games and they don't really strike me as very human like, unlike games played by the Novags which to me seem very human like (even including the little Star Opal). I guess we all see different things in the machines we like - personally I think the Fritz on the Nintendo DS is far and away more human like than Expert Travel and much more fun to play against, even though both machines are probably around the same playing strength.

What I don't like about the Expert Travel machine is that it does not seem to move purposefully or with a "pseudo" plan. It just seems to look at the position, decide if there is a material winning tactic involved - and if not then it seems to make any move that gives a good evaluation, even if it is a positional disaster. For example, it does not seem to care much about tempi, good piece development, mobility, blocked pieces or weak pawn structures at all. Fritz on the DS, though possibly weaker tactically, really seems to play with a sense of purpose seems to coordinate it's forces far better than this Expert Travel machine.

For some reason, I prefer the playing style of the more humble 16K program found in the Miami, though I do have to acknowledge that Expert Travel is capable of finding tactics that elude the Maimi and does have a stronger endgame.

But you do make some good points, and I think with all the different level combinations, it will make a good bedside machine for casual games with it's decent sized board.