More horrors from the dedicated chess computer blooper reel

[JonP01]

This position occured in a quick game (approx 30 seconds per move) between a Fidelity Excellence (white) and Mephisto Miami.

[D] 6k1/1b3p2/3p3p/1Pp3p1/2P1P3/8/5KP1/5B2 b - - 0 0

As you can see, Morsch's baby has done it's thing in the middle game and has a slight material advantage (and the opening was a gambit that left black a pawn down by the time it left's it's openings library, so black has clearly done very well).

In the above position, black unleashes 1...Bxe4?!

The reason I put the ?! there is because if black wants the pawn, black really has to have forseen the consequences of white wanting to get his bishop onto the A8 - H1 diagonal. With white's dangerous passed "b" pawn, black has to get his king into play the second white signifies he wants that diagonal.

So white, being a Fidelity machine and strong in the endgame replies 2. Be2.

Now it's been a while since I had my Mephisto Miami out to play, but I still thought I was familiar with it's strengths and weaknesses. Still, not even I was prepared for black's horrific reply:

2...d5??

The ONLY move for black at this point was Kf8. Now black is powerless to stop white queening the b pawn.

Thinking this gross error might have had something to do with the 30 second thinking time, I tried it on 40 moves in 2 hours. It still played 2...d5??

I guess this is a perfect example of why the original Travel Master and it's descendants have been criticised for playing a very uneven game. Niether of my other top machines (Amber or Chessmaster Gameboy) make this error. Whilst Novag Amber might take the e pawn, it finds the correct Kf8 for the next move. Chessmaster Gameboy does not even take the e pawn at all, and instead brings it's King to f8 immediately.

Perhaps this ubiqitous Morsch program (which has by now found it's way into countless iterations of dedicated machines) ought to be marketed as a middle game tactical analyser that automatically switches off once an endgame is reached! There is no doubting the program's tactical prowess and aggressive play, but it is probably the most unbalanced program of the modern era.

I don't have a Saitek Cosmos or Saitek Chess challenger, but please tell me that program does not make this same elementary mistake!

House of Horrors Regards

Jonathan

[Larry]

Hi Jon,
snip<Travel Master and it's descendants>snip

I don't have the Miami, but I've got a Mephisto Modena coming
from Germany next week. By the rating, I'd say it's also one of
the Travel Master descendants, so when it gets here I'll set this
position up (I've got an Excellence) and see if it makes the same
shallow move. I've actually got a TravelMaster but it's kaput, like
most of them.
The comps of the TravelMaster era were notoriously uneven in
their strength from one phase of the game to the next. The
Super Conny comes to mind.
all the best
Larry

[JonP01]

Hi there Larry,

Well I can hardly complain though. I just bought a spare "Miami" last week (nowadays called the Advanced Travel Chess). At a wallet crushing $44 US, that is still amazing ELO for the dollar. I am sure your Modena will not make the blunder. I still can't beat this Miami critter myself because I am usually a pawn (or even two) down in the endgame with a relatively poor position on account of my failed attempts to defend myself from it's nasty tactical onslaughts.

That said, a (probably) slightly lower ELO machine like my Chessmaster Gameboy is actuallly a tougher opponent for me as a human, because it doesn't have the weaknesses Miami has (but it doesn't have the same tactical grunt either). For some reason when I play Miami it feels very computer-like, but my Chessmaster just feels like a strong human.

Regards

Jonathan

[Robert Weck]

Hi Jonathan,

In the above position, black unleashes 1...Bxe4?!

The reason I put the ?! there is because if black wants the pawn, black really has to have forseen the consequences of white wanting to get his bishop onto the A8 - H1 diagonal. With white's dangerous passed "b" pawn, black has to get his king into play the second white signifies he wants that diagonal.

I think, you are much too harsh to our old friends!

Even modern PC-engines need 12-14 halfmoves, to discard Bxe4!

And concerning 2.-d5??: it needs 10 halfmoves, to see, that white gets a queen; how deep gets the Miami with 3mins/move in this position?

And it is well known, that Morsch programs have nearly no endgame capabilities (only modern Novags are even worse!)

here you can see a module, that contains all endgame routines, that Morsch-programs have:

http://cgi.ebay.de/Leer-Einschubmodul-f ... dZViewItem

So white, being a Fidelity machine and strong in the endgame replies 2. Be2.

would the Excellence avoid 1.- Bxe4 and 2.- d5?
And if it did; would it know, why?

I am not sure about that...


best regards,
Robert

[JonP01]

Hi Robert,

I never said that Bxe4 was a bad move as such. I consider it to be one of many viable options for black. The reason for my annotation though (as I explained) was that if Bxe4 is played, black needs to have an understanding of the consequences of the passed "b" pawn and the white bishop moving onto the a8-h1 diagonal within two moves. Clearly in this case, when Miami played Bxe4, it did not understand the ramifications of what it was doing when it played Bxe4.

So to reiterate, I have no problem with a machine or human playing Bxe4, so long as they don't follow that up with a blunder related to that particular move. In this case, Miami blundered in direct consequence of having taken the pawn though.

As I also pointed out, niether my Amber nor Chessmaster Gameboy blunder in the way that Miami did. I also tested the position on my Excellence, since you wondered if it would make the same blunder that Miami did. I can tell you that it didn't. It played the same continuation that Novag Amber did (i.e moved it's King to f8 rather than the pawn to d5). And it came up with the correct King move after about 90 seconds, at 7 ply, after initially rejecting the pawn advances. Whether Excellence "understood" why it rejected the pawn advances, well I feel justified in saying that it knew why. It did reject them before finding Kf8. I think that is reasonable evidence it could see the potential problems that lay ahead, since, unlike Miami, it continued to play accurately from that point onwards.

So in being considered harsh, I would have to question that as well, since I have had no trouble finding other machines in my collection with equal or even much lower ELO ratings and which have a significantly superior understanding of endgame play to Miami. Of course, Morsch programs are known for their redeeming strengths in terms of tactical prowess, but for me I prefer the more balanced game of my Fidelity Excellence and Chessmaster Gameboy. But then again, we have both said in this thread how bad Morsch endgames are, so

We agree on that point Regards

Jonathan