M-Chess Pro 3.5 - A Review
John Quill Taylor
by John Quill Taylor - jqta...@hpdmd48.boi.hp.com
To: rec.games.chess
M-Chess, P.O. Box 9388, San Rafael, California 94912
While it is fresh in my mind, I have decided to write down my thoughts
regarding my purchase of M-Chess Pro 3.5. Initially, I thought I might
do this and post it to the rec.games.chess group on the Internet, but I
have decided in addition to this to send this to the creators of
M-Chess. Since I spent $149.95 for this software, I expect a lot out of
it. Besides using it as a chess partner, I intend to use M-Chess
specifically for correspondence games in which my opponent and I have
agreed that we may use any advice or machines available. I will attempt
to use the program on three different machines (if the copy protection
"feature" will let me; I have not tried beyond the first machine yet).
The first machine is at home, a 486 DX2 66 MHz machine with 16 Megabytes
of extended memory. The second machine is a portable notebook computer,
a 486 25 MHz machine that does not support enhanced mode. The third
machine is at my office, a 486 60 MHz Pentium, which is typically 10-12%
faster than the 486 66 MHz machine.
I am a novice chess player, but I love the game very much. Please don't
misinterpret this analysis as gospel, as I certainly don't have the
expertise to claim that it is even correct. M-Chess, as well as many
other programs, are very well designed. I expect M-Chess is the best
chess software I own. I intend to use it frequently. I have gone on
record saying that the best human players will beat the best machines in
a match for many years to come. I have even bet money on it! This is
not because I don't appreciate the programs - they are very strong,
stronger than I am by far. I have a strong belief that the human
ability is often underestimated in chess. I think building a chess
program that is unbeatable is a very admirable and ambitious goal, and I
will welcome such a human triumph, if and when it occurs. I doubt that
will be in my lifetime, and I expect to make it to the end of the
decade!
M-Chess was ordered from ICD at toll-free 1-800-645-4710 on 15 April,
and arrived under my doormat on 22 April. Cost was 149.95 plus
shipping. Installation was uneventful and went smoothly. The
copy-protection schema looks a bit intimidating but I found it to be
"manageable" so far. So far means one copy on one machine, so we shall
see. I believe it is my right to use the program at home, at work, and
on my portable machine, all without having to disable one copy to run
another. The copy-protection as described in the documentation leads me
to believe that I cannot do this. Certainly the developers deserve to
be compensated for their efforts, but copy-protection is always a
negative thing. Surely chess minds can find a better way...
The first thing I noticed upon running M-Chess is that the mouse scrolls
menus both vertically AND horizontally. At first this was a bit
difficult to get used to, and I'm still not sure I like it. It is not
inferior to any other chess software I have used, but it has many
"features" that are a bit overwhelming. A "beginner" mode ought to
display just the basic minimum menus necessary to play. As it is, it
takes a while to just play a game at a specific time limit. It took me
about five minutes to get started with a "Game in 10" battle, which I
lasted close to 40 moves before being beat. I expected it to trounce me
more soundly, but maybe I was in "rare form."
The opening played is displayed at the bottom of the board, for the
WHOLE GAME! This seemed a bit odd when I was at move 35 and it was
still displaying "Reti Opening" on the screen.
My first confusion was the strange four arrows to the right of the
board. As it turns out, the top arrow toggles the displaying of the
score to the analysis. Simple enough. The next two arrows are for
moving forward and backward (take back). Both work well. The bottom
arrow tells the machine to begin searching using the predefined time,
or, if already searching, it forces the machine to make the current
move. It works very well also, allowing the machine to play either side
immediately. One problem I have had with this and other programs is
with infinite (or all-night) searches. If the arrow is pressed by
mistake, the move gets made and hours of searching comes to an abrupt
end, possibly by mistake. A very nice feature I've never seen would
store the hours of search data and allow me to make a move, or even exit
the program and turn of the computer, then go back to THAT SAME POINT in
the analysis that took all night or however long. Is this impossible to
do? Maybe I am missing something.
A unique (to me) feature of M-Chess is the display of the total number
of possible moves as the denominator of a fraction in which the
numerator is the current move under consideration (I think) at the
current depth. Well, even if I don't know exactly what the numbers are,
they offer relevant feedback information regarding just where the
machine is working, and it's a nice touch.
The clock seems to count down in some modes and up in others, and to be
honest, I find the clocks in most chess programs to be more complicated
than the chess itself! I guess the obsession with time is indicative of
the age we live in. M-Chess seems to have every conceivable option of
clock for any chess game. One thing I did have to get used to is that
after one hour the seconds go away and one hour can look the same as one
minute.
A worse problem occurs with the positions indicator, and it's one I've
seen over and over. On some shareware programs I've even poked into the
code to try to "fix" this. The problem is very basic: there are no
commas in the displayed number! So if I have searched maybe 8-ply the
display might read something like 3452765 and at a glance I can never
tell if it's three million or thirty-four million positions. Simply
adding commas would be so much nicer: 3,452,765 is a lot easier to read
at a glance.
A minor complaint is that when playing a game, the tail end of the
previous game's text is still displayed unless you click the
move/analysis button twice to "erase" it.
M-Chess says that it is "planning" one move and "considering" another
move, which is normal as it tries the various moves available, but the
main line shown off to the side frequently shows a line different than
that planned or considered! The main line should always be the
"planned" move, which ultimately does happen. Why the lag time on a
move that is no longer being planned? This is human behavior, not
machine behavior!
As far as I can tell, M-Chess will not resign or offer me a draw
(assuming I could win, which is probably a bad assumption). Why have
every clock under the sun but not include this very basic part of the
game? I've seen it in other programs.
There is no short algebraic mode. This is sad, because it is the format
most players use. It looks cumbersome to see e2-e4 instead of the
simplified e4.
There are a few clever (and a few not-so-clever) remarks displayed by
M-Chess Pro 3.5. One of the clever ones occurs after you are mated and
you accidentally try to continue, and it displays: "The Game's History!"
which is I guess cute. One of the not-so-clever remarks is when you are
done playing and click on "Quit" and it asks: "You Quit?" I find this
intimidating, and I'm a pretty easy-going D player. I am sure it will
offend some players. Maybe "Exit" would have been a better verb here,
since "You Quit?" implies that, well, you are a quitter!
The program overall is impressive, and so far it's never "latched up" my
machine, which is probably a breakthrough in itself for chess programs.
I have had it sort-of "lock up," however, on two occasions, in the
autoplay mode. In both cases, all the data fields went blank (no values
displayed), and the program seemed to halt. But upon clicking a few
buttons it would resume.
My first test of M-Chess was an endgame that I've used to check other
chess programs. This test was a major disappointment, and I have no
explanation for why such a strong machine would do so poorly. The
position is as follows to start: White - Kd2, Rf2, h5 ; Black - Kb1, a2,
a6, g7, h7. White is to move and win. I compared ChessMaster 2100,
Psion 2.13, and M-Chess Pro 3.5. In all cases, look-ahead was disabled.
The chart below shows the test with the "best moves" and the time it
takes the three programs to find these moves. In all cases in this
paper, I used a 66 MHz DX2 486 with 16 Mbytes of RAM and the above three
versions and brands of chess software.
- - - - - - - -
- - - - - - p p
p - - - - - - -
- - - - - - - P
- - - - - - - -
- - - - - - - -
p - - K - R - -
- k - - - - - -
White to move and win.
seconds to find the moves for three chess programs
==================================================
move Psion ChessMaster M-Chess
1. h6! 2 0 0
gxh6 0 0 0
2. Kc3 0 0 0
a1N 0 0 0
3. Rb2+ 1 0 0
Kc1 0 0 0
4. Ra2 0 8 0
Kb1 0 0 0
5. Rxa6 0 0 0
Nc2 4 0 0
6. Re6! 2 6 30 Why so long?
h5 0 0 0
7. Re2 0 0 0
Na1 0 0 0
8. Rh2 0 0 insists on Rb2+?
h4 0 0 0
9. Rxh4 0 0 0
Nc2 0 0 0
10. Rh1+ 0 0 0
Ka2 0 0 0
11. Kxc2 0 0 0
Ka3 0 0 0
12. Rh4! 2 0 0
h5 0 0 0
13. Rf4 0 0 0
h4 0 0 0
14. Rxh4 0 0 0
Ka2 0 0 0
15. Ra4# 0 0 0
In the "6 second per move" autoplay mode, Psion 2.13 finds every move
above, yet it takes M-Chess about 30 seconds to find the key move, Re6!
ChessMaster gets through this exercise very well also, only taking 6
seconds for the key move, and about 8 seconds for finding Ra2. Only
M-Chess insists upon inserting the needless move 8. Rb2+? Kc1 thus
delaying the eventual mate by several moves. It inserted this move even
for searches up to several minutes!
Does all this mean that at fifteen seconds per move, Psion, a public
domain program, can outplay M-Chess, a program costing over a hundred
dollars. In this position, perhaps. Here is M-Chess on autoplay with
no thinking on the "opponent's" time, set to play at 10 seconds per
move. I logged the entire process with the "save file" command, a very
nice feature of M-Chess!
Initial Position
White King: d2
White Rook: f2
White Pawn: h5
Black King: b1
Black Pawns: a6 a2 g7 h7
WHITE BLACK
M-Chess
10 seconds per move
1. h5-h6 g7xh6
2. Kd2-c3 a2-a1N
3. Rf2-b2+ Kb1-c1
4. Rb2-a2 Kc1-b1
5. Ra2xa6 Na1-c2
6. Kc3-d2 Nc2-d4
7. Ra6xh6 Nd4-f3+
8. Kd2-c3 Nf3-g5
9. Rh6-h1+ Kb1-a2
10. Rh1-h5 Ng5-f3
11. Rh5xh7 Ka2-b1
12. Rh7-h5 Kb1-a2
13. Rh5-a5+ Ka2-b1
14. Ra5-f5 Nf3-h4
15. Rf5-f1+ Kb1-a2
16. Rf1-f2+ Ka2-b1
17. Rf2-f6 Kb1-a2
18. Rf6-a6+ Ka2-b1
19. Ra6-b6+ Kb1-a2
20. Rb6-b5 Nh4-f3
21. Rb5-f5 Nf3-h4
22. Rf5-g5 Nh4-f3
23. Rg5-g8 Nf3-e5
24. Rg8-e8 Ne5-d7
25. Re8-e6 Nd7-c5
26. Re6-e3 Nc5-a4+
27. Kc3-c2 Na4-b6
28. Re3-d3 Nb6-c4
29. Kc2-c3 Nc4-b6
30. Rd3-d6 Nb6-a4+
And at this point I mercifully ended the program's search.
Now I loaded in an entirely different position. It is from the February
1969 Chess Review column by Irving Chernev, and the problem is by Harold
Lommer. The position is: White - Kh7, Rc8, Na5, Bf1, b5, d5, e2, f5 ;
Black - Kb6, Qb4, Bh4, e7. White to move and win.
- - R - - - - -
- - - - p - - K
- k - - - - - -
N P - P - P - -
- q - - - - - b
- - - - - - - -
- - - - P - - -
- - - - - B - -
White to move and win.
Here the brilliant move is 1.Rc6+!! Kxb5 2.e4+ Ka4 3.Rc5 Kxa5 4.Rxb4
Kxb4 5.e5 Kc4 6.Bg2 Kd4 7.f6 exf6 8.e6 f5 9.d6 etc. or 1...Kxa5 2.Ra6+
Kxb5 3.e3+!! eventually pushing the Black King all the way to the
Kingside to be summarily mated. So what does Psion 2.13 do with this?
For over thirty minutes, it was stuck upon 1.Nc4+? where I terminated
the search. For ChessMaster, things went not much better, as it did not
ever find Rc6+!! even after about twenty minutes at completion of eight
ply after searching over 21 million positions. And how long did it take
M-Chess to find the winning move? About eight seconds! If this is what
a hash table does, give me more! If it's merely a better chess engine,
why did this same engine fail miserably in my first position tested?
You be the judge.
What do these two tests show? Not much, but a lot at the same time.
Certainly more testing is in order. It does appear that M-Chess will
find the same "good" move that both Psion and ChessMaster find in the
first example, but it can and will take longer. It also appears that
M-Chess may find moves that neither of the other programs will ever
find. It is very hard to explain why M-Chess inserts an unnecessary
move in the first example that two other programs simply omit. If it is
a better move, someone please tell me why. If my tests are not fair,
please tell me why.
As a last item, I have logged the first game attempted against M-Chess
Pro 3.5 by my 11-year-old son, who has never played chess and doesn't
even like chess. He knows the rules and basic principles, and "Dad" was
whispering a few comments about "development" and "King safety." My son
is probably about USCF 900 or 1000, I would guess, and I was USCF 1300
or so a few years ago when I was rated. He did a few takeback moves. I
am shocked that he lasted so long against this program. I am afraid it
might not be as strong as I was hoping it was.
WHITE BLACK
Danny Taylor M-Chess
10 seconds per move (score is for White)
1. d2-d4 Ng8-f6 + 0.00
2. Ng1-f3 d7-d5 + 0.00
3. Nb1-c3 g7-g6 + 0.00
4. e2-e3 Bc8-f5 + 0.19
5. Bf1-b5+ c7-c6 - 0.18
6. Bb5-a4 b7-b5 - 0.41
7. Ba4-b3 a7-a5 - 0.31
8. a2-a4 b5-b4 - 0.24
9. Nc3-e2 e7-e6 - 0.05
10. Bc1-d2 c6-c5 - 0.11
11. O-O c5-c4 - 0.23
12. Bb3-a2 Nf6-e4 - 0.35
13. Ne2-f4 Nb8-c6 - 0.54
14. Nf3-e1 Bf8-d6 - 1.13
15. f2-f3 Ne4xd2 - 1.09
16. Qd1xd2 g6-g5 - 1.22
17. Nf4-e2 Qd8-c7 - 1.21
18. f3-f4 Rh8-g8 - 1.47
19. Ne1-f3 g5xf4 - 1.75
20. Ne2xf4 Rg8-g4 - 1.83
21. Nf4-h5 Qc7-d8 - 1.83
22. Ra1-e1 Bf5-e4 - 1.97
23. Qd2-f2 Ke8-d7 - 1.92
24. h2-h3 Rg4-g8 - 1.61
25. Ba2-b1 Be4-g6 - 1.79
26. Nh5-f4 Bg6-f5 - 1.07
27. c2-c3 Bf5xb1 - 0.96
28. Re1xb1 Qd8-f6 - 1.08
29. Rb1-a1 Qf6-h6 - 1.15
30. c3xb4 a5xb4 - 1.05
After 30 moves, an unrated kid is still hanging pretty tough!
31. Qf2-h4 Qh6xh4 - 1.53
32. Nf3xh4 Bd6-e7 - 1.52
33. Nh4-f3 Ra8-a5 - 1.63
34. Nf3-e5+ Nc6xe5 - 2.13
35. d4xe5 Rg8-g5 - 2.20
36. Rf1-f3 Rg5xe5 - 2.52
37. Rf3-g3 Re5-e4 - 3.40
38. Rg3-g7 Be7-c5 - 3.87
39. Rg7xf7+ Kd7-e8 - 5.82
40. Rf7xh7 Bc5xe3+ - 6.55
41. Kg1-h1 Be3xf4 - 6.55
42. Rh7-h8+ Ke8-e7 - 6.96
43. Rh8-h7+ Ke7-d6 - 8.02
44. g2-g4 Re4-e2 - 8.75
45. g4-g5 Re2xb2 -10.81
46. g5-g6 b4-b3 -10.62
47. Rh7-h8 Bf4-e5 -12.99
48. g6-g7 Rb2-h2+ -14.76
49. Kh1-g1 Be5xg7 -14.76
50. Rh8-h7 Bg7xa1 -20.67
51. Kg1xh2 c4-c3 -24.32
52. Rh7-b7 b3-b2 -26.03
53. h3-h4 Ra5xa4 -23.75
54. h4-h5 Ra4-h4+ -25.81
55. Kh2-g2 Rh4xh5 -28.69
56. Rb7-b6+ Kd6-e5 -30.34
57. Rb6-c6 b2-b1Q -35.33
58. Rc6xc3 Qb1-h1+ -99.96
59. Kg2-f2 Rh5-h2+ -99.98
60. Kf2-e3 Qh1-e4+ mate
A final note: I am at my office now, and I am unable to load M-Chess on
my office machine because of the copy protection. I am supposed to
de-install it at home first, but I am running an analysis at home! This
is a major drawback of M-Chess. Should it be illegal to run the same
software in two places, even at different times? What about at the same
time by the same person who paid for it? If anything ever goes wrong
with my hard disk, do I lose the program? Am I not allowed to make one
duplicate copy for backup? Enough questions for now.
Sincerely,
__
John Quill Taylor / /\
Writer at Large / / \
Hewlett-Packard, Storage Systems Division __ /_/ /\ \
Boise, Idaho U.S.A. /_/\ __\ \ \_\ \
e-mail: jqta...@hpdmd48.boi.hp.com \ \ \/ /\\ \ \/ /
Telephone: (208) 396-2328 (MST = GMT - 7) \ \ \/ \\ \ /
Snail Mail: Hewlett-Packard \ \ /\ \\ \ \
11413 Chinden Blvd \ \ \ \ \\ \ \
Boise, Idaho 83714 \ \ \_\/ \ \ \
Mailstop 430 \ \ \ \_\/
\_\/
"When in doubt, do as doubters do." -jqt
Martin Olesen
One more serious drawback of the program if you want to do
analysis with it is that it only displays a singular line,
no branches what so ever. So one is forced to take its analysis
is gospel. I would have liked an ability to inspect its
variation tree.
This is particularly inconvenient if one is using the program to
analyze an opening position.
Martin Olesen
ole...@math.uiowa.edu