Losing Moves

Of course it does make (quite) a difference whether white plays Na1 because
1) he has to, and the knight never returns into play
2) he recaptures a rook, and the knight then returns to b3 or c2
3) it is the start of a knight maneouvre such as Nb3-a1-c2-e3(-f5-h6 mate).
So we might need a database algorithm to check for how long the knight remained on the ugly corner square ... . And it might be interesting to point out some games where Na1 turned out to be a winning move!?

30. Na1 here must be one of the strongest Na1's out there, and of course played in an engine game :-)

http://www.chessgames.com/perl/chessgame?gid=1442234

4.Qxf7 is certainly the "winningest" move a little further down the food chain

I seem to remember Karpov playing Na1 once, and it being hailed as the ultimate in subtle Karpovian maneuvering...

I reckon playing e8=Q (or any other pawn) must be a sign that something is going well, especially before move 40.

Correct me if I'm wrong, but I think you're referring to the famous Nb1! vs Spassky, Candidate's Matches.

Nice stats.

Of absoutely no value to improving anyone's chess ... but fun!

Something like Kg5 (with either color) might be the best candidate for a losing middlegame move. But databases don't help because the same move might be winning in an endgame.

And there are always exceptions to general rules, what about 34.Kg5 1-0 in Short-Timman, Tilburg 1991?
http://www.chessgames.com/perl/chessgame?gid=1124533

I always thought o-o was the worst move. That's why I always leave my king safely in the middle :)

I seem to recall a double exclam, middlegame Na8 by Keres. But I can't remember if he was white or black.

White scores 55% and it is the losingest move. Isn't something wrong here?

Kapalik, the post is confusing (it took me a couple of tries to understand it too). White scores 55% over all games as a whole. He scores 45% in games in which he plays Na1.

wow, what an fun statistic. I wonder what move never (or very rarely) lost (draw or win is ok) in above 2400+ games.

If we are allowed to go with something like gxh8=Q+ I bet that's a winner!

Promotions to Q:
a8Q .72
b8Q .72
c8Q .70
d8Q .73
e8Q .70
f8Q .69
g8Q .73
h8Q .79

gxh8Q+ .80

With Na1 I think you are thinking about an exercise in Jansa and Hort or something. Can't remember the position

I have scored 100% in games where I played Qxg7# :)

I have a couple of suggestions for the winningest moves that don't include #, too: how about R1xf7(or, even better R1xf7+) or Rexb7 and things like that, which imply 2 rooks on the 7th. Also, moves where a knight moves to the 6th rank -like Ncd6- can be good, I imagine.

Also, biship sack like moves probably will do great in 2400+ games - otherwise they wouldn't get played, so things like Bxh7(/g6/f7) should be good candidates.

As for the losingest, I would suggest the moves that imply the replys to the standard sacks, that is moves like gxh3 and Kxh3 and Kxh2.

Two Na1 examples, one plus and one minus:

The "plus" example is a theoretical position from the Dragon Yugoslav Attack, which is discussed in an early chapter of Rowson's The Seven Deadly Chess Sins I think - in a section about how it helps to "talk to your pieces."

I don't recall the specific move sequence or position, but White had castled long. His K was on c1 and N's on b3 and c3, while Black probably had played a B to f5 to reinforce the pressure exerted by a R on c8. For some reason, Nd4 wasn't playable. So, White's best move in the given position was the prophyllactic Na1!, according to Rowson. The N was accomplishing nothing on b3, while defending the c2-pawn with it freed White's c3-knight to advance and contribute to a K-side attack (such as by trading itself for Black's often essential f6-knight).

The "minus" example occurred in a game I recall reading in Chess Life 6 or 7 years ago, Yudasin-Krush, from some Marshall Chess Club event, probably the first incarnation of the New York Masters (back when it was held every week). White was outplayed positionally and eventually lost. The key middlegame moment came when he was virtually forced to retreat a N to a1, where it was as good as trapped.

Come to think of it, something like N8xg6(+) or N8xe6(+) could be great for white, too.

Thomas, I looked up underpromtion to knight too. Often it's done as a joke because whichever piece is promotioned is going to be captured on the next move. The stats are mostly in the low .60s.

this is probably the most boring thread ever here.

Sex! Drugs! Rock n' roll! Kramnik! Toilet! Kirsan eats babies!! Modern chess sucks! Fischer was better!

From a file of Alekhine games (+500) where he won with White in less than 41 moves, you get White 15,862 moves (399 different moves) and the following subtotals (but are they useful?):

Total 15862
Nf3 539
d4 515
Nc3 478
e4 456
O-O 401
c4 361
Re1 273
Ld3 249
Lg5 245
Rd1 244
Ne5 222
f4 219
Lc4 193
Ne4 184
Nd4 174
e3 172
Rc1 171
Qe2 157
Lb5 156
Le3 148
Lf4 145
a4 143
Nd5 143
e5 142
h3 138
cd5 132
Nd2 130
c3 127
g4 127
h4 119
Qf3 116
Qc2 115
a3 113
b4 113
d5 109
ed5 105
Nf5 100
Qd2 97
Le2 93
g3 92
Ng5 92
bc3 90
f3 90
Qh5 90
Ld2 89
Lf6 87
de5 86
Lb3 84
O-O-O 84
Ne2 82
Rb1 80
Qd3 78
b3 68
Ld5 68
Le7 68
f5 67
Lc6 66
Nf6 66
Le4 65
Rf1 65
Qg4 63
Qb3 61
dc5 59
Nd6 59
Qd4 59
Nc4 58
c5 57
Kh1 57
Qe4 57
Ld4 55
Qe3 54
La4 53
Qf4 53
Nb5 52
Nc6 52
Qg3 52
cd4 51
d3 51
Lb2 51
Le5 51
Ng3 51
Lh6 50
Lc2 49
a5 48
Qa4 48
h5 47
Ld6 47
Nf4 47
Lf3 46
Qe5 46
Lg2 45
Nf7 45
Qg5 45
Ne6 44
Qc4 44
Rd7 44
Re3 44
fe5 43
La3 43
Nc5 43
Qh6 43
Lh4 42
Re5 42
Rf7 42
Lc5 41
Qg6 41
e6 40
Lf7 40
Qh4 40
Rb7 40
Lc3 39
Lg3 39
Qc3 39
Re7 39
Rf3 39
d6 38
Lf5 38
Ne3 38
Qd5 38
ab5 37
Lg6 37
Ra7 37
ef6 36
Kh2 36
Ng6 36
Qf5 36
Rd8 36
Rg1 36
g5 35
Rd6 35
Kg2 34
Le6 34
Qe6 34
Rc7 33
Rd4 33
Ne7 32
Nh4 32
Rd5 31
Re4 31
Rf6 31
b5 30
Nb3 30
Qd6 30
de6 29
Ng4 29
Qf6 29
Rd3 29
Kb1 28
Kf1 28
Ld7 28
Qc5 28
Ra1 28
Re6 28
Rc8 27
ef5 26
Kf2 26
Na4 26
Nd7 26
Lb4 25
Qb5 25
Qc6 25
Qc7 25
Qf2 25
Re8 25
Rf5 25
Rg3 25
f6 24
Rf4 24
Rg7 24
ed6 23
Kd2 23
Rg6 23
ed4 22
Lh7 22
Nb6 22
Qf7 22
Qh3 22
Rc6 22
Re2 22
Rf8 22
Rh7 21
gf5 20
Ke2 20
Nh6 20
Qb4 20
Qe1 20
Rh3 20
bc5 19
c6 19
Lg7 19
Qe7 19
Ra8 19
Rc2 19
Rc3 19
Kd3 18
Qb7 18
Ra6 18
dc6 17
hg5 17
Nc7 17
Qa5 17
Rb6 17
Rd2 17
Rh4 17
Nh5 16
Qc1 16
Qh7 16
Rg4 16
Rh1 16
Rh6 16
a6 15
Ke3 15
Lf2 15
Qa3 15
Qa6 15
Rb3 15
Rc5 15
ef4 14
Lh5 14
Ng7 14
Nh3 14
Qd7 14
Qg7 14
Ra5 14
Rb8 14
Rc4 14
ab4 13
g6 13
gf3 13
hg6 13
Kd1 13
Lb7 13
Lc1 13
Lf8 13
Nf1 13
Rg5 13
fe6 12
h6 12
Kg1 12
Lg4 12
Qb2 12
Ra4 12
Rb5 12
Rf2 12
ab6 11
fe3 11
fe4 11
Kc2 11
Kg3 11
La6 11
Na3 11
Na5 11
Qa7 11
Qb6 11
Rh5 11
Rh8 11
ab3 10
Ke4 10
La5 10
Lc7 10
Lf1 10
Lh3 10
Nd3 10
Qd8 10
Qe8 10
Rb4 10
ef7 9
fg6 9
gf6 9
Lb1 9
Nh7 9
Qa8 9
Qb8 9
Qc8 9
Qd1 9
Qg2 9
Ra3 9
b6 8
bc6 8
fg5 8
gh5 8
Kc3 8
Kc4 8
Kd4 8
Lb6 8
Nd1 8
Nf8 8
Qh8 8
Rb2 8
ba5 7
d7 7
fg7 7
hg3 7
hg4 7
Kf3 7
La7 7
La8 7
Nb7 7
Nd8 7
Ne8 7
Nh2 7
Qf8 7
Ra2 7
cb6 6
de4 6
e7 6
fg3 6
gf4 6
gh6 6
La2 6
Nc8 6
Qg8 6
Rg2 6
c7 5
dc4 5
ef3 5
hg7 5
Ka1 5
Kd5 5
Ld1 5
Le1 5
Le8 5
Na7 5
Nb1 5
Nb4 5
Nc2 5
Qb1 5
Qh2 5
Rg8 5
ba6 4
cd3 4
ed7 4
Kb2 4
Kh3 4
Ld8 4
Nf2 4
Qa1 4
Qf1 4
a7 3
ba3 3
bc4 3
cb3 3
cb4 3
cb5 3
dc7 3
de7 3
f7 3
fg4 3
gh7 3
Kc5 3
Ke1 3
Kf4 3
Lb8 3
Lh8 3
Qa2 3
bc7 2
cb7 2
cd6 2
dc3 2
fe7 2
g8=Q 2
h7 2
Kb3 2
Kb4 2
Kb5 2
Ke5 2
Kg4 2
Lc8 2
Na2 2
Na8 2
Nb8 2
Nc1 2
Ne1 2
Nh8 2
ab7 1
ba4 1
ba7 1
c8=Q 1
cd7 1
dc8=Q 1
de3 1
e8=N 1
e8=Q 1
ed3 1
fe8=Q 1
fg8=Q 1
g7 1
gf7 1
Ka2 1
Ka6 1
Ka7 1
Kb6 1
Kc1 1
Kf5 1
Kh4 1
Lg1 1
Lg8 1
Qg1 1
Rh2 1

I think the subject of this thread has previously been treated by Tim Krabbê on his chess site (which has sadly become less active). But I'm too lazy to search.

Mr. Christian Sánchez and Mr. Uff Da, we just love your work. Please teach us how to these statistics, like in the style "chess for dummies"? Thank you in advance.

"Others seem to find it exciting and thrilling, very thought provoking, lots of fun. They love it. They really, really do." (Luke, August 4, 5:58 PM)

See what I mean? People just love this stuff! Mig, you should do more of this! It's so cool. Yawn. zzzzz.

Why don't you keep sleeping, Luke? Rather than waking up once in a while and posting, just to tell us that you are bored ... personally I am also getting bored about posts expressing nothing but boredom ,:)

Thomas –

As you are thinking of your witty rejoinder, perhaps this can help:

Luke, please do us all a favor and go to sleep….forever.

Luke, you bore the hell out of me.

Luke, who cares a rat’s ass what you think?

Luke, just STFU and go away.

Luke, you are such a stupid person ( Manu)

Sigh. (chesshire cat)

Good luck with your snappy comeback. I know you can do it, Thomas.

Luke, no matter how much rubbish you post, you will not fulfil your attention deficit/low self esteem needs on this blog. Engage a little more with the real world, or else seek professional help.
Your naked cries for attention are increasingly painful to behold, and I say that in all kindness.
You are not doing yourself any favours.

I think the posters should ask themselves some questions before posting.

1. What idea will I like to express?
2. Who would be interested in knowing about this?
3. How can I express this in an interesting way?

If they don’t have a clear idea to write about and can’t think of anyone this might interest, them they shouldn’t post.

But if they have an idea which they think might interest other people, then go ahead an express this in a capturing way.

I would be interested to know how presumed magic moves like Nd5 work out when they can be sustained for more than 5-6 moves. There are several of these kind of moves that ancient wisdom tells us almost always gives an advantage...

stendec,
Maintaining Nd5 for at least five moves scores 0.59, which is significantly better than the 0.56 for having it there for only one move (e.g., it is immediately captured, on its way elsewhere, or chased off).

Nf8 is great (0.69) if it is there for only one move, but if it stays for another move, its success drops to 0.63.

guest: To research, use the "Position" or "Manoeuvres" [sic] search mask in Chessbase.

"A knight on the sixth rank is worth a rook."

The game statistics agree. When White is down by knight vs. rook for at least two moves but has knight is on the sixth rank, White scores 0.52. Stats are better for c-f files (0.54) than for the rook and knight files (0.48).

I guess that early king moves into the other half of the board are the losingest, in a statistical sense. E.g. if black plays something like Ka4 before move 20, it is most probably because he is being mated.

But maybe that doesn't apply to games of IMs or better, because they have probably already resigned earlier, in case of such an accident.

I searched in a database with ~900,000 games:

bKa4 up to move 20: 100% white wins (8).

Of course this doesn't make very much sense, because it's not a move Black wanted to play, but was forced to.

I wonder if there's a useful distinction to be made here between winning as a *process*, and winning as an *event*.

To ask which move is the "winningest", is to ask which move is correlated most strongly with the event of having won. To ask whether having your knight on d5 for five moves is associated with winning, is really to ask about the process of winning (i.e., to begin to ask how the singular event of "having won" is brought about through a sustained process).

I've always been horribly partial towards bxa1=N.

I should think that most any move the features the underpromotion of a pawn is likely to be overwhelmingly winning - on a percentage basis perhaps slightly more so than promotion to a queen.

I thought so too, Calvin, but usually in real games, underpromotion is done as a kind of joke, in a situation in which whatever the pawn promotes to will be captured immediately, so the stats are not nearly as strong as promoting to queen.

Great stats (0.35) for bK on e6 before move 10, but the wK on e3 before move 10 is not good (0.47).

First you get a PGN file of Alekhine games (filtered). Then you open it with Word, and use the find-and-replace feature to delete headers, move numbers, x's, +'s, #'s (but don't mix white and black moves!) and the disambiguation symbols.
Next you copy-and-paste the column of white and black moves into Excel:

e4 e5
Nf3 Nc6
Bb5 a6

Then you separate it on two columns, sort it, do "subtotal". The end.

Interesting idea, Christian. Comparing this list with the parallel losing list would make it more useful.

I having a new segment on my late nite talk show. It's called "stupid database tricks". I want you to come on the show Mig, The audience is sure to love it! I know a weiner when I see it!

The Best losing move....Marry early like Karjakin!

Overheard in the Marshall Chess Club one afternoon circa 1970. Two regulars are playing an informal game and two other regulars, old-timers, are kibitzing.

A move is played on the board.
Old-timer kibitzer#1: "Knight to king bishop four. Alekhine played that move in 1928!"

Old-time kibitzer #2: "Gee, Boris, I didn't know you knew so much about opening theory!"

OTK #1: "Who said anything about opening theory? I'm sure Alekhine played a knight to king bishop four at least once in 1928!"

My database has five games with Nf4 by Alekhine in 1928. All five wins, all five in simuls. I wonder if the tendency to play Nf4 against patzers can be said to characterize a player's style? This ought to be measured! Frequency, result and average rating of opponents in games with or without Nf4 respectively, to be calculated for each of the world champions. Anyone?

But, bondegnasker, you don't know whether the move being kibitzed by OTKs #1 and 2 was made by White or Black. If the latter, then it would have been ...Nf5!

fly: You're right, I forgot about that! But actually, I can't find any games by Alekhine where he played Nf5 as black in 1928. I suppose that confirms the hypothesis that it's a simul move, the big shots always play the white pieces in simuls.

Black moves (which lose).

Total 15344
Nf6 669
d5 374
O-O 372
Nd7 351
Nc6 348
e6 313
Le7 306
e5 290
Re8 264
c5 260
c6 252
Rd8 212
Nd5 198
Qe7 195
a6 193
g6 193
h6 191
Ld7 171
d6 168
b6 166
Rc8 157
Ne4 153
Kh8 148
b5 147
f5 146
Lb4 146
Qc7 145
Ne5 141
Ne7 136
Lc5 130
Le6 130
f6 128
Lb7 123
Qd7 120
Kf8 114
Ld6 114
Kf7 112
cd4 105
Kg7 103
ed4 101
Lg4 101
Qf6 99
Kg8 93
Qb6 93
ed5 92
Lf5 92
dc4 90
Rb8 90
Qa5 87
Rf8 83
g5 76
Nc5 76
Lc3 75
a5 74
de4 74
Ng6 73
Kh7 72
Qd5 72
Nb6 71
Qd6 71
Ke7 67
Qd8 67
Lf6 65
Lg7 65
Rg8 64
h5 62
Nf8 60
Rd7 60
Ld5 59
Re7 59
Lc6 58
Qe8 58
Nd4 57
Nc3 56
Rf7 56
Kd7 55
Nh5 55
cd5 53
Na5 52
Ne8 52
Qe6 52
Nc4 49
Qe5 49
Qf7 49
de5 48
gf6 48
bc6 47
Nd6 47
Lf8 46
Ng4 46
Lg5 45
Nf5 45
fe6 44
Kd8 44
Ke8 44
Ld4 43
Qc5 43
Le5 42
Lf3 42
Lg6 40
Qc6 40
Ne6 39
O-O-O 39
Qd4 39
Qg5 39
Lb6 37
Lc8 37
b4 36
Rc7 36
Nf4 35
Qc8 35
La6 34
Nb4 34
Qb7 33
Re6 33
ab5 32
Ra8 32
e4 31
ef4 31
Le4 31
Rc6 31
Qf5 30
Qh4 30
Rd6 30
Nc7 29
Nd3 29
Qb4 29
Qg6 29
Ra7 29
Rf6 28
Kh6 27
Ld8 27
Le8 27
Nd8 27
Rd5 27
Kg6 26
Qb5 26
Qf8 26
Re4 26
Re5 26
Lb5 25
Qe4 25
ef5 24
Rb7 24
c4 23
fg6 23
Ld3 23
Nh7 23
Qg7 23
gf5 22
Kb8 22
Kf6 22
Na6 22
Nb8 22
Nf7 22
Qb2 22
Qc4 22
Qf4 22
Rg7 22
Ld2 21
Ng8 21
fe4 20
Lc4 20
Ng5 20
Qa6 20
Qb8 20
Qh5 20
Rc5 20
Rd1 20
hg5 19
Kd6 19
Ne3 19
Qc3 19
Rh8 19
dc3 18
fe5 18
Lh5 18
Nc8 18
Ng7 18
Rc2 18
Re1 18
Rf4 18
bc5 17
dc5 17
gf4 17
Kc8 17
La5 17
Qd3 17
Rb2 17
d4 16
gh6 16
Ke6 16
Le2 16
Lf7 16
Nb3 16
Rd4 16
Rf5 16
Rg6 16
Rh7 16
ab6 15
cd6 15
Nf3 15
Nh6 15
Rb6 15
hg6 14
Qa4 14
Qa8 14
Qg4 14
Rg5 14
dc6 13
h4 13
Kc7 13
Lh6 13
Nb7 13
Qg3 13
Rd2 13
Lf4 12
Qa2 12
Qh6 12
Rc1 12
Rd3 12
f4 11
Kg5 11
Qd1 11
Qe3 11
Ra2 11
Rg4 11
a4 10
ab4 10
cb4 10
cb5 10
fg4 10
Kc6 10
Le3 10
Lg2 10
Lh3 10
Qf3 10
Qh3 10
Rc3 10
Rc4 10
Re2 10
ba4 9
bc4 9
Lc7 9
Na7 9
Nb5 9
Qa3 9
Qa7 9
Ra6 9
Rb4 9
ef3 8
g4 8
hg4 8
Kb7 8
Kh5 8
La3 8
Lf2 8
Na4 8
Rb5 8
Re3 8
Rf1 8
Rh6 8
fg5 7
Lb3 7
Lb8 7
Nd2 7
Ng3 7
Qb1 7
Qb3 7
Qd2 7
Qe2 7
Qg2 7
Ra4 7
ba5 6
bc3 6
Kb6 6
Kc5 6
Kd5 6
Kf4 6
Kf5 6
Kg4 6
Lh4 6
Ra1 6
Ra3 6
Ra5 6
Rg3 6
Rh5 6
b3 5
d3 5
gh5 5
La4 5
La8 5
Lf1 5
Lg3 5
Lh7 5
Nb2 5
Nc2 5
Nh4 5
Qa1 5
Qc2 5
Qe1 5
Qf2 5
Qg8 5
Qh8 5
ba6 4
f3 4
Ka8 4
Ke5 4
La7 4
Lc2 4
Ne2 4
Qc1 4
Rb1 4
Rh4 4
a3 3
ba3 3
cb2 3
de6 3
fg3 3
gf3 3
gh4 3
Ka6 3
Lb2 3
Ld1 3
Na1 3
Nh8 3
Qf1 3
Qh1 3
Rf3 3
Rg2 3
Rh2 3
c3 2
cb6 2
d2 2
de3 2
e2 2
e3 2
ef6 2
h3 2
hg3 2
Ka7 2
Kb5 2
Kc4 2
Ke4 2
Kh4 2
La2 2
Lb1 2
Le1 2
Lg1 2
Lg8 2
Lh2 2
Na2 2
Na8 2
Nd1 2
Nf1 2
Nf2 2
Qh2 2
Qh7 2
Rf2 2
a1=Q 1
a2 1
b1=Q 1
c2 1
cb3 1
cd3 1
ed3 1
ed6 1
fe3 1
fg2 1
gh2 1
gh3 1
Ke3 1
Kf3 1
La1 1
Lc1 1
Lh1 1
Lh8 1
Nc1 1
Ng2 1
Qg1 1
Rb3 1
Rg1 1
Rh1 1
Rh3 1

I remember that a few years ago, GM Neil MacDonald pointed out that Kb5 is the most successful move with white, if you happen to be a world champion.
Morphy and Capablanca each scored 100% whenever they played this move, Fischer got 9.5/10, and Spassky himself managed 3/3 against Fischer.
Also, it seems that if you lose more than once after playing Kb5 with white, you will never become world champion (for instance Keres and Korchnoi)

Someone wondered "what does this mean for Magnus Carlsen?"

Answer = nothing.

Can you use these database tricks to evaluate old saws like "develop Knights before Bishops" or "don't move the Queen too early." What move by the database should I move my queen for the first time?

tja

On more interest are midgame rules like centralize your rooks or centralize your queen. I wonder if statistically Re4 Re5 Rd4 and Rd5 are associated with winning? Same four squares for the queen? That would validate those rules.