chessgames.com
Members · Prefs · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing

(If you register a free account you won't see all these ads!)
Robert James Fischer vs Bent Larsen
"Denver Omelette" (game of the day Oct-26-2007)
Fischer - Larsen Candidates Semifinal (1971), Denver, CO USA, rd 1, Jul-06
French Defense: Winawer. Advance Variation (C19)  ·  1-0
Move:
Last:

Click Here to play Guess-the-Move
Given 62 times; par: 60 [what's this?]

explore this opening
find similar games 13 more Fischer/Larsen games
sac: 23.Rxe5 PGN: download | view | print Help: general | java-troubleshooting

TIP: If you register a free account you will be able to create game collections and add games and notes to them. For more information on game collections, see our Help Page.

PGN Viewer:  What is this?
For help with the default chess viewer, please see the Pgn4web Quickstart Guide.

Robert James Fischer vs Bent Larsen (1971)
Fischer holds out his fists to let Larsen pick who has white for the first game of the candidates match, 1971.


Kibitzer's Corner
< Earlier Kibitzing  · PAGE 10 OF 10 ·  Later Kibitzing>
Jun-08-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move (part 2 of 3)

Two things I think are noteworthy to point out:

(1) 35.Bb6, the move Fischer played, is apparently the only move to preserve White's advantage. White's advantage is its passed a-pawn, if you remove that pawn the game is, perhaps surprisingly, apparently a draw. And even one lost tempo allows Black to draw since White's king and rook are out of play at the moment. In that respect, the following line from Stockfish at d=44 is particularly instructive:

35.Bc5 g4 36.Bb6(a) Qxc2(b) 37.a5 Qb2(c) 38.Rf1 Kg6(d) 39.Bg1(e) Qa3(f) 40.Bd7 Qxa5 41.Bxf5+ Qxf5(g) 42.Rxf5 Kxf5(h) 43.Bd4 Ke4 44.Ba1(i) h5 45.Bf6(j) Kd3 46.Kg2 c3 47.Bxc3 Kxc3(k)

(a) This move, which represents the winning approach but with a loss of a tempo, shows what could happen. Although, of course, Fischer (and for that matter, neither you nor I nor anyone frequently visiting this site) would have likely played this.

(b) The elimination of White's c-pawn gives Black a powerful passed pawn of its own and provides sufficient counterplay to secure the draw. In this line (and presumably in others as well) White's a-pawn advances one move too late.

(c) The key move in Black's defense/counterplay. Not only does it prevent the further advance of White's a-pawn by threatening to capture White's DSB but it clears the way for the advance of Black's c-pawn. Taking the time to move the DSB out of harm's way is the one tempo that Black needs to secure the draw.

(d) Another fine and necessary move. 38...c3 allows 39.Bd4+ and not only does White regain the lost tempo to advance it’s a-pawn but allows White, when (not if!) necessary, to give up its DSB for Black's passed c-pawn.

(e) An odd looking but seemingly effective move. It protects White's king from back rank checks and a possible perpetual yet it still keeps an eye on the a-pawn's advance while allowing White's rook to join the game. So in effect White increases the mobility of its most powerful piece in exchange for reducing the mobility of a less powerful piece.

(f) But now the a-pawn is lost and without it any White winning chances are gone. White's attempts at counterplay on the k-side fall short.

(g) Sealing the draw. White will have to give up its DSB for Black's c-pawn, drawing Black's king away from the k-side and allowing White to establish a fortress on the k-side.

(h) And this position is a theoretical draw per the Lomonosov tablebases. But although there are many ways to draw, the rest of the line is still instructive.

(i) Most bishop moves along the long diagonal (except 44.Be5 of course!) also lead to theoretical draws.

(j) Most bishop moves along the long diagonal (except 44.Bd4 and 44.Be5 of course!) also lead to theoretical draws, as do 44.Kg2 and (perhaps surprisingly) 44.h3. In the latter case after either 44...gxh3 or 44...g3 45.Kg2 h4 White's king cannot be chased from the right corner and a stalemate results after White gives up its bishop for Black's c-pawn.

(k) And this is a clear draw, even for me, particularly after 48.h3.


click for larger view

Jun-08-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move (part 3 of 3)

(2) I'll bet that you forgot that there was a second noteworthy thing! But do note that although all 3 engines indicate that White has an advantage, exactly HOW much of an advantage is questionable. Typical evaluation ranges are:

[0.00] – [+0.49]: Equal position (i.e. chances are equal, not necessarily a draw) (=).

[+0.50] – [+0.99]: White stands slightly better. ( ).

[+1.00] – [+1.49]: White has a clear advantage ( ).

[+1.50] and up: White has a winning position (++).

FWIW I personally prefer the more conservative ranges since they lead to less disappointments:

[+1.00] – [+1.99]: White has a clear advantage ( ).

[+2.00] and up: White has a winning position (++).

Using these evaluation ranges Stockfish indicates that at [+3.37] White has a winning position. Komodo indicates that at [+1.21] White has a clear advantage but does not necessarily have a winning position, and Houdini indicates that at [+0.76] White stands slightly better.

I suspect that this is strictly a function of the search depth. From a given starting position the deeper an engine searches the more the advantage of a winning position become clear, and the higher its evaluations become. Unless, of course, AylerKupp's Corollary to Murphy's Law (AKC2ML) is in effect (see my forum's header).

I should add that I did these analysis (as well as all the earlier ones) with each engine having 5-piece Syzygy tablebase support. Unfortunately the 6-piece Syzygy probing code does not work on my archaic 32-bit computer, it needs a 64-bit computer. Had I been able to use the 6-piece Syzygy tablebases, Houdini and Komodo might have reached the proper conclusion earlier, or at least have had higher evaluations for White when I stopped their analyses.

So, in (thankfully!) conclusion, it appears to me that 34...g5 was the losing move and not the first of a series of inaccuracies. And after 34...g5 there does not appear to be any way for Larsen to save the game, assuming perfect play by Fischer. Which, of course, he was able to achieve.

But it was certainly close, only one tempo separating the loss from a draw! Just think, since this was the first game of the match, if Larsen had played 34...Ke6, 34...Ke5, or even maybe 34...g6 then Larsen might have drawn the game and put up a better resistance. Of course, it could have just meant that Fischer would have won the match by a score of 6 1/2 – 1/2 instead of 6 – 0. As Fischer himself would have put it, "If Larsen had played differently, he simply would have lost the match by a different score." But it might have given Larsen confidence and not cause him to try so hard in the remaining games.

And, had Larsen drawn this first game, it would have made life more difficult for Taimanov since he then would have been the only GM to lose a match to Fischer by the score of 6-0.

I'm still now sure why 34...Ke6, 34...Ke5, or 34...g6 would have drawn while 34...g5 loses, but that's the subject of another investigation. And definitely not today!

Jun-08-17  Howard: Thanks very, very much as usual!!
Jun-10-17
Premium Chessgames Member
  4tmac: Great work AK! 34...g5 may be the losing move. I am "sure" 34...Ke5! (Ke6) maybe even g6 or even <QxP!?> draws .... had to post this irrelevent Kodinifish spew (there is always iggy) < 34... Qxc2 35. a5 Qd2 > (35... Qd3 36. a6 c3 37. Bb7 c2 38. Rc1 Qd1+ 39. Rxd1 cxd1=Q+ 40. Bg1 Qb3 41. Ba8 (41. Bg2 f4) 41... Qb8 42. Bg2) black can resign this beautiful position


click for larger view

< continue with 34..QxP?! 35a5 Qd2! 36. a6 c3 37. Bb8 c2 38. a7 c1=Q 39. a8=Q ! > yields an amazing position (diagram) where black should have no chance! (best computer stuff example follows)....however I believe black can draw with <39...QxR+!!> even though there is *no* perp! (although the "perp threat" probably plays a roll) I will spare you this "possible draw" a mess beginning with Queen checks etc.


click for larger view

an example if black keeps his extra Q: 39... Qc5?! 40. Bc7! Qdd4 41. Bd8+! Kf7 42. Re1 Kg6 43. Be7! Qcc4 44. Qe8+ Kh6 45. Bf3 is crushing.

Jun-10-17  Howard: Just about every source I've seen calls Larsen's 32...Qxc3 the "decisive" mistake. Wonder when we'll see that viewpoint revised in future books.
Jun-10-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move (reprise)

OK, I suppose that your question triggered my inner pit bull so I tried to make absolutely sure that 34...g5 was the losing move. After 34...g5 Houdini and Komodo both indicate the following moves as being best (and seemingly almost forced or at least logical): 35.Bb6 Qxc2 36.a5 Qb2 37.Bd8+ Ke6 38.a6 Qf2 39.Bxg5 c3 40.Ra1


click for larger view

After 35.Bb6 Stockfish prefers 35...Qa3 36.a5 Kg6 and a completely different line as its path to a win. So I decided to find out if all 3 engines consider the position to be a winning one after 40.Ra1:

Black's Houdini 4 Komodo 10 Stockfish 8

Move d=31 d=32 d=50 <St.Avg>

-------- --------- -------- --------- ----------

40...f4 [+0.29] [+1.22] [+1.15] <[+0.89]>

40...c2 [+0.92] [+1.78] [+2.09] <[+1.70]>

40...Qa7 [+1.22] [+1.87] [+2.25] <[1.87]>

40...Qc5 [+18.83] [+250.00] [+128.39] <[+147.60]>

40...Qb6 [+150.72] [+250.00] [+128.42] <[+178.34]>

No! The beginning of the winning line is apparently NOT 35.Bb6 Qxc2 36.a5 Qb2 37.Bd8+ Ke6 38.a6 Qf2 39.Bxg5 c3 40.Ra1 as presumably indicated by Houdini and Komodo based on their increasing evaluations (and more than a little faith from me). At least none of the 3 engines, including Stockfish which rejected this line, could find any definitive win after 40...f4, and I let them run a long time.. If a definitive winning line is to be found either an earlier deviation from this line is necessary, perhaps after the game's 35.Bb6 Qxc2 36.a5 Qb2 37.Bd8+ Ke6 38.a6 where Larsen with 38...Qa3 deviated from Houdini's and Komodo's 38...Qf2. If not, then Stockfish's 35.Bb6 (this still seems desirable/necessary) 35...Qa3 36.a5 Kg6 should be investigated further using Houdini and Komodo. But, since that deviates from the game's continuation, it's sort of outside the scope of your original question. Several possibilities for earlier wins have been suggested.

A note about those "funny" evaluations in case you don’t know. The Syzygy tablebases indicate a won position (though not necessarily a mate, just overwhelming material advantage) by returning an evaluation of [128] (the GUI may change the sign according to the GUI settings). Their evaluation is not the result of the engines' evaluation function, at least not entirely. And, of course, averaging artificial evaluations only results in artificial averages.

Different engines then manipulate this [128] and report it differently. For example:

1. Houdini adds its evaluation to the 128. So the [+150.72] for 40...Qh6 indicates that Houdini evaluated the position at 150.72 – 128 = [+22.72].

2. Komodo adds some factor (which I can remember) to a won position and reports a mate as [+250.00].

3. Stockfish does something similar to Komodo but does not differentiate between won positions and mate, it reports them all as [+128.xx].

At least this is what I think the 3 engines do. I had a good explanation of how they treat the Syzygy [128] evaluation but I lost it and can't find it on-line again.

Jun-10-17
Premium Chessgames Member
  AylerKupp: <<4tmac> Great work AK!>

Thanks, but not quite yet. As you'll see above and below I am no longer sure that 34...g5 was the losing move. More analysis coming.

Jun-10-17
Premium Chessgames Member
  AylerKupp: <Howard> More analysis, 38...Qa3 response (part 1 of 2)

As part of the continuing investigation, here is how the same 3 engines evaluated the position after 35.Bb6 Qxc2 36.a5 Qb2 37.Bd8+ Ke6 38.a6 Qa3 where both Houdini and Komodo with 38...Qf2 deviated from Larsen's 38...Qa3. After all, both Houdini and Komodo have indicated that there is no apparent win after 38...Qf2, although White is then definitely better.


click for larger view

White's Houdini 4 Komodo 10 Stockfish 8

Move d=22 d=25 d=32 <St.Avg>

-------- --------- -------- --------- ----------

39.Re1+ [+14.52] [+10.38] [+128.36] <[+51.09]>

39.Bb7 [+2.24] [+3.84] [+5.98] <[+4.02>

39.Be7 [0.00] [+0.06] [0.00] <[+0.02]>

39.Rd1 [0.00] [0.00] [0.00] <[0.00]>

39.Bc7 [0.00] [------] [0.00] <[0.00]>

39.Bh4 [------] [0.00] [------] <[0.00]>

White has at least 2 likely winning lines after 38...Qa3, with 39.Re1+ being by far the most forcing. Stockfish's evaluation of [+128.36] is again an artifact of the Syzygy tablebases and is not (entirely) the result of Stockfish's evaluation function.

Both 38...Qf2 and 38...Qa3 serve the same purpose, preventing the immediate advance of White's a-pawn and clearing the way for the advance of Black's c-pawn. But, since Larsen returns his queen to the g1-a7 diagonal after 39...Qc5, 38...Qf2 apparently saves a tempo and then 39.Bb7 (as in the game) allows 39...c3 which prevents the 40.Rb1, 41.Bb6 continuation that let to Larsen's resignation since after 40.Rb1 c2 White's rook must abandon the b-file, preventing 41.Bb6.

So maybe after 38...Qf2 there is no win for White and it's 38...Qa3 and not 34...g5 which is the proverbial last mistake that throws away the draw.

So 38...Qa3 (see below) leads to a loss and in an even more forcing way than Fischer's 39.Rb1. All 3 engines came up with the same winning line and position, although they differed in how they finished the kill. I didn't bother letting them run a long time since the results seemed so conclusive after 39.Re1+ But here are fragments of their winning line since they are interesting and it may not be obvious as to why they are so conclusively winning.

Jun-10-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Not necessarily 38...Qa3.

Now that I think I've shown that 38...Qa3 is <a> losing move, perhaps <the> losing move and the last mistake, what needs to be shown to confirm it is that 38...Qf2 as indicated by Houdini and Komodo does not necessarily lead to a losing position for Black.

So here is the 3 engines' analysis of the position after 35.Bb6 Qxc2 36.a5 Qb2 37.Bd8+ Ke6 38.a6 Qf2:


click for larger view

White's Houdini 4 Komodo 10 Stockfish 8

Move d=30 d=36 d=49 <St.Avg>

-------- --------- -------- --------- ----------

39.Rb1 [+4.28] [+3.01] [+9.34] <[+5.54]>

39.Bxg5 [+0.91] [+1.25] [+0.99] <[+1.05]>

39.Ra1 [0.00] [+0.87] [+0.25] <[+0.37]>

39.Bg2 [0.00] [+0.93] [+0.12] <[+0.35]>

39.Bb7 [0.00] [+0.34] [0.00] <[+0.11]>

Unfortunately, I suppose, after 38...Qf2, 39.Rb1 also seems to lead to a win so the gain of a tempo by 38...Qf2 instead of Larsen's 38...Qa3 and 38...Qc5 is apparently not significant. So if 38...Qa3 is a losing move, the alternative 38...Qf2 is not any better. So the losing move must have come earlier. Oh well, time to backtrack.

Jun-10-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Backtrack #1

Since neither 38..Qa3 nor 38...Qf2 seem to have been the losing moves in the game continuation after 34...g5 (both lead to winning positions for White), time to start backtracking. The seemingly natural place to start is after 35.Bb6 Qxc2, logical as 35...Qxc2 might seem.


click for larger view

Here is what the 3 engines came up with:

White's Houdini 4 Komodo 10 Stockfish 8

Move d=28 d=32 d=44 <St.Avg>

-------- --------- -------- --------- ----------

36.a5 [+0.43] [+2.61] [+6.61] <[+3.22]>

36.Bd8+ [0.00] [+0.44] [0.00] <[+0.15]>

36.Bg2 [-0.36] [0.00] [-0.21] <[-0.19]>

36.Bd4+ [-0.78] [0.00] [-0.79] <[-0.52]>

36.Bb7 [-1.62] [0.00] [-1.75] <[-1.12]>

Now, that's a shocker! Both Komodo and Stockfish consider 35.Bd6 Qxc2 36.a5 as winning for White while Houdini considers the position equal. Furthermore, all 3 engines consider Fischer's move, 36.Bd8+, as providing no more than equal chances. And anything other than 36.a5 or 36.Bd8+ seem to lead to a better game for <Black>, not White. It shows how complex the position is.

Here is the PV for each of the 3 engines:

Houdini 4: [+0.43], d=28: 36.a5 <Qa2> 37.Bb7 Qb2 38.Bd8+ Ke5 39.a6 c3 40.Bxg5 c2 41.Re1+ Kd6 42.Bf4+ Kd7 43.Rg1 Qb4 44.Bc1 Qd4 45.a7 Qxa7 46.Rg7+ Kd6 47.Bf4+ Kc5 48.Be3+ Kd6 49.Bxa7 c1Q+ 50.Rg1 Qc7 51.Rb1 Qc2 52.Ra1 Qd3 53.Bb8+ Ke6 54.Re1+ Kf6 55.Bg3 Kg6 56.Be5 Kf7 57.Bc6 Qd2 58.Bg3 Kg6 59.Re6+ Kg5 60.Be1 Qb2 61.Bf3

Komodo 10: [+2.61], d=32: 36.a5 <Qb2> 37.Bd8+ Ke6 38.a6 Qf2 39.Rb1 c3 40.Bb6 Qa2 41.Re1+ Kd6 42.Bb7 c2 43.Bg1 Kc7 44.Re8 g4 45.Rc8+ Kd7 46.a7 f4 47.Rxc2 Qxc2 48.a8Q f3 49.Bxf3 gxf3 50.Qd5+ Ke8 51.Qxf3 Qb1 52.h3 Ke7 53.Kh2 Qb2+ 54.Bf2 Qc2 55.Kg3 Qg6+ 56.Qg4 Qd6+ 57.Kg2 Qc6+ 58.Qf3 Qc4 59.Be3 Qa2+ 60.Kg3 Qe6 61.Bc5+ Kd7

Stockfish 8: [+6.61], d=44: 36.a5 <Qb2> 37.Be3 Qa3 38.Bxg5+ Kf7 39.Bd5+ Ke8 40.Re1+ Kd7 41.a6 Qa5 42.Rd1 Qa4 43.Rd2 f4 44.Bf6 Kc7 45.Be5+ Kd7 46.Bxc4+ Ke7 47.Rd4 Qc6+ 48.Kg1 Qh6 49.Bd6+ Ke8 50.Rd5 f3 51.Bb5+ Kf7 52.Kf2 Qh3 53.Bg3 Qg2+ 54.Ke3 Qg1+ 55.Ke4 f2 56.Rf5+ Kg6 57.Rxf2 h5 58.Bd3 h4 59.Kd5+ Kh5 60.Rf5+ Kg4 61.Rf4+ Kg5 62.Rxh4 Qb6 63.Re4 Qb3+ 64.Bc4 Qb2 65.Re5+ Kf6 66.Bh4+ Kg6 67.Re6+ Kh5

The first difference is that Houdini evaluates 36...Qa2 as Black's best response to 36.a5 and both Komodo and Stockfish evaluate 36...Qb2, a move that I thought was so logical, as Black's best response to 36...a5. <So either Houdini is wrong (possibly as a result of running the analysis to a lower search depth) or both Komodo and Stockfish are>. The best way to find out is to run a Houdini analysis after 35.a5 Qb2 and Komodo and Stockfish analyses after 35.a5 Qa2 and see if their evaluations are reversed. Coming up next.

Jun-11-17
Premium Chessgames Member
  AylerKupp: <Howard> More analysis, 38...Qa3 response (part 2 of 2)

Ooops! I forgot to post the second (and best!) part of Fischer vs Larsen, 1971 (kibitz #230). And without it some of the subsequent posts don't make much sense. Yes, I don't know whether they make sense anyway, but here is the intended second part:

Houdini 4: [+14.52], d=22: 39.Re1+ Kf7 (if 39...Kd6, 40.Be7+) 40.Bd5+ Kg6 41.a7(a) h6(b) 42.Rg1+ Kh7 43.a8Q Qxa8 44.Bxa8(c) ...

(a) The point, the a-pawn cannot be prevented from queening. If 41...Qxa7 then 42.Re6+ Kf7 43.Re7+ (42...Kh5 43.Bf3#) 43...K(any) 44.Rxa7 and Black's queen is lost. Maybe Larsen saw this and played 38.Qa3 to prevent it) 43.Re7+.


click for larger view

A side point that I initially missed and maybe you might also. I thought that Houdini misevaluated this position because after 43.Re7+ Black could play 43...Qxe7 and after 44.Bxe7 Kxe7 45.Bxc4


click for larger view

This position is a draw per the Lomonosov tablebases since even if White captures all of Black's pawns his bishop does not control his h-pawn's queening square. So I thought that maybe this apparent misevaluation was due to Houdini having only 5-piece tablebase support. But I missed that after 43.Re7+ Black is also in check by the Bd5 so 43...Qxe7 is not a legal move. Imagine seeing this OTB given the pressures of a WCC Candidates Match and the likelihood that, given the move numbers, both players (particularly Larsen) could likely have been in time pressure! The ability of these top players never ceases to amaze me.

(b) So Houdini, correctly evaluating the line starting with 41...Qxa7 as a win for White, allows White's a-pawn to queen, presumably evaluating this option as the lesser of two or more evils.

(c) And now it's totally hopeless for Black, R+2B+P vs. 4P. Even you or I could probably beat Larsen as White from this position.


click for larger view

Komodo 10: [+10.38], d=25: 39.Re1+ Kf7 40.Bd5+ Kg6 41.a7(a) h6 42.a8Q Qxa8 43.Bxa8(b) ...

(a) Komodo reaches the same conclusion as Houdini.

(b) Reaching the same losing position as Houdini

Stockfish 8: [+128.36], d=32: 39.Re1+ Kf7 40.Bd5+ Kg6 41.a7(a) Kh5 42.a8Q Qxa8 43.Bxa8(b) ...

(a) Stockfish also reaches Houdini's and Komodo's conclusion.

(b) And the same winning position.

So what now remains to be seen is if any of the 3 engines can find a win for White after 38...Qf2.

Jun-11-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – The Validation

OK, who’s right, Houdini 4 or Komodo 10 / Stockfish 8? Here is what the 3 engines came up with on their latest "assignment":

Houdini 4 after 36.a5 Qb2:


click for larger view

White's Houdini 4
Move d=28
-------- ---------
37.Be3 [+2.65]
37.Bd8+ [+0.33]
37.Rg2 [0.00]
37.Bc5 [0.00]
37.Re1 [0.00]

So Houdini validated Komodo's and Stockfish's evaluations; after 36.a5 Qb2 White had a winning position. Here is Houdini's PV:

Houdini: [+2.65], d=28: 37.Be3 Qa3 38.Bxg5+ Ke5 39.Re1+ Kd4 40.Bb7 h6 41.Bxh6 c3 42.a6 Kc4 43.Bg2 f4 44.Bxf4 c2 45.Bf1+ Kd5 46.h4 Kc5 47.Kg2 Qa2 48.Kg1 Qg8+ 49.Bg5 Qb3 50.Kf2 Qf7+ 51.Ke3 Kb4 52.Bd3 Kc3 53.Rc1 Qe6+ 54.Be4 Qxa6 55.Rxc2+ Kb3 56.Rc6 Qa7+ 57.Kf3 Qg1 58.Bd5+ Kb2 59.Bf6+ Ka3 60.Be4 Qh1+ 61.Kf4 Qh2+ 62.Kg4 Qg1+

Both Houdini and Stockfish evaluated 37.Be3 as superior instead of Fischer's 37.Bd8+ but Komodo indicated that White also had a winning position [+2.61] after 37.Bd8+.

Komodo 10 and Stockfish 8 after 36.a5 Qa2:


click for larger view

White's Komodo 10 Stockfish 8

Move d=34 d=38 <St.Avg>

-------- -------- --------- ----------

37.Bd4+ [+4.62] [+10.21] <[+7.42]>

37.Bb7 [+1.49] [+1.09] <[+1.29]>

37.Bd8+ [+0.35] [0.00] <[+0.18]>

37.Rg2 [0.00] [0.00] <[0.00]>

37.Be3 [0.00] [------] <[0.00]>

37.Re1 [------] [0.00] <[0.00]>

So Komodo and Stockfish validated Houdini's evaluation, after 36.a5 Qa2 White had a winning position. So all were right, White had a winning position after 36.a5 and <either> 36...Qa2 (Houdini) or 36...Qb2 (Komodo, Stockfish). To paraphrase Fischer, "White may play differently but he just wins differently:.

Here are Komodo's and Stockfish's PVs:

Komodo: [+4.62], d=34: 37.Bd4+ Ke6 38.Ra1 Qd2 39.Bg1 Qc3 40.a6 Qxa1 41.a7 Ke5 42.a8Q Qxa8 43.Bxa8 f4 44.Bc6 h6 45.Kg2 c3 46.Ba4 Kd5 47.Kf3 Kc4 48.Bc2 Kd5 49.Bb6 Ke5 50.Bd8 Ke6 51.Ba5 Kf7 52.Bxc3 Kg8 53.Be4 Kf7 54.Kg4 Kf8 55.Bg6 Kg8 56.Bd4 Kf8 57.Bf6 Kg8 58.Bc3 Kf8 59.Be4

Stockfish: {+10.21], d=38: 37.Bd4+ Kg6 38.Ra1 Qd2 39.Bg1 Qd6 40.Bb7 Qf6 41.a6 Qxa1 42.a7 Kf7 43.a8Q Qxa8 44.Bxa8 Ke6 45.Bd4 Kd6 46.Kg2 Ke6 47.Kf3 Kd6 48.Bb7 h6 49.Kg2 h5 50.Ba8 Ke6 51.Kg3 Kd6 52.Kf3 Ke6 53.Bb7 g4+ 54.Kg3 Kd6 55.Ba8 Ke6 56.Kf4 h4 57.Bc6 g3 58.hxg3 h3 59.Bb7 h2 60.Be5 c3 61.Bxc3 Kd6 62.Bd4

The (drum roll) conclusion to follow.

Jun-12-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Forward sliding.

Somewhat embarrassingly, after reviewing my so-called "evidence" for determining the losing move I realized that my logic was faulty and that I could not come to a defensible conclusion. So I have decided to revert to the "creeping barrage" approach where I step up one Black move at a time until I find the move when all 3 engines indicate that White will have a winning position <with the move actually played in the game>. That or when I finally decide that I can't figure out what the losing move was, throw up my hands, and give up.

FWIW I believe I first used the term "creeping barrage" here: Team White vs Team Black, 2015 (kibitz #2300) in reference to <kutztown46>'s approach to forward sliding.

In the meantime, this was another attempt to try to find the losing move. And in my haste, and due to some less than clear thinking after finding that my losing-move-finding logic was faulty, it just recycles old ground.. Here is what the 3 engines came up with after 36.a5 Qb2:


click for larger view

White's Houdini 4 Komodo 10 Stockfish 8

Move d=28 d=33 d=40 <St.Avg>

-------- --------- -------- --------- ----------

37.Be3 [+2.65] [+1.35] [+8.93] <[+51.09]>

37.Bd8+ [+0.33] [+1.26] [+3.73] <[+4.02>

37.Rg2 [0.00] [------] [0.00] <[+0.02]>

37.Bc5 [0.00] [------] [0.00] <[0.00]>

37.Re1 [0.00] [0.00] [0.00] <[0.00]>

37.Bd5 [------] [0.00] [------] <[0.00]>

37.Bg2 [------] [0.00] [------] <[0.00]>

At least a few things became clearer:

(1) Stockfish evaluates both 37.Be3 and 37.Bd8+ as leading to a winning position for White. So, if the analysis had been done only with Stockfish, we could have concluded that, while Stockfish evaluated 37.Be3 as being a better move, the move that Fischer played, 37.Bd8+, also led to a winning position for White and therefore 36...Qb2 was the losing move.

(2) Houdini evaluates 37.Be3 as leading to a winning position for White, but it did not evaluate the move that Fischer played, 37.Bd8+ as leading to a winning position for White. So if the analysis had been done only with Houdini, we could not conclude that 36...Qb2 was the losing move.

(3) Komodo did not evaluate either 37.Be3 or 37.Bd8+ as leading to a winning position for White, so as far as Komodo was concerned 36...Qb2 was not the losing move regardless of White's continuation.

This reinforces once again my opinion that, if you want to reach definitive conclusions about a move or a line using chess engines, you <MUST> use multiple engines and try to reach common ground, either by consensus or the preponderance of the evidence; i.e. averaging the evaluations or some other more sophisticated results aggregation techniques.

(4) I am at this moment not willing to conclude what Black's losing move was until all 3 engines indicate that a winning position for White is achievable following a specific Black response to the previous White move, although that might change as time goes on without reaching a reasonable conclusion. So, while boring and time consuming, the "creeping barrage" seems the most thorough approach, sliding forward one Black move at a time as played in the game.

Jun-13-17
Premium Chessgames Member
  diceman: Ayler, do you ever use “annotation” mode?
(I assume your programs have something like that)

I find on my relatively weak, old Fritz program, it does a lot better at finding mistakes/turning points in that mode.

It starts at the end of the game and works forward. I don't know how the software stores/uses its data, but I think it does a better job because it “saw” the position that resulted from its earlier moves.

Jun-22-17
Premium Chessgames Member
  AylerKupp: <diceman> No, and I don't even know what "annotation" mode is. And if I were to take you literally, how does it start at the end of the game and move forward since there are no moves after the end of the game?

Maybe you're thinking of what I know as <backward sliding>. This involves starting the analysis at either the end of the game or whenever the engine correctly sees a certain win. Then you back up one move (2 plies) and start the analysis again. If you have a large hash table then the analysis supposedly goes very quickly because most of the position evaluations are already in its hash table so they don't need to be recalculated. You then wait until the actual move played shows up as the engine's top move, stop, the analysis, back up one move, and repeat the process. Eventually the actual move does not show up as the engine's top move, and that is where the player's mistake was.

At least that's how I think it works. I personally have never tried it although I keep saying to myself that I should, as soon as I finish the analysis that I'm working on. Of course, as soon as I do that I start another analysis, so the moment never comes.

This process is documented in the book "Modern Chess Analysis" by Robin Smith (since deceased so I can't ask him for clarification. Yes, I thought of that), as applied to Short vs Timman, 1991. You can read more about it here: Short vs Timman, 1991 (kibitz #126). Maybe what I should do is use that game and walk through the process until I truly understand it.

In this game I would have started with the final position and worked backwards. Maybe I'll try it when I finish this analysis using my current "creeping barrage" approach. Assuming, of course, that I am not sick and tired of this game by then. I think that the reason I'm having all this trouble is that I tried to jump ahead unsystematically and got thoroughly confused. Or maybe it's just my nature. Another rationalization is that I tried to use multiple engines to confirm the results, and with my old computer with limited memory I had to reduce the size of each engines' hash tables so that all the code and data for each engine would fit in memory and avoid constant access to the disk. So the backward sliding approach would probably not have worked for me in this situation. But, like I said, it's only a rationalization for not thinking about doing as you suggested.

BTW, many players on this site do a combination of repeated forward sliding and backward sliding. They call this "sandpapering", a visualization of the process of sliding sandpaper back and forth over a surface to make it smooth. A clever name, I think.

Jun-22-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Beginning the creeping barrage (part 1 of 2)

Well, after some interruptions to deal with real life, including addressing <diceman>'s comment above, it's time to get back to the important stuff. As I mentioned earlier, since the analyses after 36.a5 Qb2 did not result in a set of conclusive evaluations leading to a winning position for White, I decided to try the "creeping barrage" forward sliding approach. I did decide that since most other moves lead to an equal position it's not necessary to use MPV=5 any longer, either; MPV=3 or even MPV=2 would likely suffice, leading to a slight speed-up of the analysis and the ability to reach deeper search depths in the same amount of time.

So, to reconfirm a previous analysis, I repeated it starting with the position after 34.Bxa7. As a bonus (or a curse!), I just got a copy of Komodo 11.01 so I've included it's analysis so that you can see how it compares with Komodo 10. And, FWIW, in the latest CCRL 40/40 tournament list, Komodo 11 is rated 36 Elo points higher than the version of Komodo 10 I've been using. So here is what the 4 engines came up with using MPV=3 after 34.Bxa7:


click for larger view

Black's Houdini 4 Komodo 10 Komodo 11 Stockfish 8

Move d=28 d=30 d=30 d=42 <St.Avg>

-------- --------- -------- --------- --------- ----------

34...Ke6 [0.00] [+0.84] [+0.25] [+0.00] <[+0.27]>

34...Ke5 [+0.03] [+0.70] [+0.36] [+0.21] <[+0.33]>

34...g6 [+0.16] [------] [+0.61] [+0.24] <[+0.34]>

34...Ke7 [------] [+1.14] [------] [------] <[+1.14]>

And, filling in the gaps by inserting evaluation values one centipawn greater than the lowest evaluation, we get the following. But remember that these averages would be optimistic

Black's Houdini 4 Komodo 10 Komodo 11 Stockfish 8

Move d=28 d=30 d=30 d=42 <St.Avg>

-------- --------- -------- --------- --------- ----------

34...Ke6 [0.00] [+0.84] [+0.25] [+0.00] <[+0.27]>

34...Ke5 [+0.03] [+0.70] [+0.36] [+0.21] <[+0.33]>

34...g6 [+0.16] [+1.15] [+0.61] [+0.24] <[+0.54]>

34...Ke7 [+0.17] [+1.14] [+0.62] [+0.25] <[+0.55]>

And these would be the move rankings independent of the evaluation:

Black's Houdini 4 Komodo 10 Komodo 11 Stockfish 8

Move d=28 d=30 d=30 d=42 <Avg>

-------- --------- -------- --------- --------- ----------

34...Ke6 1 2 1 1 <1.3>

34...Ke5 2 1 2 2 <1.8>

34...g6 3 4 3 3 <3.3>

34...Ke7 4 3 4 4 <3.8>

True Rank: 1 = [ 34....Ke6, 34...Ke5 ]; 2 = [ 34...g6, 34...Ke7 ]

Jun-22-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Beginning the creeping barrage (part 2 of 2)

I think that the following observations are, or may be, significant:

1. 34...g5 as played by Larsen did not show up in any engine's top 3 moves.

2. At d=28 Houdini's evaluations of its top 3 moves were almost the same on the previous run; 34...Ke6=[0.00], 34...Ke5=[+0.03], and 34...g6=[+0.18]. So the results were almost deterministic when using MPV=3, and the difference might be attributed to using MPV=5 in the previous run.

3. At d=30 Komodo 10's evaluation of its top 3 moves were somewhat different on the previous run; 34...Ke5=[+0.83], 34...Ke6=[+0.94], and 34..g6=[+1.04] and the move rankings were different, 34...Ke5 was evaluated higher than 34...Ke6, the reverse of this run. I don't know why.

4. At d=43 Stockfish's evaluation of its top 3 moves were also somewhat different on the previous run; 34...Ke6=[0.00], 34...Ke5=[+0.08], and 34...g6=[+0.20]. But at least its move rankings were the same.

5. At d=30 Komodo 11's evaluations were very different, much lower, than Komodo 10's and much closer than Houdini's and Stockfish's. And its move rankings were the same as those 2 engines. So using Komodo 11 will probably make it easier to get 3 engines to agree on the move rankings so I think I'll use Komodo 11 from now on. :-)

Jun-23-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Continuing the creeping barrage. (part 1 of 2)

These are the 3 engine evaluations after 34.Bxa7 g5 35.Bb6 using MPV=3:


click for larger view

Black's Houdini 4 Komodo 11 Stockfish 8

Move d=30 d=31 d=45 <St.Avg>

-------- --------- --------- --------- ----------

35...Kg6 [+1.38] [+2.12] [+4.54] <[+2.68]>

35...Qa3 [+1.64] [+2.12] [+4.63] <[+2.80]>

35...Qb4 [------] [------] [+4.77] <[+4.77]>

35...Qxc2 [+0.71] [+0.77] [------] <[+0.74]>

And, filling in the gaps by inserting evaluation values one centipawn greater than the lowest evaluation, we get the following. But remember that these averages would be optimistic

Black's Houdini 4 Komodo 11 Stockfish 8

Move d=30 d=31 d=45 <St.Avg>

-------- --------- --------- --------- ----------

35...Kg6 [+1.38] [+2.12] [+4.54] <[+2.68]>

35...Qxc2 [+0.71] [+0.77] [+4.78] <[+2.09]>

35...Qa3 [+1.64] [+2.12] [+4.63] <[+2.80]>

35...Qb4 [+1.65] [+2.13] [+4.77] <[+2.85]>

And these would be the move rankings independent of the evaluation:

Black's Houdini 4 Komodo 11 Stockfish 8

Move d=30 d=31 d=45 <St.Avg>

-------- --------- --------- --------- ----------

35...Kg6 2 2 1 <1.7>

35...Qxc2 1 1 4 <2.0>

35...Qa3 3 2 2 <2.3>

35...Qb4 4 4 3 <3.7>

True Rank: 1 = [ 35....Kg6, 35...Qxc2 ]; 2 = [ 35...Qa4 ], 3 = [ 35...Qb4 ]

Jun-24-17
Premium Chessgames Member
  diceman: <AylerKupp: <diceman> No, and I don't even know what "annotation" mode is.>

Basically generates an annotated game.

<And if I were to take you literally, how does it start at the end of the game and move forward since there are no moves after the end of the game?>

Well, since you start at the end. Forward is backward.

Maybe I should have bastardized digital language and come up with “Norward?” :)

<Maybe you're thinking of what I know as <backward sliding>. This involves starting the analysis at either the end of the game or whenever the engine correctly sees a certain win. Then you back up one move (2 plies) and start the analysis again.>

The way this works is, you set a time, and sensitivity. The time determines how long it will think per move, sensitivity sets how much it will annotate. 1 pawn sensitivity will probably only show major blunders, 1/10 pawn sensitivity
will show more alternative moves.

It takes the start of the game and looks at its
database. It adds relevant lines, and alternative lines. When it is out of the database it marks the
first new move “N.”
(no computer calculation yet)

It then jumps to the end of the game
and analyzes norward with the “N” new
move being the last one it looks at.
(you do have the option to turn off
the database, and it will analyze until
move 1)

It adds standard chess annotation !/? as
well as, “only move/better is” and so on.
It will display better variations, and so on.
It also has some canned commentary such
as “White is ruining his position.”

When a game ends in mate. You can see it
quickly jump norward until it cant calculate
a mate any more.
(it will only show analysis if there was a quicker mate)

<you set a time> This looks like an average.
If you set it to 30 seconds, it wont
break analysis, and move exactly at 30 seconds.

It seems to call turning points better.
I think because, when it is analyzing move 29, its already seen move 30 analyzed.

Jun-27-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Conclusion (part 1 of 3)

Well, after a lot of huffing and puffing and some wrong turns along the way, I've come to the conclusion that after 34.Bxa7, <34...g5 was the losing move>.


click for larger view

My reasoning is supported by Houdini 4, Komodo 10/11, and Stockfish 8 as follows:

1. <Stockfish> After 34.Bxa7, Stockfish's top 3 responses by Black were 34...Ke6, 34...Ke5, or 34...g6 and it could not find a winning position for White Fischer vs Larsen, 1971 (kibitz #239)). It evaluated the resulting positions at d=42 equal; at [0.00], [+0.21], and [+0.24] respectively. Yet after 34...g5 Stockfish at d=45 evaluated the position after 35.Bb6, the move that Fischer played, as winning for White, [+4.54] (Fischer vs Larsen, 1971 (kibitz #241)).

2. <Komodo> After 34.Bxa7, Komodo's top 4 responses by Black were 34...Ke6, 34...Ke5, 34....Ke7 (Komodo 10), and 34...g6 (Komodo 11). And they also could not find a winning position for White Fischer vs Larsen, 1971 (kibitz #239)). At d=30 they evaluated the resulting positions as follows:

Komodo 10: 34...Ke5 at [+0.70], 34...Ke6 at [+0.84], and 34...Ke7 at [+1.14].

Komodo 11: 34...Ke6 at [+0.25], 34...Ke5 at [+0.36], and 34...g6 at [+0.61]. Komodo 11 is rated somewhat higher than Komodo 10 and its evaluations are more in line with Stockfish's and Houdini's, so I think that they are probably a little bit more accurate.

Yet after 34...g5 Komodo 11 at d=31 (Fischer vs Larsen, 1971 (kibitz #241)) evaluated two of Black's top responses to 35.Bb6 as winning for White, 35....Kg6 and 35...Qa3, [+2.12] for both of them. Only for the move played by Larsen, 35...Qxc2, did it fail to find a win for White.

But sliding forward and starting the analysis after 34.Bxa7 g5 35.Bb6 Qxc2 36.a5 (the move that Fischer played in the game), Komodo 10 at d=32 evaluated the position as a win for White at [+2.61] (Fischer vs Larsen, 1971 (kibitz #232)). So, following the moves played in the game, 34...g5 was also considered by Komodo 10/11 to be the losing move.

Jun-27-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Conclusion (part 2 of 3)

3. <Houdini> Houdini had by far the hardest time showing that 34...g4 was the losing move. After 34.Bxa7 Houdini's top 3 responses at d=28 were the same as Stockfish's; 34...Ke6, 34...Ke5, and 34...g6, all indicated as equal at [0.00], [+0.03], [+0.16] respectively (Fischer vs Larsen, 1971 (kibitz #239)). But, unlike Komodo or Stockfish, it could not find winning positions for White after its top 3 responses to 35.Bb6; 35...Qxc2 (as played in the game), 35...Kg6, and 35...Qa3, evaluating the resulting positions at [+0.71], [+1.38], and [+1.64] respectively (Fischer vs Larsen, 1971 (kibitz #241)). So, for completeness, not only must 35...Qxc2 be eliminated as a saving move, but also 35...Kg6 and 35...Qa3.

35...Qa3 is easy to dispose of as a saving move. In Fischer vs Larsen, 1971 (kibitz #230) Black was destroyed after 36.Re1+, with Houdini at d=22 evaluating the resulting position at [+14.52].

35...Kg6 was not investigated earlier but, at least on the surface, there doesn't seem to be anything to recommend it. It took quite a while (more than 21 hours!) but at d=34, Houdini evaluated the resulting position as winning for White, [+2.07]. Here is Houdini's winning line: 36.a5 Qa3 37.Re1 f4 38.Rb1 Qa2 39.Be4+ Kf7 40.Rb5 Qa1+ 41.Bg1 g4 42.Rf5+ Ke7 43.Rxf4 Qxa5 44.Rxg4 h5 45.Rg7+ Ke8 46.Ra7 Qe1 47.Bd5 Qc1 48.Bc6+ Kf8 49.Rc7 c3 50.Rd7 h4 51.h3 Qb1 52.Be4 Qe1 53.Bd3 Qg3 54.Bf1 Qe1 55.Bg2 Qc1 56.Bd5 Qd1

In contrast, it took Komodo only a little over 4 minutes to find a line with a winning evaluation for White ( [+2.20] ) and it took Stockfish only a little over 13 minutes at d=32 to do the same ( [+2.68] ). Perhaps Houdini 5 can do better.

So back to 35...Qxc2. Houdini couldn't find a winning line for White as played in the game after the following:

(a) 36.a6, evaluating the resulting position at only [+0.43] (Fischer vs Larsen, 1971 (kibitz #232))

(b) 36.a6 Qb2 37.Bd8+, evaluating the resulting position at only [+0.33] (Fischer vs Larsen, 1971 (kibitz #234)) (although it did consider that White would be winning after 37.Be3, evaluating the resulting position at [+2.65], so maybe Fischer missed a shorter win.)

(c) 36.a6 Qb2 37.Bd8+ Ke6 (not posted), evaluating at d=29 the resulting position at only [+0.77].

(d) After 36.a6 Qb2 37.Bd8+ Ke6 38.a6 Qa3 (also not posted) it DID evaluate the resulting position as winning for White, [+7.83] at d=30. But it did not consider the position as winning for White after 38...Qf2, evaluating the resulting position at only [+0.41].

So, would 36.a6 Qb2 37.Bd8+ Ke6 38.a6 Qf2 have been a saving line for Larsen? No, none of the 3 engines considered it as such, evaluating the resulting positions as winning for White, [+4.28] at d=30 (Houdini), [+3.01] at d=36 (Komodo 10), and [+9.34], d=49 (Stockfish) (Fischer vs Larsen, 1971 (kibitz #231)), although it would have required Fischer to find 39.Rb1 instead of playing 39.Bb7 as he played after 38...Qa3. All 3 engines considered the positions after 38.a6 Qf2 39.Bb7 as equal; [0.00] (Houdini), [+0.34] (Komodo 10), and [0.00] (Stockfish).

Jun-27-17
Premium Chessgames Member
  AylerKupp: <Howard> The Losing Move – Conclusion (part 3 of 3)

After 32.Bxc6 Qxc3 all 3 engines all 3 engines evaluated the position as OK for Black; [0.00] at d=28 (Houdini), [+0.53] at d=27 (Komodo 10), and [0.00] at d=40 (Stockfish). So it doesn't seem like 32...Qxc3 was the losing move and this was, after all, the original question. The losing move must therefore have come later.

Instead, it looks to me that 34...g5 was indeed the losing move. At least according to Komodo 10/11 and, particularly, Stockfish 8. Houdini 4 had a much greater difficulty arriving at that conclusion, and even that can be considered conditional on White finding 39.Rb1 after 34...g5 35.Bb6 Qxc2 36.a6 Qb2 37.Bd8+ Ke6 38.a6 Qf2.

Why did Houdini 4 have such difficulty compared to Komodo and Stockfish? I don't know. But in these positions Houdini took the longest time to reach a given search depth, and apparently the positions after 34...g5 required in-depth searching. Maybe Houdini 5 would do better; and if I get it (another story), I might try it and see.

I should also try <diceman>'s suggestion and use backward sliding and see if that results in the same conclusion, only to kick myself if it does with a lot less time and effort I spent on this analysis.

But the most important question is: Why was 34...g5 the losing move? Again I don’t really know. But this was a very sharp conclusion to this game and maybe 34...g5 represented a loss of tempo that Black could not afford, even though in some line Black was able to put up a struggle by later advancing both his f- and g-pawns.

And 34...g5 looks good on the surface. It not only provides support for Black's passed f-pawn if it advances but it also provides Black's king another escape square at g6. Perhaps someone else (not me!) might want to investigate the alternatives that these 3 engines suggested after 34.Bxa7; 34...Ke6, 34...Ke5, 34...g6, or even the immediate 34...f4.

Finally I think we all need to remember that these moves were all played towards the end of the first time control when both players were likely short of time. And the fact that they each found almost all the only moves to either win or prolong the game under these conditions is simply amazing.

Jun-27-17
Premium Chessgames Member
  AylerKupp: <diceman> This is clearly a feature of your GUI. Which GUI do you use? FWIW, I use Arena 3.5 which does a reasonably good job as a GUI and has lots of features, but nothing like this as far as I know. It also has the great advantage of being free, which was a big consideration when I first got started with computer-assisted chess since I wasn't sure whether it would be something that would keep my interest so I didn't want to make a financial investment on it. Now I have the investment of time and learning that I put into Arena so I am reluctant to change GUIs.

With Arena, if you want to do backward sliding, you have to do it manually yourself (or maybe it has a feature I've never uncovered) and, as I said before, I've never tried it. But maybe I will.

The process you describe seems like "my" backward sliding, particularly your comment that when it's analyzing move 29 it has already seen move 30 analyzed. It seems to have a few useful additional bells and whistles like adding the standard chess annotations like , etc. I wonder what evaluation is required before it gives it an annotation of ++ or --.

But you make a reference to a database that it looks at before any computer analysis is done. How does that database get populated? I assume that it comes with your GUI, you can populate it using a commercial database, or it's done by previous analyses you've conducted, or a combination of the three.

Maybe, if you have the time and the interest, you'd like to try it with this game. Then we'll see if it agrees with my conclusion about 34...g5 being the losing move.

I liked your "Norward" and your "Forward is backward". The first one I think is a portmanteau, in which part of multiple words are combined to form a new word. The second one sounds 1984ish; e.g.

War is Peace
Freedom is Slavery
Ignorance is Strength

Jun-27-17  Howard: That's far more analysis than I'd probably have time to go through, but (trust me !) I truly appreciate it nonetheless !!!
Jun-28-17
Premium Chessgames Member
  AylerKupp: <Howard> I doubt that anyone would have the time, or the interest, to go through all the analysis I posted. It was one of those situations where, once I got into it, I couldn't let go; my pig-headedness got the better of me. And it took a lot more effort and time that I originally expected, partly due to my getting sidetracked. But I did learn one valuable lesson; if you want to make sure that you reach your goal, you need to be systematic about it.

If you only have time to read one post, read my conclusion above (Fischer vs Larsen, 1971 (kibitz #243)). After all, you were the reason that I spent all this time an effort so requiring you to read those three posts seems to be fitting "punishment" :-). Besides, I would be interested to find out if you agree with my reasoning.

Other "fun" analyses are Fischer vs Larsen, 1971 (kibitz #230) which shows the demolition of Black's position of 38...Qa3 by 39.Re1+ that Fischer apparently missed (although I can't really quibble with his 39.Bb7 which caused Larsen to resign 2 moves later)

And this is one analysis after 37.Bd8+ Ke6 38.a6 Qf2 (instead of Larsen's 38...Qa3) evaluated by Houdini at [+0.41], d=30 that I didn't post, and is perhaps noteworthy for (a) a remarkable find by Houdini for a possible saving resource for Black in a sideline and (b) an incredible miss by Houdini of a winning sequence for <Black> later in the same sideline which I saw immediately and Houdini later confirmed with some forward sliding:

1. [+0.41], d=30: 38...Qf2 39.Bxg5 c3 40.Ra1 f4 41.Bh4(a) Qc5 42.Bf3 c2 43.a7 Qxa7 44.Rxa7 c1Q+ 45.Kg2 Qc2+ 46.Bf2 Qg6+ 47.Kh1 Qb1+ 48.Bg1 Qd3 49.Ba8 Kf6 50.Ra2 Qb1 51.Rf2 Kg5 52.Bf3 Qa1 53.Kg2 Qg7 54.Rc2 Kf5+ 55.Kh1 Qa1 56.h3 Qe1 57.Rf2 Kg5 58.Kh2 Qc3 59.Kg2 Kh6 60.Kh1 Kg5 61.Kg2 Kh6 and it seems that Houdini is headed towards a draw by repetition since it can't figure out how White can make progress.


click for larger view

(a) An attempt at deflection since after 41...Qxh4 42.a7 White's a-pawn queens and survives and Black's c-pawn doesn't. 41.a7 might seem simpler but after 42...f3!! (exclamations mine, of course) Houdini finds at least a perpetual check in the open board: 43. Bxf3 (forced, otherwise after 43.Rg1 White is immobilized and cannot force the queening of his a-pawn) 44...Qxf3+ 43.Kg1 Qg4+ 44.Kf2 Qf5+ 45.Ke2 Qe5+ 46.Be3 Qh5+ 47.Kd3 Qb5+ 48.Kxc3 Qc6+ (a big surprise, Houdini misses 48...Qe5+ followed by 49...Qxa1 and <Black> wins! Restarting the analysis from the position after 48...Qe5+ Houdini confirms this!) 49.Kd3 Qb5+ 50.Kc2 Qc4+ 51.Kd2 Qb4+ 52.Kc2 Qc4+ 53.Kd2 etc.

This once again shows how essential it is to check each and every computer line for possible misevaluations.

Jump to page #    (enter # from 1 to 10)
search thread:   
< Earlier Kibitzing  · PAGE 10 OF 10 ·  Later Kibitzing>

Now on DVD
NOTE: You need to pick a username and password to post a reply. Getting your account takes less than a minute, totally anonymous, and 100% free--plus, it entitles you to features otherwise unavailable. Pick your username now and join the chessgames community!
If you already have an account, you should login now.
Please observe our posting guidelines:
  1. No obscene, racist, sexist, or profane language.
  2. No spamming, advertising, or duplicating posts.
  3. No personal attacks against other members.
  4. Nothing in violation of United States law.
  5. No posting personal information of members.
Blow the Whistle See something that violates our rules? Blow the whistle and inform an administrator.


NOTE: Keep all discussion on the topic of this page. This forum is for this specific game and nothing else. If you want to discuss chess in general, or this site, you might try the Kibitzer's Café.
Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors.
Spot an error? Please submit a correction slip and help us eliminate database mistakes!
<This page contains Editor Notes. Click here to read them.>
This game is type: CLASSICAL (Disagree? Please submit a correction slip.)

Featured in the Following Game Collections [what is this?]
knightfly's favorite games
by knightfly
olisibaraw's favorite games (E4)
by olisibaraw
Queen sacrifices 4
by obrit
Game 8
from Starting out : The French by diegoami
Robert Fischer's Best Games
by Jaredfchess
+6 -0 =0 vs. Larsen (Candidates semifinal)
from Match Fischer! by amadeus
wrimle's favorite games
by wrimle
French Defense: Winawer. Positional Variation
from GAMES OF THE DAY by gambitfan
It will work out?
from Concentration of Force by sevenseaman
Match Play
by frogmanjones
Travis Bickle's favorite games
by Travis Bickle
R+2B vs. Q.
from Bobby Fischer's Best Endgames. by Nasruddin Hodja
French Defense: Winawer. Positional Variation
from WINAWER by gambitfan
Memorized Games
by 000129059
akapovsky's favorite games
by akapovsky
Winawer. Advance Variation
from MKD's French Defense by MKD
FR Winawer. Advance (C19) 1-0 BF has the Bishop Pair
from Game collection: -ER by fredthebear
French
by ALL
Larsen bent by Fischer
from Greatest Games of the Century - Chessbase by lucifershammer
ghostchild's favorite games
by ghostchild
plus 139 more collections (not shown)


home | about | login | logout | F.A.Q. | your profile | preferences | Premium Membership | Kibitzer's Café | Biographer's Bistro | new kibitzing | chessforums | Tournament Index | Player Directory | World Chess Championships | Opening Explorer | Guess the Move | Game Collections | ChessBookie Game | Chessgames Challenge | Store | privacy notice | advertising | contact us
Copyright 2001-2017, Chessgames Services LLC