Maybe I must describe it in another way. Of course a mistake is necessary to decide a game. So for the best game of all time the mistake must be a very normal, logical, positionally founded move - like ...d5 in the Steinitz-Von Bardeleben game. To be the tactically best game of all time, the refutation must be hard to find, involve a myriad of sidelines. The loser must play the most stubborn defence, which gives the winner most chances to fail. Still the winner doesn't and brings the point home. Afterwards all analysts must break their teeth in trying to find unexpected salvations and shorter, easier wins.

This does not exclude the fact, that Tal ao played many brilliant games. But if his sacrifice(s) can be refuted, then his game(s) cannot be the best game of all time. This is also true for many Spielmann games.
Take for instance Spielmann-Hönlinger. Black made a positional mistake, where these days many strong amateurs would have found the right, equalizing continuation. Spielmann's refutation was absolutely brilliant, but Black's fault was too obvious. Again: not the best game of all time.

In previous posts I have mentioned a couple of games, which match these standards.

Regarding that Danish Gambit game: I don't think it is a very good game. But I challenge you to find one single other game, in which the winner gave smothered mate to a king in the centre with his only remaining knight.

Mnb. Thx for reminding me of many a fine game from Spielman. Though I do recognize quite a few from Spielman's memorable book 'The art of sacrifice in chess'.
Your example game is quite nice. But from move 14 or 15 on it's not really difficult to calculate:
the smothering mate-motiv is well know and forced. It was already a standard combinational motive at the end of the nineteenth century (though it commonly occurs mirrored with a King on h8). So while you're entitled to your own opinion (that Kasparov, Tall and Shirov never played a combination to match this game) I do not share it.
I still challenge you to define the best/strongest defence. More so because of your comment 
"Who can blame Lasker for playing ...h5?"

Applying your criteria, excludes many older games from 'the competition' because of the progress in opening-theory. If new methods are found to play for an advantage in a variations all the older games in such a variation are by definition rendered incorrect. This might mean that we must regard all games with 1.e4,e5 2.f4,exf 3.Nf3 as incorrect because white 'overlooked' black's strong reply 3...g5! 

Also thx everyone to list some memorable drawn games. Spasski-Fischer (19th game) is indeed memorable. If only for Fischer finding the only saving move Qd2!! over the board.

MNb wrote on 02/24/06 at 02:49:59:

I would be very curious what Rolf Schwartz has to say about it

@Willempie
I am more than willing to prove you, that both Steinitz and Von Bardeleben made two mistakes in their famous and beautiful game - but in a separate thread. I am looking forward to another debate with you. 

@Micawber and evt others:

By no means I want to diminish the beauty of all those Tal games, nor of Steinitz-Von Bardeleben - I maintain, that there were three errors in the critical phase of that game. Note, that I do not blame VB for playing 7...d5.

But the title of this thread is "The Best Game of all time". Such a game should fulfill my criteria imo, when it comes to tactics. If Steinitz had played 15.Qa4+ Kf7 16.Ne5+, he would have had my vote.
The same for Rosanes-Anderssen, Breslau 1863: a very beautiful game, GM's of the 21st century could be proud of playing a game like this. Still 10...bxc6 was an error and should have been replaced by 10...Nxc6.

So these games are not better than say
Spielmann-Eljasjov, München 1903
Spielmann-Nimzovitsj, München 1905 (the 19 moves game)
Spielmann-Janovsky, Karlsbad 1907
Mieses-Spielmann, Regensburg 1910 (3rd game)
Spielmann-Rubinstein, Karlsbad 1911
Spielmann-Réti, Abbazia 1912
Rubinstein-Spielmann, San Sebastian 1912
Rubinstein-Spielmann, Bad Pistyan 1912
Spielmann-Moller, Goteborg 1920
Spielmann-Grünfeld, Teplice Sanov 1922
Kmoch-Spielmann, Semmering 1926
Tartakower-Spielmann, Semmering 1926
Spielmann-Nimzovitsj, Semmering 1926
Spielmann-Hönlinger, Vienna 1929
Spielmann-Grünfeld, Karlsbad 1929
Spielmann-Capablanca, Karlsbad 1929
Pirc-Spielmann, Rogatska Slatina 1931
Spielmann-Bogoljubow, Semmering 1932 (9)
Gereben-Spielmann, Sopron 1934
Sjoholm-Spielmann, Kalmar 1941.
All very beautiful games, but not the best game ever according to my standards.

For Micawber I have two fascinating draws:
Spielmann-Capablanca, Moscow 1925
Spielmann-Lasker, Moscow 1935.

Here is another golden oldie: Pillsbury-Lasker, Nürnberg 1896. Who can blame Black for playing 18...h5? The punishment afterwards is 100% correct and needed no less than 4 sacrifices. Moreover Lasker did not err anymore.

RotGut wrote on 02/23/06 at 16:59:20:


I generally agree with this view. If a sac,even if its wrong, is sufficient to give the opponent huge problems otb its 'correct' enough in my book at least. Aftrerall that's what chess is all abt in the 1st place: creating problems & difficulties for your opponent. 

The Tal-Botvinnik game quoted by micawber was a good example of this. That unsound sac was sufficient to trouble probably the most logical & scientific player ever. With the clock ticking, to find an adequate defense or refutation is next to impossible. Hence most of these are found in intense post mortem analysis. And of course the most impt thing was Tal won.

But of course many other aggressors wouldnt sac like Tal (save maybe Topalov). Fischer's & Spassky's sacs generally hv some kind of conclusive basis, whereas Tal's quite often leaves an unclear position. So Tal relies more on intuition rather than concrete assessment of the final position at his event horizon so to speak. More often than not he has his opponent's clock as his ally. But still, his games are a joy to behold & very interesting nonetheless.

Anyway, Tal did say "There are 2 types of sacrifices. Correct ones and mine"


"If correctness is an issue, than the correct choice would be a drawn game where attack and defence held the balance. However no drawn game has been suggested in this thread!" - Try that Spassky-Fischer 1972 game (cant remember which no.)where Spassky saced 2 minor pieces for initiative, & Fischer counter sacs with a whole rook. All action being played in a space of 5 moves by each side. After the storm, the game eventually ended in a peaceful draw.
No outright blunders by both players here.

later edit: it was game 19. The 2nd Alekhine game of the c'ship.

Willempie,
Thx for your clarification and additional info on von Barleben.

I am not sure that I agree with one of MNB's criteria: The strongest defence.
I will give an example. Suppose the loser had at one point the choice:
a) a relative simple variation that resulted in a lost but technically dificult endgame.
b) a complex variation where white should play several difficult unique moves
that resulted in a clearly won position.
Which of the two is the strongest defence? Which of the two is the best practical defence?
The two dont necesserily coincide.

Another point about correctness. In the Geller-Euwe game I choose, white could have played a
stronger defence (though black would still have good winning chances). Euwe could have eliminated
this variation with an intermediair move. Bronstein relates that it took several grandmasters a week! to find this possibility and adds that Geller had no chance at all to find it "over the board". He also puts the question if this diminishes the beauty of the game. He answered the question himself; he didnt think so.
So I think I am still in good company with my game of choice. [even if it doesnt match your criteria exactly].

The criteria should eliminate quite a few Tall games. But he played several games that reside in our
collective memory. These games could easily qualify as best game. An example: the spectacular piece sac Nf4!?, against Botwinnik in their WC-match. 
Eventually this sac was proved incorrect. But it took great effort to find the refutation. Again it was hardly possible to do so in OTB. The fact that there was a refutation hardly diminishes the pleasure to replay this game and read Tall's honest anotations. Perhaps we've grown more critical now that everyone with a powerfull computer and  fritz can find out in 1 night, what took several GM's days to establish.    But not only correctness but risk and entertainment-value should be considered as well.
If correctness is an issue, than the correct choice would be a drawn game where attack and defence held the balance. However no drawn game has been suggested in this thread!

MNb wrote on 02/23/06 at 02:48:47:


I guess most of Tal's games do not meet your requirements. Quite often his combinations are much later proven wrong. That being said, I'm pretty sure there are better & simpler ways to maintain his initiative (though I cant ind any yet) & his opponents always go very wrong when meeting one of his sacs (most probably due to sheer shock), neither do they find good moves later in the game due to time pressure.

But as 1 author wrote , something like "tho the sac was proven incorrect it hardly detracts the play of Tal, and the most impt thing was, he won"

Surely the best game is a titanic struggle between 2 players at the peak of the game.

No one would argue, for example, that a first-round knockout is the BEST boxing match ever-- regardless of how brilliant the victor performed.

Therefore i am shocked to see so many lopsided and <30 move examples in this thread.  The Kramnik-Anand game is a beauty because it is a monstrous struggle, and both players are top 5 or better in the world.

Another good choice no one has mentioned is Capa - Marshall NY 1918.  It has an interesting story and features the world debut of the Marshall Attack.