Carlsen was 8 years old (4 months before his 9th birthday I guess) when playing this game, which means it was right after he started getting interested in chess for real. As we know, he learned the game a few years earlier.

Of his games with normal thinking time, I know of nine games who are missing (there may be a few more missing club games). These are:

Skei Grand Prix Group B (Gausdal), 30 July - 1 August 1999:

1) <Unknown player 0> (not Øhman, Odd Hansen or Linda Hensen)

Arnold Grand Prix (Gausdal), 10-14 January 2000:

2) <Santa Rause 1> (not asked)

3) <Anders Bekker-Jensen 1> (ABJ does not have the game)

NTG Grand Prix Group B, 4-6 February 2000:

4) <Oddbjørn Christensen (NOR 992) - Carlsen ½-½> (OC does not have it)

ASKO Pinse Tournament Group B, 10-12 June 2000:

5) <Carlsen (904) - Jo Vederhus (NOR 1304) 1-0> (Vederhus' does not have it)

Asker SK Club Tournament Group A (4), 2 (?) October 2000:

6) <Carlsen (1323) - Vegard Hole (NOR 1442) 1-0> (Hole does not have it)

Asker SK Club Tournament Group A (6), 23 (?) October 2000:

7) <Carlsen (1323) - Neil Benn (NOR 1574) 1-0> (Benn: probably lost to the annuls of history)

Norwegian Individual Championships Group Junior (Kristiansund), 8-14 July 2001:

8) <Carlsen (2084) - Torill Skytte (NOR 1661) 0-1> (TS does not have the game)

9) <Carlsen (2084) - Lars Gunnesdal (NOR 1889) ½-½> (LG does not have it)

1. d4 e5

<Kids, do not try this at home: the Englund is unsound.>

2. dxe5 Nc6 3. Bf4 f6

<Elsner vs E J Diemer, 1987 saw 3...g5.>

4. exf6

<A tiny bit greedy, developing with 4. Nf3 is better. 4...g5 5. Bg3 g4 6. Nd4 fxe5 7. Nxc6 bxc6 8. h3 is a possible continuation.>

4...Qxf6

<4...Nxf6 makes more sense, as it develops. White is objectively better, Black's opening is a full-blast disaster.>

5. Qd2??!?

<Nonsense. Either 5. Qc1 or 5. e3 are better. The first one will not lose a pawn, the second one helps the much-needed development. Maybe even 5. g3 can be thought of.>

5...Qxb2 6. Qe3+ Nge7??

<6...Be7, certainly, with the idea of putting it on f6.>

7. Be5?? Qxc2??

<Why? Just why? The bishop can be taken.>

8. Bxc7 Nb4?!

<A weird move. Maybe something like 8...d5 is better, developing while White struggles with the total lack of coordination.>

9. Nc3 Na6?

<9...Qb2 10. Rc1 Nxa2 11. Nxa2 Qxa2 with a later b6 is straightforward and obvious, where Black is not worse at all.>

10. Rc1 Qb2 11. Qe5??

<I watched the position after 11. Rb1 for a while, and could not decide if White is simply better or much better. So I asked the electronic monster to crunch it for a couple of hours. At depth 55 it goes for 11...Qc2 12. Bf4 h6 13. Nf3 Qg6 14. g3 Qe6 15. Qd2 Ng6 16. Nd4 Qf7 17. Bg2 Nxf4 18. gxf4 Bb4 19. Rxb4! Nxb4 20. Ncb5, and Black's underdevelopment is very bad, but not 100% lost yet. DISCLAIMER: trusting an engine output without any need of criticism is a very bad habit. Maybe 11. Bd6 or 11. Bf4 are also playable.>

11...Qxc1+

<:D>

12. Nd1 Nxc7 13. e4 Kd8 14. Nf3 Ng6 15. Qg5+ Qxg5 16. Nxg5 Bb4+ 17. Ke2 Nf4+

<Exploiting the king's horrific position with 17...Re8 is precise. Black is winning, anyways.>

18. Kf3 Rf8 19. Nxh7 Nfd5+

<Simplifying to a won late-middlegame, by sacrificing the exchange. Clever decision by Randjelovic!>

20. Nxf8 Bxf8 21. exd5 Nxd5 22. Bc4 Nb6 23. Bb5

<To be honest, I cannot see the reason of this move. White's position is already resignable, and neither d7 is weak, nor the bishop pins or X-rays anything there.>

23...a6 24. Bd3 Nd5 25. Re1 Nb4 26. Bg6 Kc7 27. Re5 Bd6

<Simply take that forgotten pawn on a2, and it is curtains.>

28. Re4

<28. Re2 will not save the game, as Black can develop that ugly bishop and rook without much interference.>

28...Nxa2 29. Ne3 Bxh2

<29...b5 is the most obvious and straightforward, of course.>

30. Rc4+

<30. Ra4 b5! 31. Rxa2 Bb7+ and 30. Bf7 b5! 31. Bxa2 Bb7 32. Nd5+ Kd6 33. Rd4 Re8 both look clever, but they are equally winning for Black.>

30...Kb8 31. Nd5

<I tried to make 31. g3-ideas work, without success. White is much worse.>

31...b5 32. Rc5

<The absolutely very last hope for some play is 32. Rxc8+ Kxc8 33. Bb1. Black is still winning by a wide margin, but there is a minimal chance that in time trouble the full point will somehow get blundered away.>

32...Bb7 33. Bf5 Bd6 34. Rc2 Bxd5+ 35. Be4 Bxe4+ 36. Kxe4 Nb4 37. Rd2 Be7 38. Rxd7

<Snatching a pawn!>

38...Bc5 39. Rxg7

<Snatching a pawn!>

39...a5 40. Rg5 Be7 41. Rxb5+

<Snatching a pawn! Giving a check!>

41...Kc8 42. Re5 Ra7 43. Rh5 a4 44. Rh1 a3 45. Ra1 Nc2 46. Ra2 Ra4+ 47. Kd3 Nb4+ 48. Kc3 Nxa2+ 49. Kb3 Rf4

<Quicker is 49...Nb4, but White is dead lost anyways.>

50. Kxa2 Rxf2+ 51. Ka1 Rxg2 52. Kb1 a2+ 53. Ka1 Bf6#

<Ouch! :)>

There's Marshall's <I shan't need the rook> game, is that what you're thinking of?

J Esser vs Marshall, 1911

I'm not familiar with the Tal game, but it certainly sounds in character. Of course there's this recent example of Carlsen rallying to win after blundering a piece.

This is the first annotated game in Tal's autobiography.

Tal vs J I Zilber, 1949

On move 12, Tal makes an unsound (but deliberate) piece sacrifice. After 12....Qf4+ 13.Kb1 gxf6, Tal wrote <Black has an extra bishop, for which I now do not see any compensation....White continues as if nothing has happened.>

It could be this one. (there are so many to choose from!)

Smyslov vs Tal, 1959

Tal as Black plays on from here.

click for larger view

Smyslov is so furious he loses the plot and at one point slams the clock so hard it tipped over some of the pieces. He walked into a perpetual check. An amazing save by Tal (who was also in severe time trouble) in a position many players would just resign.