Stockfish 16 says:

50. Qxb8

(50. Qxb8 (1.Qxb8 Rxb8 2.Rxb8 Kc7 3.Nxe6+ Kxb8 4.Nd4 Kc7 5.Kg3 Kd7 6.c3 Kd8 7.Nxc6+ Kd7 8.Nb4 Kc7 9.h4 Kd7 10.c6+ Kc8 11.Kf2 d4 12.cxd4 Kc7 13.d5 Bxd5 14.Nxa6+ Kc8 15.Nb4 Be4 16.Ke3 Bxg2 17.Kd4 Bf1 18.Kc5 Kc7 19.Nd5+ Kc8 20.Kd6 Kb8 21.Nb4 Bh3 22.Ke7 Be6 23.a6 Ka7 24.Kd8 Bg4 25.c7 Kb6 26.c8R Bxc8 27.Kxc8 Kc5 28.a7) +7.45/47 776)

The unlucky Janowski fell victim to a comparable idea a second time (37.b7) in this game:-

It looks at first glance that the Black Q can escape with 51...Qe7 but of course 52.Rb7+ skewers her. Maybe that's what Janowski missed when he erred with 47...Qe6?.

Mason was always a problem for Janowski. At the close of Mason's career, Janowski was riding high in tournament play while Mason was usually towards the bottom of the standings. But after splitting their results earlier in their respective careers (two wins each with two draws in six games), Mason won their final three encounters, including here at Monte Carlo 1902 (where Janowski finished first and would--all other things being equal--have won the tournament had he managed to win this game) and later that same year at Hanover 1902 (where Janowski took first and lost only to Mason). Perhaps Mason, who reportedly had problems remaining sober, managed to pull himself together when he played the often obnoxious Janowski.

1. e4 e5
2. Nf3 Nc6
3. Bb5 a6
4. Ba4 Nf6
5. 0-0 Be7
6. Nc3

This completely sound but modest effort was popular at the time. It has, of course, since been eclipsed in popularity by 6. Re1.

6... d6

"A reserved reply." (Tartakower/Dumont).

The more usual 6...b5 is better. The text leaves"

click for larger view

7. BxN+

Probably best. White obtains no edge after 7. d4 b5 8. dxe5 [and White is actually worse after 8. Bb3 Nxd4 9. NxN exN 10. Bd5 (and certainly not 10. Qxd4? c5 losing a piece!].

7... bxB
8. d4 exd4

Tartakower/Dumont claimed this needlessly abandoned the center, but if 8...Nd7 (as they suggest) White is even better off than in the game with 9. dxd5 dxe5 10. Qe2.

9. Nxd4 Bd7

click for larger view

10. b3

While, as Tartakower/Dumont correctly note, no "direct enterprise would lead to a concrete result," the text has little to recommend it other than being sufficient for equality. Probably best for White here were 10. Bg5 or 10. Re1 or maybe 10. Bf4 leading in each case to a tiny edge for White.

10... Qb8

A strange effort by Janowski (called "original" by the Tournament Book and "a highly artificial maneuver" by Tartakower/Dumont. Simplest and best was 10...0-0 (called "obvious" by Tartakower/Dumont). The rationale for this move by Janowski was revealed two moves later by his 12...0-0-0?!

11. Bb2

A logical continuation. If Mason wanted to try to exploit Janowski's last move, he might instead have continued 11. Ba3

11... Qb7

Continuing his dubious plan.

12. Re1 0-0-0?!

"...the originality of this plan does not signify its soundness." (Tartakower/Dumont). As this source goes on tos say, 12...0-0 was surely sounder and better.

The text left:

click for larger view

The best I can say for 12...0-0-0 is that it tempted Mason to overplay his hand and potentially relinquish the advantage he now enjoyed. But Janowski in turn overplayed HIS hand and soon managed to lose a pawn and eventually the game.

13. Qd3

Perhaps 13. Qf3 or 13. Nf3 or maybe 13. Nf5 would have presented more challenging problems for Janowski, but given what soon ensued, this is likely all moot.

13... Rhe8

"?"--(Tournament Book)

"In order to continue 13...c5 14. Nf5 Bf8." (Tartakower/Dumont).

But the above variation would fun into 15. Nd5 with an overwhelming game for White. 13...c5 is a reasonable line, but Mason would probably have responded 14. Nde2 or 14. Nf3 with the better game. And after 13...c5 14. Nf5, even so great a lover of Bishops as Janowski would surely have played the far better 14...BxN with a difficult but tenable game.

Interestingly, nobody mentions Black's best here: 13...Ng4 so that if then 14. Nf5 Black can respond 14...Bf6 15. Na4 BxB 16. NxB g6 with some chances to hold his own.

The text (13...Rhe8), even if not absolutely best, doesn't look all that bad to me. Perhaps more importantly, it goaded Mason into:

14. b4?!

click for larger view

"!"--(Tournament Book)

"An important move" (Tournament Book)

"With this pretty offer of a pawn White insists on his rights to the initiative on the Queen's side." (Tartakower/DuMont)

I must dissent. Mason got away with the text only because Janowski, for once, responded passively instead of adopting the aggressive 14...c5, which seems to turn the tables in favor of Black (the contrary commentary notwithstanding).

Instead, Mason could have retained his edge with the stolid but effective 14. Nf3.

14... Bf8?

Incredible from Janowski. Hard to believe that Janowski, of all people, shunned the obvious aggressive 14...c5! Tartakower/Dumont called this "too dangerous," but after either (a) 15. Nf3 cxb4; or (b) 15. Nb3 cxd4 16. Nd5 NxN 17. exN Qb5; or (c) 15. bxc5? dxc5 16. Nf3 [or 16. Nb3 Be6 17. Nd5 (surely not the recommended 17. Qf3? c4!) NxN 18. exN Rxd5] Be6 Black is better--and in some cases close to winning--in all variations.

After the wimpy text, Janowski should have been in trouble.

15. Rab1

15. a3 is also good for White.

15... d5!

"Trying to open up his game." (Tartakover/Dumont).

If instead 15...c5, then White is better after 16. bxc5 dxc5 and now 17. Ba1! instead of 17. Ba3 as given by Tartakower/Dumont, which could have run into 17...Qa8 (much stronger than 17...Qa7 as given in Tartakower/Dumont).

16. e5!

Seizing the initiative.

16... Ng4

Anything else would court disaster, and leaving:

click for larger view

Here Mason should surely have played the very strong 17. f4! But Mason missed this chance and instead played:

17. Nf3?

Now Janowski would have been very much back in the game with 17...f6, but instead made what could have been a losing blunder:

17... g6?

click for larger view

18. h3

Of course.

18... Bf5

As good or bad as anything (I examined 18...Nxf2?! but Black likewise seems to be in trouble in this line).

After the text, the position was:

click for larger view

19. Qd2?

This should have proved a useless move. If instead 19. Qf1! Mason's Queen would have remained on the f1..c4 diagonal and thus left him able to fend off the dangerous counter-play otherwise available to Black, e.g., 19. Qf1! d4 20. hxN dxN 21. gxB cxB and now, with most of the minor pieces off the board, White is in control, e.g., 22. f6! c5 (probably the best try) 23. Qc3 [available because of 19. Qf1] and now if 23...Qxb4 24. Qxf7 is murder and if instead 23...Rd5 24. a3 cxb4 25. axb4 and now if 25...Bxb4 26. Rxb2 after which Black's goose is cooked.

But after Mason's weak 19. Qd2?, the position was:

click for larger view

Now, with 19...d4! (not mentioned by any commentators I have read), Janowski would have been very much in the game. But instead, Janowski played the very weak:

19... Nh6?

It is true that 19...Bh6?, discussed by Tartakower/Dumont would have been even worse, but not because of their suggested 20. Qd1? because Black then is basically OK with 20...d4! 21. Ne2 Nxe5 22. NxN Bg7, but in light of the crushing 20. Qd4! which leaves Black no decent recourse.

20. a3

20. Ne2 or 20. Ba1 would also have been strong here.

The text left:

click for larger view

Janowski was probably already lost here. His best chance lay in 20...Be6. But here the usually aggressive Janowski decided to reshuffle his minor pieces with:

20... Ng8?

This was much too slow, and Mason took due advantage with:

21. Nd4

21. Na4 immediately, eyeing c5,was perhaps even stronger. But Mason's move was also excellent.

21... Be6
22. Na4

click for larger view

22... Nh6

Continuing his bizarre Knight antics. I'm not saying it's best, but from Janowski I would have expected something such as 22...h5 (which, among other things, would anticipate a possible g4 by White).

23. Qc3

Another nothing move. The Tournament Book claimed that 23. Nb3 (heading for c5) was best, but this would still leave Janowski some play after 23...Nf5 24. Nac5 BxN 25. NxB Qb5.

But far stronger for White here was the indicated 23. g4! which would reduce Black to hopeless reshuffling. But the text led to an exciting turn of events, the position now being:

click for larger view

23... Nf5!

"Voluntarily giving up the pawn, because after 23...Bd7 24. Nb3 White conquers the dominating square c5." (Tartakower/Dumont)

True, but Black now has some counter-play with either 24...Qb5 or 24...Nf5, whereas White could instead have answered 23...Bd7 with 24. g4!

After Janowski's much superior 23...Nf5!, the board exploded with all sorts of tactical possibilities:

24. Nxc6

click for larger view

24... Rd7

<vonKrolock> has made the key point. Janowski's move--seconded by the computers (Stockfish and Fritz) is the "safe" move which does not immediately lose more material. But in a game between humans, I much prefer 24...d4?! True, this involves the sacrifice of the exchange, but Black is no longer a sitting duck. A possible line might be 24...d4 25. NxR RxN 26. Qd3 Be7 27. f3 h5. Is this a theoretically losing line for Black? Probably. But playing Black I would prefer that to a slow painful death. The best thing I can say about Janowski's move is that Mason did not play the strongest lines from here (neither did Janowski).

And given the upcoming questionable moves by both sides, it appears that the approaching move-30 time control may have become a relevant factor.

25. Nd4 NxN
26. QxN Bf5

click for larger view

27. Rbc1

Was Mason afraid to create complications with the time control looming? If not, it is difficult to see why he didn't try the seemingly obvious and indicated 27. Nc5. The text was not a blunder, but it did little to advance Mason's chances. Another indication that, over the board, 24...d4 may have been a much better try for Black.

27... Qc6

Janowski may also have had clock troubles, since 27...Qb5 appears more active and providing better chances for counter-play. The text left:

click for larger view

28. Nc3?

Again by-passing 28. Nc5 for the tepid text. With Mason and Janowski, one can never be sure what is going on. But the text appear hopelessly passive in a position in which White--absent time control--should be thinking of ways to blow Black off the board.

28... h5

Not fully taking advantage of Mason's lapses. Perhaps 29...Bg7 or 28...Bh6 or even 28...Qb6 should have been considered.

29. Ne2?

Was this why Mason played 28. Nc3? Did he really think supporting f4 was a better plan than attacking on the Queen-side with Nc5. Or was this all just time-pressure nonsense?

29... Bh6!

Suddenly, Janowski's beloved two Bishops spring into action.

30. f4

If this was truly Mason's grand plan I may have misjudged his talent. Among other things, Mason has now permitted Janowski to get the Queens off the board with 30...Qb6. In any case, Mason's move left:

click for larger view

30... Bf8?

Just awful. With 30...Qb6, the Queens come off and Black with his two Bishops would have more than a prayer of a chance of exploiting his two Bishops. Although Mason nearly botched the win with his horrific 41st and 44th moves (when the move-45 time control was now in sight), it does not appear that Black has much of a chance after the very weak text move.

31. Qf2

Now for a while, Mason resumed his patient strong play.

There is, as pointed out on this site by <Phony Benoni> an alternative game score that records 31. Qd2 being played here. But as he correctly concludes, it seems likely that 31. Qf2 was the move Mason actually played. Among other things, it does not allow Black to force the trade of Queens.

31... Be6?

Another doubtful effort by Janowski. 31...Qc4 looks like the best chance to attempt to punish White for his prior hesitant play.

32. Ne4 Qb6

Probably best here, but no longer as strong as it would have been two moves earlier.

33. Rb1 Bf5

click for larger view

34. Bc3

"Instead of liquidating by 34. NxB QxQ+
35. KxQ gxN etc., White considers his centralized Knight more precious than an opposing Bishop." (Tartakower/Du Mont). Indeed, Mason's Knight proves to be a monster in the eventual endgame. So, yes, the text was superior to 34. NxB. But perhaps even stronger was 34. Re3! Be4 35. Nb3 with all sorts of threats.

In any case, the text was certainly not a mistake.

34... Be4
35. a4

"?"--(Tournament Book)

"Into consideration came 35. Bd2 intending to maneuver Nc5." (Tournament Book)

35. Bd2 looks too slow. Black, instead of doing nothing, might try 35...f6 followed by a sacrifice of the exchange after 36. e6 Rxe6 37. NxR QxN with better chances than in the game.

After Mason's actual 35. a4, which was probably best, the position was:

click for larger view

35... c6?

Better chances were afforded by 35...g5 or maybe 35...f6. But Janowski's move turned out not to be so bad after all when Mason responded:

36. a5?

"?--(Tournament Book)(<Max Lange>)

As pointed out by <Max Lange>, far better for White was 36. b5!

<FLClove> claims that 36...b5 would run into 36...c5. But White would doubtless answer this with 37. Nc6! (rather than the much weaker 37. bxa6 Qxa6 given by <FLClove) after which Black seems to be in serious trouble.

36... Qa7

This left:

click for larger view

37. Rb2

As Tartakower/Du Mont pointed out, 37. Nxc6? QxQ+ 38. KxQ would have been a mistake, especially if Black then continued 38...Rc7! instead of 38...Bxc2? as given by Tartakower/Du Mont (which would have allowed White to regroup with 39. Rbc1!).

Mason might, in addition to the text, have considered 37. Bd2.

37... Rb7
38. Reb1

Yet again, 38. Nxc6? would have been a horrible mistake in light of 38...QxQ+ 39. KxQ Rc7! (far superior to 39...Kd7? as given by Tartakower/Du Mont) after which Black is suddenly winning since White now loses a piece.

38... Kd7?

38...c5 or 38...g5 or perhaps 38...Rc7 look like Black's best legitimate chances here.

39. Kh1

Getting the King off the a7...g1 diagonal. But now White has to deal with some new pins (of the d4 Knight and the g2 pawn).

39... Reb8

39...Be7 seems better. After the text, the position was:

click for larger view

Starting here, however, the players appear to have been troubled by the approaching move-45 time control.

40. Qe1?

Missing the very strong 40. f5!

40. Qd2, called strong by Tartakower/Du Mont and by the usually reliable <Pawn and Two>was also far inferior to 40. f5.

Janowski, however, failed to exploit Mason's weak move, and played:

40... Be7?

Now was surely the time for 40...c5!

The text left:

click for larger view

41. Nb3?

"A trap..." (Tartakower/Du Mont)

"Actually, this trap has a large flaw in it. Mason should not have played 41. Nb3. White has a strong advantage, but needs to find a different continuation for the win." (<Pawn and Two>)

The supposed "trap" Tartakower/Du Mont claim to have spotted was 41...Bxb4 42. BxB RxB 43. Nc5+ QxN 44. RxR supposedly winning the exchange for White. But as <Pawn and Two> pointed out on this site 18 years ago, Black on the above line would just play 44...Qxc2. Black then has excellent counter-play (including an extra pawn plus two dangerous connected passed pawns) and is certainly not worse.

But I part company with <Pawn and Two> is in the best line after 41...Bxb4 (which was--far from being a mistake--was Black's best). The line given by Tartakower/Du Mont assumes that White would respond to 41...Bxb4 with 42. BxB. But (as <Pawn and Two> has shown), much better for White--and retaining his edge--would be 42. Bd4! Play could then continue--as per the line given by <Pawn and Two> 42...BxQ 43. BxQ RxN. But then White should play, not 44. cxR as given by <Pawn and Two> but 44. RxR; e.g., 44...RxR 45. RxR Bxa5 and now White, up the exchange for two pawns, would have excellent winning chances with either 46. c3 or 46. Bc5 (since 46...Bxc2? would then get crushed by 47. Rb7+

This all returns us to the question of 41. Nb3. Did Mason have anything better?

He did. Rather than seeking some uncertain action on the Queen-side, Mason should first have consolidated his probably still winning position with 41. Qe2 or 41. Kh2 (getting out of the pin on the long diagonal) or maybe even 41. Qf2.

Once Mason played 41. Nb3? Janowski, as discussed above, could best have responded with 41...Bxb4, falling into "the trap." But Janowski now played a move that could have cost him the game almost immediately:

41... Rb5?

The very best that can be said for this move is that "it set the stage for a dramatic finish." (<Pawn and Two>), the position now being:

click for larger view

42. Bd4!

Now Mason was very much back in the driver's seat.

42... Bc5
43. Bc5 BxB?

This exchange should have allowed Mason to finish off the game with dispatch (43...Ke8 was plainly more prudent, though at least theoretically insufficient to save the game. But after 43...BxB the position was:

click for larger view

44. bxB?

Hoffer's analysis that 44. NxB+ would have been a mistake stood, so far as I have found, until 2006 when <Pawn and Two> refuted it on this site. In fact, and again as was pointed out by <Pawn and Two>, 44. bxB? should have allowed Janowski to save the game. By contrast, and Hoffer notwithstanding, 44. NxB+ wins easily: 44...RxN (Hoffer's move, which is best but unavailing) 45. bxR QxR 46. RxQ RxR. Hoffer said that the resulting position (in which Black has Rook and Bishop for White's Queen) "...is not good for White." But <Pawn and Two> has annihilated this analysis by finding White's winning continuation: 47. Qh4! Black is now dead, e.g., 47...Rxc2 48. Qf6! Rxg2 [looks strong, but...] 49. Qxf7+ Kc8 50. Qf8+ Kc7 51. Qd6+ Kb7 52. Qd7+ Kb8 53. Qd8+ Kb7 54. Qb6+ [the position for White almost plays itself] Kc8 55. Qxc6+ Kd8 56. Qf6+ Ke8 57. h4! and now Black's discovered check avails him not at all.

Of course, with the clock ticking, one cannot fault Mason for missing all this.

44... Rb4

"?"--(Tournament Book)

This move may or may not have been best, but Janowski was now OK in any event. His move was certainly not any kind of blunder.

45. Qh4 Kc8

Perhaps 45...Ke8 was better, but Black should still be able to hold the game, especially since move 45 had now been reached:

click for larger view

We now come to the part of the game that has made it the celebrated treasure it has so justly become:

46. Qf6 Qd7

Threatening mate in two.

click for larger view

47. Kh2!

A nice quiet move that avoids the threatened mate and which gave Janowski a chance to go disastrously wrong. 47. Kg1 was also good.

The position was now:

click for larger view

A more difficult position than it may appear at first blush. And it was here that Janowski made his fatal mistake:

47... Qe6?

"?"--(Tournament Book)

"Giving White a wonderful combination"--(Tournament Book)

"An error that cost him the game"--(<Pawn and Two>)

To recognize just how tricky this position was/is, one need only look at the analysis by Tartakower/Du Mont, who--with all the time in the world, suggested here 47...Kc7. This move, as pointed out by <Pawn and Two> gets crushed by 48. Nd4!, e.g., 48...RxR 49. RxR RxR 50. e6! fxe6 51. Nxe6+ [a cute fork!!] Kb7 [if instead 51...Kc8 52. Qh8+ as per <Pawn and Two>] 52. QxR+ Kc8 53. Qf6 after which Black's game crumbles.

Janowski actually had several ways to hold the game here, the simplest probably being 47...d4. The Tournament Book's suggested 47...Bxc2 was probably also good enough to avoid defeat.

But after Janowski's blunder, the position was:

click for larger view

"With his next three moves, Mason finished off the game with a beautiful winning combination"--(<Pawn and Two>)

48. Qh8+

"!"--(<Pawn and Two>)

48... Kd7

"After 48...Kc7 White has the same brilliant reply" (Tartakower/ Du Mont)

The position now before Mason's transcendentally brilliant 49th move was:

click for larger view

49. Nd4!!!

click for larger view

"!!"--(Tournament Book)

"!"--(<Pawn and Two>)

"The sacrifice of the Queen yields an ending of singular beauty"--(Tournament Book)

"A liquidating sacrifice. It was of course foreseen by White on his preceding move"--(Tartakower/Du Mont)

"The ending that followed was completely hopeless for Janowski."

49... RxR

Even worse (if that's possible) would be 49...RxQ 50. RxR. And--needless to say--if instead 49...RxN?? then White mates in two beginning with 50. Rb7+!

50. QxR!

click for larger view

"!"--(Tournament Book)(<Pawn and Two>)

"The point! The Black Queen will find herself in an ambush and be lost"--(Tartakower/Du Mont)

50... RxQ
51. RxR

click for larger view

What a spectacular position! Black, despite his Queen, is entirely helpless.

51... Kc7

51...Qe8 was no better. And 51...Qe7, of course, runs into 52. Rb7+

52. NxQ+ KxR

click for larger view

"There follows an ending in which the Knight's dominating position is the deciding factor"--(Tartakower/Du Mont)

"The ending is hopeless for Black"--(Tournament Book)

53. Nd4

"!"--(Tournament Book)

53. Nd8 and 53. Ng5 also win easily. But the quiet text most neatly displays Janowski's utter helplessness here.

53... Kc7
54. g4

One of many ways for White to win. It left:

click for larger view

54... h4

Also wretched for Black would have been 54...hxg4 55. h4! [55. hxg4 also wins easily, but 55. h4! is prettier] 55...Kd7 56. Kg3 Bf5 57. h5!

55. c3

Giving Black not the slightest chance.

55... Kd7

click for larger view

56. Kg1!

Quiet...and deadly!

56... Kc7

One can only imagine Janowski's exasperation at being reduced to this sort of agonizing waiting move.

57. Kf2 Kd7

click for larger view

58. f5

One of many ways to win.

58... gxf5
59. gxf5

59. Nxf5 is also devastating. But the text likewise gave Janowski no chances whatsoever:

click for larger view

59... Kc7
60. Ke3

Abandoning his h-pawn in return for: (a) the Black h-pawn; (b) protecting his f-pawn; and (c) allowing his King to penetrate. All in all, a delightful winning concept by Mason.

He could also have won with 60. f6 Kb7 (or 60...Kd7 61. e6+ fxe6 62. Nxc6!) 61. e6 fxe6 62. Nxe6+ Kc6 63. Ke3 Bh5 64. Kd4 Kd7 65. Ke5 Bf7 66. Nd4 Kc7 67. Nf3

60... Bx2
61. Nf3

61. f6 is even simpler., e.g., 61...Bxh3 62. e6 Bxe6 63. NxB+ Kd7 64. Ng5 etc.

61... Bxh3
62. Nxh4

click for larger view

62... Bg4

62...Kd8 would only delay the inevitable.

63. Kf4 Be2
64. Nf3

64. Kg5 also wins quickly. The text left:

click for larger view

64... BxN

Leading to a completely lost King and Pawn ending. But the alternatives were no better: (a) 64...d4 65. Nxd4 Bc4 66. Kg5 Bd5 67. e6 Kd8 68. Kf6 Ke8 69. exf7+ Bxf7 70. Nxc6; (b) 64...Bd1 65. Nd4 Kb7 66. Kg5

65. KxB

click for larger view

Here the game was reportedly adjourned for a second time. The unpleasant ending is described in the Tournament Book:

"The game was...to be played out the following morning; but Janowski did not turn up, and the game was scored to Mason, his opponent's time having run out."

Thus did Janowski lose his chance to win (or place 2nd) at Monte Carlo 1902.

Janowski won his two remaining games in this tournament against Marshall and Teichmann, enabling his to take 3rd place.

At the next major tournament, Hanover 1902, Janowski took first ahead of Pillsbury and lost only one game: to Mason (who finished 17th out of 18, winning just one other game there).