Qh6 seems to have been his move. This is from the beginning of his career; see Janowski vs Samisch, 1925, played near the end.

And I wonder if we have a descriptive notation typo here. After <24.d6>

click for larger view

<24...Bxc4> is laughingly bad. I think even Janowski would have played 25.dxe7, or even 25.Rxc4 in response.

So I wonder if the move was actually recorded as 24...BxP, and was interpreted as 24...Bxc4 instead of 24...Bxd6, which makes some sense. <25.Nxf6> would also be a natural response; in the score here, it's kind of iffy, isn't it?

I know what you mean. Part of the problem, I think, is that we all want to use hanging pawns safely - ie by slowly and securely pushing them up the board like the English scrum pushing the French back.

But that rarely works. The other bloke digs his heels in and the hanging pawns get stuck, ganged up on and taken.

The real answer to hanging pawns seems to be to use them to grab space and then to find an opportunity to throw them forwards, even if it means sacrificing them for the greater good.

What if 24...Rxd6 giving the exchange for a pawn and to blunt the attack? after running through some continuations quickly I don't see a forced win for White.

24...Bxc4?? Dear Lord this deserves a third and maybe a fourth question mark, which is clearly not matching Steel's calibre.

Clearly the game was registered as 24...BxP and misinterpreted, when the actual move has been 24...Bxd6, justifying Janowski notations "This merely plays into White's hands." An immediate, unthought response to 24.d6, probably because of time pressure.

And that is what Janowski does here. He takes risks with his centre pawns and allows them to be liquidated. But in return his pieces flood into the black kingside for a lethal mating attack.>

I believe you're right.

The Oxford Companion to Chess (1st edition) gives these two games as examples of how the side with hanging pawns can parlay them, by a timely advance and liquidation, into a winning attack:

I think you are very correct Phony..
One of the advantages of algebraic notation is that it avoids any possible ambiguity like that, although later descriptive can be very specific (i.e., B/7xP or B/6xP) but whichever play recorded the game, he probably thought it was obvious which bishop was capturing, and simply jotted down BxP. Note that black played BxP on both moves 24 and 25, so it certainly appears the move order just got reversed...

I think the actual move order was 24. d6 Bxd6 25. Nxf6 Bxc4 26. Qh6 . The final position is the same after either move order, but it is just a lot more sensible following the order given.

after 24.d6

click for larger view

d6 was obviously played as a decoying move, to lure the bishop away from the guard of f6. After Bxd6 Nxf6 the knight is immune because of Bxf6 mate.

click for larger view

The mating net is closing, and black tried the desperanzo Bxc4 in an attempt to eliminate one of the attackers (Be6). But of course, Qh6 puts blacks lights out.

Exchanging pieces is a normal way for the defender to minimize or neutralize an attack, but the beauty of this game is that black is OUT OF TIME. He is several tempi behind and is left defenseless.