A rambling rook in a blitz game
Blitz games are usually a lot of fun while they're being played, but best forgotten once they're finished. Last week, however, I played a blitz game on Chess.com which actually featured a funny theme you don't hear much about.
It's the so-called rambling rook (the name was coined by Dutch author Tim Krabbé).
Here's the position after Black's 39th move:
White's decision to finally capture the rook was, of course, motivated by a lack of time on the clock to figure out a win, but, as I found out after the game, the position is, in fact, won for White.
I first learned about the rambling rook phenomenon in the well-known Dutch chess book Nieuwe Schaakcuriosa (1977), by Tim Krabbé. (It's called 'Dolle Toren', or literally 'Mad Rook', in Dutch.)
Krabbé devotes an entire chapter to these 'mysterious forces in the rook'. In my opinion, it's one of the best chess pieces he as ever written. Funny, instructive and full of philosophical and literary asides. What more can one wish for?
He starts his chapter with the following archetypical game:
Like sand after a walk on the beach, this rook gets between everything, causing annoyance and anger. For whoever is confronted with the mysterious forces of a rambling rook was often (...) completely winning. Until, in a moment of recklessness, he stalemated the opponent - apart from that last rook, which is protected by the stalemate rule, and which can therefore cheakily haunt the king virtually everywhere.
Virtually, for in many cases it's possible to tame the rook. Except that such discoveries have been done on a case by case basis - nobody has ever formulated any principles. (...)
Krabbé goes on to formulate many of these principles himself throughout the course of the chapter, most of them based on the scenario where two rooks are stalemating the opponent's king (as in my blitz game).
The first principle is very straightforward and educational, and was actually described by Horwitz back in 1881:
The white king walks to the intersection of the two rooks, which is f3. After that, it's all over.
The second situation requires a little detour:
This is just the very basics, but it's enough to understand how my opponent could have won our little blitz game. There, the intersection of the two rooks is the square d1. So, it's sufficient for the White king to walk to d1, after which Black runs out of checks. Simple if you know it!
Back in 1977, there were no computers to check for errors and wins in such endgames. Nowadays, this is easier, but it's funny to see how little computers still understand about rambling rooks. Sure enough, any engine will show an immediate win for White in my blitz game even with the rambling rook on the board (it's actually Mate in 12 after 1...Rc5+), but if you feed the position of Post-Nimzowitsch to Stockfish, it seems to never realize that 3...Rxf3! is by far the best practical chance for Black.
Instead, it tries to 'hang in there' with something like 3...Rh8 which is obviously hopeless. It's also interesting that although the evaluation after 4.Kxh4 Rf4+ is +7.0 for White, it doesn't see a win for White even after fifteen minutes (which is when I got bored and switched the engine off), and the evaluation actually drops back to 6.8 after some minutes.
Maybe it's different on more powerful computers, but I suspect that concept of the rambling rook seems not to have made it to the core competencies of modern chess engines yet. Isn't that almost as funny as the rambling rook itself?