The rook ending I lost three times before I bothered to learn it

Three times in two seasons I reached the same kind of rook ending a pawn ahead, and three times it ended in half a point or none. The fourth time I had finally done the homework, and the game was over in eleven moves.

1.Three losses that were all the same loss

The pattern was consistent enough to be funny by the third time. I would come out of a long middlegame a pawn ahead, both rooks still on the board, four pawns against three on one wing, and about twelve minutes on my clock. A book would have called the position winning. What I actually did was push the extra pawn far too early, let my rook drift in front of it, and then watch my opponent set up a defence behind me while I spent my last minutes hunting for something clever that was never there.

I only noticed the pattern because I write the ending type at the top of the scoresheet before the set goes back in the bag. Three sheets in a row said the same thing: rook and four against rook and three, all on one wing. That is not a run of bad luck, it is a hole. A stronger player on our team looked at the third game for about ninety seconds, asked me where my rook was supposed to be, and then went back to his own post-mortem. It was the politest correction I have had.

2.The defence with the rook on the third rank

The method that had been beating me is simple to state and takes a while to trust. The defending rook sits on its third rank and stays there, cutting the attacking king off from the squares in front of the pawn. As long as the pawn has not crossed that rank, the defender shuffles along and nothing can be arranged. The moment the pawn steps onto the third rank, the rook abandons the blockade, drops to the far end of the board, and checks the king from behind for as long as it likes.

Reading that in a book took two minutes. Believing it at the board took a season. The awkward part is the switch: a rook that has been useful on the third rank suddenly has to run to the eighth, and every instinct says stay near the pawn. I now set the position up on the pocket set and play the defence ten times in a row against myself, badly, until the switch stops feeling like a concession. Playing the losing side on purpose taught me more than playing the winning side ever did.

3.Escorting the pawn home

The winning technique from the other side has the same shape: the pawn cannot walk by itself and the king cannot walk without cover. With my pawn on e7, my king on e8 and my rook loose on the first rank, the idea is Re1-e4, which looks pointless and is the whole game. The rook on the fourth rank gives the king a screen to step out behind. When the checks come from the side, the king walks towards them and the rook interposes on the fourth rank at the right moment.

What goes wrong in practice is counting. If the defending rook is one file further away, the screen is not needed and a much simpler plan wins. If it is one file closer, the screen has to be built on the other side. I used to reach these positions and start calculating from scratch with eight minutes left, which is how you turn a win into a draw. The technique is worth memorising precisely because it removes the calculation at exactly the moment when the clock has taken calculation away from you.

4.How I drill it now

The drill is fifteen minutes, twice a week, on a pocket set on the kitchen table. I set up one of four positions, play both sides at speed without writing anything down, and then check only the moves where the two sides disagreed with each other. It is dull in the way that scales are dull. The point is not to discover anything; the point is that the switch of the rook to the back rank should happen before I have decided to do it.

Since the autumn I have had this ending five more times, three with the extra pawn and twice defending. Four went the way the book says they should, and the fifth I lost on the clock in a position that was still drawn, which is a different page in this notebook. That is not a transformation, but it is a hole closed. The next one on the list is the same ending with the pawns split, which I understand considerably less well than I understand this one.

Four positions only, rotated in order, no favourites.
Play the defending side first, always.
No clock during the drill; the clock comes at the board.
Check the engine after both sides are played, not during.
One line in the notebook per session, even if it says nothing new.