The last five minutes: giving away positions I had already won

1.Where the points actually went

I went through two seasons of scoresheets over a long weekend in June, thirty-one serious games, and sorted them by where the result was decided rather than by result. Eleven were decided in the last five minutes of one side's clock, and nine of those eleven went against me. In seven of the nine I had a position an honest observer would have called better. That is a great many half points sitting in one very small part of the game, and none of them are an opening problem.

The second thing the sort showed is that the collapses were not random. They came in positions where a plan had run out and a new one was needed, and they came after long periods of thinking earlier in the game. Every single one of the nine had at least one move earlier where I had spent more than twelve minutes. The clock did not run out because the position was hard at the end; it ran out because I had already spent the money somewhere in the middle.

2.A budget for the middle of the game

The budget I now keep is crude and written on the corner of the scoresheet before the game starts. Ninety minutes for forty moves, in my head, means the first fifteen moves get twenty minutes, moves sixteen to thirty get fifty, and whatever is left belongs to the ending. If I am behind the budget I am not allowed a long think until I am back on it. This is not a system for playing well; it is a system for not arriving at move thirty-two with six minutes and a decision to make.

What has been surprisingly hard is the permission the budget gives to play a move I am not sure about. A budgeted position often needs a decision in four minutes that I would happily give twenty, and taking the four is uncomfortable every time. The compensation is that the ending, which is where my study actually is, gets played at a speed where I can use it. I would rather be slightly wrong on move eighteen with a clock than exactly right on move eighteen and helpless on move thirty-eight.

• Write the budget on the sheet before move one.
• Behind budget means no long thinks until you are level.
• Reserve the last third of the clock for the ending, not the crisis.
• A four-minute decision made is worth more than a twenty-minute one you cannot afford.

3.What to do when the flag is close

Under two minutes the useful skills are not chess skills. The first is to stop calculating variations and start playing moves I already know: a checking sequence I have seen before, a rook swinging to a rank I recognise, the defensive setup from a drill. The second is to keep the pieces in contact with each other, because a coordinated position survives bad moves and a scattered one does not. The third, and the one I fail at most, is to keep writing the moves down.

The hardest thing to learn has been that offering or accepting a draw is a move like any other. For a long time I treated it as a small defeat, which meant I played on in positions where my remaining minutes could not support the position I had. Three times last season a draw was clearly the right result and I declined it out of something that was not confidence. Two of those three became losses within ten moves. I keep that count on the same page as the budget.

4.Six months later, honestly read

Since February I have played fourteen serious games with the budget. Two were decided in the last five minutes, one each way, which is a large change from the previous rate and a very small sample. My results overall are almost identical to last season, and I want to be careful about what that means. The wins have not arrived; what has gone is a particular kind of loss, and it has been replaced by a number of rather dull draws in positions I used to lose or occasionally win.

That trade sounds unglamorous and I will take it for another season before deciding. The next thing to look at is the first fifteen moves, where the budget currently says twenty minutes and my scoresheets say I regularly use thirty-two. The repertoire page is supposed to have solved that and clearly has not. If there is a pattern in this notebook at all, it is that the fix for a problem in one part of the game usually turns out to live in a different part entirely.