We made thirty-seven predictions. Nine of them were written so that nothing could ever prove them wrong.

The paper has written down three hundred and three things over ten weeks. Most are not predictions. Most are facts we checked at the time, or opinions we held. Only thirty-seven were predictions: something we said would happen, that had not happened yet, that somebody could later check.

Eleven of those thirty-seven had been marked. All eleven were marked right.

On Monday night we went through the other twenty-six for the first time. Four right. Two wrong. Two half right. Nine we cannot call yet, because the day has not arrived. And nine had nothing in them to check.

Take the clearest one. In July we wrote that a bank holiday tied to England's World Cup run "may be granted". England lost their semi-final to Argentina on 15 July, so no holiday came. But we had written "may". Had England won, we would have marked ourselves right. Written that way, the sentence wins either way.

Another was a list of three things worth watching. A list is not a claim about anything.

Two belong to other people. One says National Grid expects data centres to use a far bigger share of Britain's electricity by 2035. That is National Grid's forecast. Writing it on our list does not make it ours. The same goes for one recording what a music industry campaigner said the government ought to do.

The last four are opinions. They say a risk exists, or that something is possible, with no date and no number attached. One of them sits very close to this paper's own core belief: that a lot of British office work could already be done by far fewer people. That may well turn out to be true. As written, there is no way to test it, which means it looked like a checked position and never was.

Add the two we wrote down after they had already been announced, and eleven of the thirty-seven were never at risk. Nearly a third of the list was free.

There is a simpler reason none of this was spotted. The list stopped being written. The last prediction went on it on 23 July. We have published ninety-five pieces since, and added nothing.

Two other records went the same way. We keep a list of events we expect and mean to cover. It stopped on 18 July. We keep a record of our own mistakes. It stopped on 29 July. Three records that exist to keep us honest all fell quiet inside eleven days of each other.

One entry deserves saying the other way round, with a qualification attached. A lot of the world's oil is shipped through one narrow sea passage, the Strait of Hormuz. In June a deal reopened it. Our foreign affairs writer said it was a short truce rather than a peace, and that drivers here would not feel the benefit soon. He was right. The deal broke down within weeks, and oil is still well above where it sat before the fighting began.

The qualification is that we were covering the strait daily at the time, across seven pieces in six days. The prediction went on the list while the story was already running. That is a better entry than the Wimbledon ones, because the outcome was genuinely unknown when it was written. It is not the clean forecast the list makes it look.

Two others are the shape they should all be. Each names a document or a place, a date, and one thing that would kill it stone dead.