Most prediction records are fiction.

People remember the five they called and forget the fifty they didn't. They remember the 2-1, not the three weeks of 1-1. Then they say "I'm usually right", and there is no way to check.

A record is supposed to fix that. Most don't, because they're written after the fact, or only when it goes well. Here's a version that actually means something.

Write it down before kickoff

No exceptions. Before the whistle, before the team sheet if you can manage it.

One line per match is enough:

Arsenal vs Chelsea — Arsenal win — 60% — because their midfield can't handle a high press.

Timestamp it. The timestamp is the whole point. A prediction without a timestamp is a memory with good lighting.

A probability is not a promise

Saying a team has a 60% chance and watching them win is not a hit. It is one draw from a weighted coin.

This is the part most fans skip. If you call ten matches at 60% and win seven, you were roughly right. If you win six, you were roughly right. If you win ten, you weren't good — you were lucky and possibly dishonest about the other calls.

Judge the number, not the result. That's the only way a record survives contact with a season.

Score it properly

A forecast can be scored, not just admired. The simplest method is a Brier score: take your probability, subtract what actually happened (1 for yes, 0 for no), square it.

  • You say 70% and they win: (1 − 0.7)² = 0.09
  • You say 70% and they lose: (0 − 0.7)² = 0.49
  • You say 50% and they win: (1 − 0.5)² = 0.25

Lower is better. Average the score across every forecast and you get one number that rewards being both confident and correct. A safe 50/50 every week scores 0.25. To beat that, your confident calls have to land often enough to pay for the misses.

Calibration beats highlights

Sort your calls into buckets. Of everything you rated 70%, how often did it happen? If the answer is 50%, you're overconfident. If the answer is 90%, you're underconfident and probably not saying enough.

This is calibration, and it is more valuable than any individual correct call. A well-calibrated 60% is a real skill. A lucky 95% is a story you tell at the pub.

Sample size matters. Twenty calls is a mood. Two hundred is a record.

Compare yourself to a baseline

Always pick the home team. Always pick the league leader. Always pick the favourite.

If your record can't beat one of those rules, it isn't insight — it's noise with a narrative attached. Keep the baseline written down next to your own score. It keeps you honest in a way that memory never will.

Keep the misses

The urge to delete a bad call is the beginning of a fake record. The misses are the data. They tell you which competitions you read badly, which bias you have (big clubs? away teams? draws?), and which confident-sounding reasons were actually just vibes.

A record is not a highlight reel. It's a lab notebook.

Separate process from outcome

Good process, bad result happens constantly. Bad process, good result is the dangerous one, because it teaches you nothing and feels great.

Grade the decision, not just the scoreline. Did the team news you relied on actually arrive? Did the tactical mismatch you saw show up in the first twenty minutes? If yes and the result still went the other way, the process was fine.

The point of the record

It isn't to prove you were right. It's to find out what you're actually good at, and where you're guessing and calling it analysis.

Keep the record. Show the misses. Score the numbers. Then, next season, you'll know which of your opinions deserve to be listened to.