Methodology

Every number on this site is explained here. No black boxes.

The DEFCON rule

Since 2025/26, FPL awards 2 points for defensive contribution, at most once per match. Defenders need 10 or more combined clearances, blocks, interceptions and tackles. Midfielders and forwards need 12 or more, and their count also includes ball recoveries.

Hit rate

A hit is a match where the player reached the threshold. We only count matches with 60 or more minutes played; a short cameo says nothing about a player's engine. The headline hit rate blends recent form with the season: 60% weight on the last 5 qualifying matches, 40% on the whole season. With fewer than 5 qualifying matches we use the season rate alone.

Near miss

A near miss is a match finishing 1 or 2 actions short of the threshold. A defender who keeps landing on 9 is closer to points than his hit rate suggests. High near-miss rates are a buy signal, not a flaw.

Expected points and value

DEFCON xPts per match is the blended hit probability times 2. Value is xPts divided by price in millions. Consistency is 1 minus the coefficient of variation of the player's actions: steadier engines score closer to their average every week.

Decisions

The Buy, Hold and Avoid calls follow fixed rules. Buy needs a hit rate of at least 60% and a next-5 fixture run averaging difficulty 3 or easier. Below a 40% hit rate, or with an injury, suspension or availability flag, the call is Avoid. Everything else is Hold. The one-line reason always shows the numbers behind the call.

Data source and refresh

All data comes from the official FPL API. Profiles are rebuilt twice a day, at 06:10 and 18:10 UTC. Each page shows the gameweek and time of its last update.

Known limitations

  • Early in a season the sample is small; hit rates move a lot until roughly gameweek 6.
  • Fixture difficulty uses FPL's own FDR, which is a blunt instrument against low-block opponents.
  • Minutes risk uses FPL status flags, which can lag press-conference news.

What we do not do

No black-box machine learning. No hidden weightings. No paid boosts to rankings. If a number cannot be explained on this page in plain English, it does not ship.