READING RECENT FORM: WHY THE LAST 12 GAMES MATTER

A full-season average is a single number that tries to describe months of games played under wildly different circumstances — early-season rust, mid-season injuries, role changes, coaching adjustments, and late-season rest. By the time you reach the business end of a campaign, that average is a blend of several different versions of the same player, some of whom no longer exist.

Consider a forward who spent the first two months of the year coming off the bench, then moved into the starting lineup and doubled his minutes. His season average will sit somewhere in the middle of those two realities — comfortably describing neither. Anchor to that number and you will consistently under-rate him now that he starts.

When we analyse a matchup, the question is not 'what did this player do on average since the season opened?' It is 'what is this player doing right now, in the role he currently occupies?' Those are frequently different answers, and the gap between them is where most naive projections quietly go wrong.

We use a rolling window of the most recent twelve games (L12) as our default lens. Twelve is a deliberate compromise. It is large enough to smooth out a single hot or cold night — the sort of variance that afflicts every athlete — but short enough to stay sensitive to a genuine change in form, role, or health.

Shorter windows are seductive because they feel current, but three or five games are statistically noisy. One blowout in which the starters rest early, or one foul-plagued night, swings a five-game average far more than it should. You end up chasing ghosts, reacting to randomness as if it were signal.

Longer windows have the opposite failing. A season-to-date or 30-game view is stable, but sluggish. It takes weeks to acknowledge a player who has just been promoted into a bigger role or returned, rounded into shape, from injury. By the time the long average catches up, the edge you were trying to capture has already been priced in by everyone else.

Twelve games tends to cover two to four weeks of a typical schedule, which maps neatly onto how quickly roles and rotations actually change at the professional level. It is recent enough to be relevant and long enough to be trustworthy.

Not every game inside the window deserves equal weight. A performance from last night tells us more about current form than one from three weeks ago, even if both fall inside the same L12 sample. To reflect that, our engine applies a gentle recency weighting so the most recent outings carry slightly more influence — without letting any single game dominate the picture.

The emphasis on 'gentle' matters. Overweight the last game or two and you are back to the noise problem, effectively analysing a three-game sample wearing a twelve-game costume. The goal is a smooth gradient: yesterday matters a little more than last week, which matters a little more than three weeks ago, with no cliff edges.

This weighting is most valuable for players whose role is trending. A guard who has seen his minutes climb steadily over the past fortnight will show a rising projection well before his flat, unweighted average acknowledges the change — which is exactly when that information is most useful.

A single average, however well constructed, is still a snapshot. The direction of travel is often as informative as the level. Two players sitting at the same L12 figure can be moving in opposite directions — one climbing out of a slump, the other sliding into one — and that trajectory frequently persists into the next game.

When you read a player's recent-form line, look at three things together: the L12 average itself, the slope of the trend across those games, and how tightly the individual games cluster around the average. A high number built on volatile nights behaves very differently from a modest number that shows up almost every time.

The most common mistake is treating one enormous outlier as the 'real' player. A 45-point explosion is memorable, but if the eleven surrounding games say 18, the explosion is the exception, not the identity. Good analysis lets the bulk of the sample speak and treats the spike as what it usually is: a single, unrepeatable night.

The second trap is ignoring the schedule that produced the numbers. Twelve games against weak opposition are not the same as twelve against elite defences. Recent form is the starting point of an honest analysis, but it is never the whole story — which is why we layer matchup context and consistency on top of it, each of which deserves its own discussion.