HoopsJunkie

Low leverage removed

Low leverage removed is the default stat filter on HoopsJunkie and is marked by the spray bottle icon () in the site header. It removes the final stretch of games where the outcome has already been decided in attempt to ensure ratings and on/off numbers describe the possessions where both teams were still competing.

The idea is similar to removing garbage time, but instead of fixed margin thresholds and on-court starter checks, it uses our win probability model directly via the leverage score attached to every possession.

This catches situations margin rules miss such as games that are effectively over in the 3rd quarter, or which are uncompetitive in the 4th quarter but a coach leaves starters in to ensure a lead isn't lost. It does this while leaving genuinely competitive endings untouched.

How a stretch is identified

We scan each game's possessions in order and look for the final unbroken stretch where the outcome was settled. A stretch opens when leverage falls below 0.1 (average leverage is 1.0). Because coaches concede games before the scoreboard fully does, the entry threshold relaxes when the lineups say the competitive part is over: below 0.25 once two or fewer starters remain on the floor, and below 0.4 once either team has pulled all five starters.

A stretch only resets if the game returns to a genuinely competitive state (leverage climbing back above 0.25, or 0.5 while a team is running its full bench). A brief flicker of life during a blowout does not un-flag anything, and only the final stretch of the game ever counts. If a real comeback develops, the earlier stretch is discarded entirely, mirroring how Cleaning the Glass disqualifies garbage time.

Finally, short stretches only count if the benches were in. A close game's last few dribble-out possessions technically carry no leverage, but removing them would be pedantry; a conceded game's short tail is exactly what the filter exists for.

In practice the filter removes about 7 to 8 percent of possessions across a season, compared with roughly 3 percent for garbage time, and covers over 99 percent of what the garbage rule catches.

Why it is the default

Ratings with low leverage removed answer the question most people are actually asking: how good is this team or player when the game is being contested? Decided-game stretches are often played at half intensity by bench lineups and the leading team usually coasts. Removing them slightly compresses the league's net rating spread (strong teams give back the padding they earn against beaten opponents) while barely moving anyone by more than a point.

The other two filters remain one click away in the header: garbage time excluded (the classic margin-and-benches rule) and raw, which counts every possession. On game pages the margin chart grays out the removed stretch, and flagged plays carry the spray bottle icon in the play-by-play, so you can always see exactly what the filter is doing.