Garbage Time: How It Quietly Distorts the Box Score

Every basketball fan knows the feeling. It’s late in a blowout, the starters are in warmups on the bench, the deep reserves are jogging through possessions, and the outcome was settled five minutes ago. We call it garbage time, and we mostly ignore it. But the box score doesn’t ignore it. Those meaningless minutes get counted exactly like the meaningful ones — and that quiet accounting trick distorts player numbers, poisons lineup data, and fools anyone who takes the raw totals at face value.

What garbage time actually is

Garbage time is the stretch of a game after it has been functionally decided — when the trailing team’s chance of winning has collapsed to near zero and both coaches know it. Its signatures are unmistakable: benches empty, starters sit, intensity drops, defenses stop scheming and start coasting, and the players on the floor are often the last men in the rotation getting their only run of the night. The effort is real in the sense that nobody is loafing on purpose, but the competitive conditions are gone. It is not the game the starters played — it is a different, lower-stakes game grafted onto the end of the same scoreboard.

The problem is that the stat sheet has no column for "this happened in garbage time." A bucket in a one-point game and a bucket up 28 with two minutes left are recorded identically. That equivalence is the source of every distortion that follows.

How it inflates the box score — in both directions

Garbage time bends counting stats and rate stats alike, and it does so for two different groups of players at once.

For benchwarmers, garbage time is a stat buffet. A deep reserve who would otherwise log a handful of minutes a week suddenly gets extended run against other deep reserves, with the defense disengaged. Points, rebounds, and assists pile up against weak resistance. Worse, because many rate stats divide by minutes or possessions, a player who only appears in these soft conditions can post gaudy per-minute or per-possession numbers that evaporate the moment he faces real competition. The raw line says "productive"; the context says "padded."

For starters, garbage time distorts in the opposite direction — by omission. A star who builds a huge lead and then sits the fourth quarter accumulates his strong numbers in fewer minutes. His per-game totals get suppressed (he didn’t play the garbage minutes), while his per-minute efficiency may look even better than it should, because he only played the high-leverage stretch where he dominated. Either way, the minutes a player did or didn’t play in garbage time skew how his season reads.

Why it especially poisons plus-minus and lineups

If garbage time merely nudged a few counting stats, we could shrug. The real damage is to the on/off and lineup numbers, where the distortion is concentrated and severe.

Consider what happens in the final minutes of a blowout. A bench lineup that had nothing to do with building the 25-point lead is suddenly on the floor. If the game stays a blowout, that unit’s plus-minus is roughly flat for those minutes — but if the winning team coasts and the margin shrinks from 25 to 14, that bench unit absorbs a big negative swing it didn’t earn, against the other team’s reserves who are happily padding. Net ratings for those lineups lurch around based on minutes that decided nothing. A handful of garbage-time possessions can move a five-man unit’s net rating by a startling amount precisely because the sample is small and the conditions are abnormal.

This is the trap we lay out in detail in our piece on plus-minus and RAPM noise: raw plus-minus is already dominated by who you share the floor with and who you play against, and garbage time is the most extreme version of that confound. A reserve’s glowing plus-minus is frequently a fact about when he played — soft opponents, decided games — not about how well he played. Filtering garbage time is one of the first things any serious lineup analysis does, because leaving it in means letting the least important minutes of the season shout over the most important ones.

How analysts filter it out

Because the distortion is real, the analytics community has spent years trying to fence garbage time off and keep only the "meaningful minutes." There is no single official definition, but the common approaches fall into a few families:

• Score margin plus time thresholds. The classic rule of thumb: exclude possessions where the lead exceeds some cushion with little time left — treating the game as decided once the margin climbs past a certain number of points inside the final minutes. The exact numbers vary by source, but the logic is always "a lead this big this late is safe."
• Win-probability cutoffs. A more refined version ties the filter to a win-probability model: drop the possessions where one team’s chance of winning has crossed some high threshold — say, well past 95% — so the cutoff adapts to time and margin together rather than relying on fixed point gaps.
• Curated "meaningful minutes" splits. Some data providers publish filtered datasets that strip blowout garbage out of the box score and lineup numbers, so the totals already exclude the junk.

All of these are trying to answer the same question — "which possessions were the players actually competing over?" — and all of them improve the signal in efficiency stats like offensive and defensive rating, where a few minutes of disengaged blowout basketball can drag a team’s per-100 numbers around.

An illustrative example

To make the effect concrete, here is a deliberately invented stat line for a hypothetical reserve over a stretch of games — the numbers exist only to show the shape of the distortion, not to describe any real player.

The raw line looks like a hidden gem: double-digit scoring and a sparkling plus-minus that screams "play him more." Strip out the garbage-time minutes — the soft run in decided games — and the same player is a modest scorer with a slightly negative impact, which is a far more honest description of his value. Nothing about the player changed between the two rows. Only the minutes we chose to count did. That gap, manufactured here for clarity, is exactly the kind of illusion that filtering is designed to dissolve.

The honest caveat: every cutoff is arbitrary

Here is the part that intellectual honesty demands we say out loud: there is no true, bright line where competitive basketball ends and garbage time begins. Comebacks from supposedly safe leads happen. A "decided" game occasionally un-decides itself. Any threshold — a fixed margin, a time cutoff, a win-probability number — is a judgment call, and reasonable analysts pick different ones. That means a filtered stat is only as defensible as its filter, and two sources can disagree about a player’s "real" numbers simply because they drew the garbage-time boundary in slightly different places. The filter is a major improvement over counting every blowout minute equally — but it is an estimate, not a measurement, and it deserves the same humility as any other modeling choice.

The takeaway

Garbage time is the box score’s blind spot: minutes that decide nothing, counted exactly like minutes that decide everything. It inflates benchwarmers’ totals, flatters or suppresses starters by what they didn’t play, and does its worst damage to plus-minus and lineup data, where a few coasting possessions can swing a unit’s net rating wildly. The fix — filtering by score margin, time, or win probability to isolate the meaningful minutes — genuinely improves the numbers, but every cutoff is a judgment call, not a law of nature. So when a deep reserve’s raw numbers look too good to be true, check when he earned them. The scoreboard counts garbage time. The smart analyst doesn’t.