June 9, 2024. Game 2 of the NBA Finals, Boston Garden. Jayson Tatum, the Boston Celtics’ primary creator who had averaged 26.9 points on 53% true shooting in the regular season, finishes the night with 18 points on 6-of-22 shooting. The Mavericks have switched everything at the point of attack, double-teaming Tatum on every catch above the break, and forcing the ball out of his hands to Boston’s second and third options. The Celtics win the game anyway, mostly because Jrue Holiday and Derrick White produce career-defining defensive performances at the other end. Tatum’s playoff true shooting that series drops to .519, well below his regular season number. The 2024 Finals MVP — Holiday — was, by every conventional measurement, the Celtics’ third or fourth best player in the regular season. The story of the playoffs is, frequently, the story of which regular season numbers survive a coordinated defensive opponent and which do not. Playoff scaling is the technical term for the structural question underneath every postseason narrative.
Playoff scaling — the question of how well a player’s or team’s regular-season performance projects into a seven-game series against a quality opponent — is, in my opinion, the single most underappreciated analytical frame in basketball coverage. The NBA’s 82-game regular season is, in a real sense, a different sport from the four-round playoff bracket. The defensive intensity is higher. The rotations are tighter. The scouting reports are deeper. The opposing coaches have weeks to study film and design specific game plans. A player who feasts on third-string defenders in the regular season may not see those defenders in the playoffs at all. A scoring system that works against ill-prepared defenses can be solved by a coordinated playoff opponent in two or three games. The math of which numbers travel and which don’t is the project of playoff-scaling analysis.
I have been writing about NBA analytics since 2014, with growing focus on the gap between regular-season and playoff production, and the question I find myself most often interrogating in my own writing is the one this article is going to unpack. Playoff scaling — which regular-season numbers survive May, which collapse, and how to read postseason performance through the lens of structural translation rather than narrative inflation, is the subject of this article.
The origin: where playoff-scaling analysis came from
The empirical study of how regular-season performance projects into the postseason dates to the early 2000s, when the first wave of NBA analytics began producing the data infrastructure that made the comparison possible. Dean Oliver’s Basketball on Paper (2004) included early observations about playoff variance, though Oliver’s primary focus was on possession-based analysis rather than the specific scaling question. John Hollinger’s writing at ESPN in the mid-2000s frequently surfaced the playoff-vs-regular-season gap for individual players.
The deeper analytical work came from FiveThirtyEight’s NBA coverage in the 2010s, particularly Neil Paine’s writing on playoff regression. Paine’s pieces consistently demonstrated that certain types of regular-season production translated more reliably to the playoffs than others. Cleaning the Glass (Ben Falk’s analytics site) built much of its public reputation on garbage-time-adjusted statistics that, in a sense, anticipated the playoff-scaling problem by removing the regular-season-only noise that the playoffs filter out.
By the late 2010s, the analytical community had developed a reasonably stable picture of which kinds of regular-season production survived the playoffs. Three-point shooting variance, garbage-time scoring, and production against bottom-tier defenses were known sources of regular-season inflation. Defensive performance against high-usage opponents, late-game shot-making, and consistency across multiple defensive looks were known indicators of playoff translatability. The vocabulary was, by 2020, sophisticated enough to support serious public-facing writing about postseason expectations.
The 2024 Finals — Boston winning a championship while their primary scorer struggled, with their third and fourth pieces producing the decisive performances — was a particularly clean illustration of the framework’s predictive value. The analysts who had written before the playoffs that Tatum’s specific shot profile would face challenges against a switching defense were largely correct. The pieces that had predicted Boston would still win, on the strength of their depth and defensive versatility, were also largely correct. The playoff-scaling lens had separated signal from noise in a way the conventional preseason coverage had not.
How playoff scaling works: in plain language
The basic insight is that the playoffs are a different statistical environment than the regular season, and the components of regular-season production translate to that environment unevenly. The most useful frame is to ask, for each type of production: does this depend on opponent quality, opponent preparation, or both?
Volume scoring against poor defenses does not translate. A player who racks up 32-point games against bottom-quartile defensive teams may not see those teams in the playoffs at all (they were probably lottery teams). The 28 points per game from those games is, structurally, the equivalent of a 22-point pace against playoff-quality competition.
Efficiency under high-quality defensive pressure translates well. A player whose true shooting percentage holds above .580 against top-ten defensive teams in the regular season is, in most cases, going to maintain that efficiency in the playoffs against teams of similar quality. The skill that produces the regular-season number against good opponents is, almost by definition, the same skill that produces playoff success.
Three-point shooting is the famously variable component. A team that shoots 38% from three across the regular season may shoot 32% or 44% in a specific playoff series, mostly due to variance. The “three-point variance” effect makes playoff outcomes less predictable than the underlying team quality would suggest. The teams with the most three-point-dependent offenses are the ones whose playoff outcomes are hardest to predict from the regular-season profile.
Defense generally translates, but with caveats. A team’s defensive scheme can be schemed-against in the playoffs in ways the regular-season schedule doesn’t reveal. A team that switches everything in the regular season may be exposed by an opposing coach who designs an offense specifically to attack switches. The defensive metrics are stable; the strategic vulnerability is variable.
The critical component: opponent-quality filtering
The single most important conceptual move in playoff-scaling analysis is to filter regular-season statistics by opponent quality before projecting them forward. The standard 82-game season includes roughly 30 games against playoff-quality opponents and 50+ against teams with poor defenses, lottery rosters, or rest-day rotations. Aggregating across all 82 games produces averages that inflate the playoff projection.
The fix is to look at each player’s and team’s per-100 production specifically in games against top-ten or top-twelve defensive teams. The filtered numbers are usually meaningfully different. A scorer averaging 28 points per game in the regular season at 56% true shooting may, against top-ten defenses, average 23 points at 51% true shooting. That’s a different player profile, and it’s the profile that’s more relevant for playoff projection.
Cleaning the Glass built much of its analytical reputation on these filtered statistics. The site’s “garbage time filter” and “opponent quality filter” produce stat tables that are, in my experience, dramatically more predictive of playoff performance than the unfiltered versions on Basketball-Reference or NBA.com.
Playoff-scaling metrics vs the alternatives: a comparison
The major analytical frameworks for projecting playoff performance:
| Framework | What it does | Where it shines | Where it breaks |
|---|---|---|---|
| Opponent-filtered TS%% | True shooting against top-10 defenses | Identifies scoring resilience | Smaller samples; noisier at player level |
| Late-game performance | Per-possession output in clutch minutes | Captures pressure response | Tiny samples; vulnerable to coincidence |
| Lineup data vs playoff teams | Net rating in games against likely playoff opponents | Direct match-up projection | Requires careful filtering of who counts as “playoff team” |
| Defensive versatility (positions defended) | Number of positions a player guards adequately | Predicts switch defense success | Hard to quantify precisely |
| Career playoff history | Player’s prior playoff TS%%, scoring, defense | Most direct evidence available | Roster context changes; samples limited |
The honest reading uses three or four of these in concert. A player who looks elite by all filters is, in expectation, a playoff-ready piece. A player who looks elite by one or two and average by the rest is more uncertain — the writer should name the uncertainty rather than smooth it into a single projection.
What the data needs: inputs
Playoff-scaling analysis requires game-by-game play-by-play data filterable by opponent and game-state, defensive rating splits for opposing teams, and per-100-possessions production for individual players. The standard public sources — Basketball-Reference, NBA.com/stats, Cleaning the Glass — provide most of this. The proprietary versions (Synergy, Second Spectrum) provide deeper coverage detail (specific defensive schemes faced, types of shots taken) that public analysis usually can’t access.
The historical playoff dataset is, for projection purposes, both rich and constrained. The NBA has 78 years of playoff data with consistent rules and a relatively stable game shape since the late 1990s. A player’s playoff history across two or three years is usually a more reliable indicator of how they’ll perform in May than any single-season projection. But playoff samples are small — a player’s career playoff total may include only 40-60 games against varied opponents — so the inference is always probabilistic rather than definitive.
Building the analysis: a working framework
The practical workflow:
- Pull each player’s per-100 production against top-10 defensive teams for the current season. Compare to their season-long per-100 production. The gap tells you about playoff-quality opponent resistance.
- Look at career playoff history. A player with three playoff appearances all showing efficiency drops of 5+ TS% points is structurally different from one whose playoff numbers match their regular-season profile.
- Identify the team’s defensive versatility. Teams that can switch effectively across multiple matchups are harder for opposing offenses to scheme against in a seven-game series.
- Project specific matchup risks. A team built around three-point variance is more vulnerable to a series of cold-shooting nights than a team with multiple scoring profiles.
- Cross-reference with coaching playoff history. A head coach with a track record of making in-series adjustments is more likely to win a 4-3 series than one who tends to stick with regular-season rotations and schemes.
Where this gets weird: common mistakes
Single-game extrapolation. A player who has one elite playoff game gets headlines and “this is the playoffs” narratives. The math says one game is one game. The same player’s series-long performance, three or four games in, is much more informative.
Coach-specific reputation transfer. A team that hires a coach with playoff success at a different organization is not, automatically, a playoff-successful team. The coach’s contributions are real but partially context-dependent. The new roster, new staff, and new opponents are different challenges.
Three-point variance attribution. A team that wins a playoff series partly on hot three-point shooting is often described as having “found a level” or “clicked at the right time.” The variance was usually variance, not skill acquisition. The teams that consistently sustain playoff three-point performance across multiple seasons are doing something replicable; teams with single-series hot shooting are usually not.
Star power assumption. A team with two elite stars and a thin supporting cast may, in the regular season, post a strong net rating. In the playoffs, when defenses can focus exclusively on those stars, the supporting cast is often the swing factor. Teams with depth tend to over-perform their regular-season metrics in playoff series; teams without depth tend to under-perform.
Defensive scheme exposure. A defense that has been effective all season can be exposed by an opposing offense that has time to design specific counter-actions. The teams whose defenses have been “tested” against varied competitive offensive systems are usually more playoff-stable than teams whose defensive numbers were earned against a softer slate of offensive opponents.
When playoff-scaling analysis shines: use cases
Series prediction. The opponent-filtered metrics combined with career playoff history produce playoff predictions that beat the betting markets in a measurable number of series each year. The edge is small but real.
Trade-deadline analysis. A team adding a veteran with strong playoff history at the deadline is making a different bet than one adding a regular-season-strong young player. The playoff-scaling frame produces more nuanced evaluation of these moves than the conventional “they got better” framing.
Roster construction. Front offices building around playoff success need different roster types than those optimized for regular-season win totals. The two-way wings, the versatile defenders, the multi-skill scorers — these archetypes are more playoff-ready than pure scoring specialists or one-side-of-the-ball players.
Coach evaluation. A coach whose teams consistently outperform their regular-season metrics in the playoffs is doing something measurable. Erik Spoelstra (Heat), Steve Kerr (Warriors), Mike Budenholzer (Bucks 2021 era) all have track records that exceed their teams’ regular-season profiles in the playoffs. The pattern is not random.
A working example: the 2022-23 Nuggets
The 2022-23 Denver Nuggets are one of the cleaner playoff-scaling case studies of the modern era. Denver finished the regular season with the West’s #1 seed and a strong but not historically elite net rating. By every opponent-filter metric, the Nuggets’ profile actually projected slightly better than their seeding suggested — Jokic’s per-100 production against top-10 defenses was elite, the team’s defensive versatility was strong, and the supporting cast (Murray, Gordon, Porter, Caldwell-Pope, Brown) all had above-replacement playoff projections.
The team swept through the playoffs more dominantly than the seeding gap implied. Denver won 16 of their 20 playoff games, the most one-sided championship run since the 2017 Warriors. The post-hoc narrative was about Jokic’s brilliance and championship inevitability. The pre-playoff analytical writing — at Cleaning the Glass, at The Athletic, at FiveThirtyEight — had largely captured the team’s playoff projection accurately. The framework worked.
The contrasting 2023-24 Nuggets, by similar metrics, projected as a strong contender but not necessarily a back-to-back winner. The supporting cast had aged slightly. The bench production had thinned. The defensive versatility was marginally lower. Denver lost in the second round to a hot-shooting Minnesota team. The playoff-scaling lens had been right both seasons, in opposite directions.
The limits: what playoff scaling cannot tell you
Playoff scaling cannot predict individual game outcomes. The framework gives series-level probability estimates. Single games are dominated by variance.
Playoff scaling cannot capture coaching adjustments in real time. A coach who adjusts mid-series can change the underlying matchup math. The framework projects from regular-season tendencies; mid-playoff scheme shifts are not modeled.
Playoff scaling cannot fully account for injuries and rest. A team whose star is playing through an injury that the public doesn’t know about will underperform the projection. The opposite — a team whose role players are fresher than expected — can overperform.
Playoff scaling cannot replace film study for matchup-specific evaluation. The numbers say which teams have favorable structural profiles. The film says which specific actions will exploit those profiles. The combination is the work.
One additional limit, less analytical and more sociological. The narrative coverage of NBA playoffs is so dominated by “moments” — buzzer beaters, hot streaks, individual heroics — that the structural framework can feel inadequate to the experience of the games. The framework is right about what’s likely to happen across a series; the games themselves can be improbable in ways no model captures. Holding both — the structural likelihood and the actual unpredictability — is the analytical work.
Frequently asked questions
Which regular-season stats translate best to the playoffs?
True shooting against top-10 defensive teams, per-100 production filtered for high-leverage minutes, defensive versatility (positions credibly defended), and lineup net ratings against playoff-quality opposition. These are more predictive than season-long per-game averages.
Why do three-point shooting teams have higher playoff variance?
Three-point shooting has higher inherent variance than two-point shooting. Over a regular season, a team’s three-point percentage stabilizes near their underlying talent level. Over a single playoff series (4-7 games), variance can dominate — a team shooting 40% from three across the regular season can shoot 30% or 48% in a single series purely due to chance. Teams whose offenses depend heavily on three-point volume are more vulnerable to these swings.
Do coaching changes mid-playoffs work?
Rarely well in the NBA. The literature on mid-season coaching changes generally finds little evidence of immediate improvement; mid-playoff changes are even rarer and don’t have the historical track record to evaluate. The closer-to-effective version is in-series tactical adjustments by the existing coach, not full coaching turnover.
Where can I see playoff-filtered stats?
Cleaning the Glass is the best paid source. Basketball-Reference’s “season splits” pages provide some opponent-filtered data for free. NBA.com/stats allows filtering by opponent quality. The Athletic’s playoff coverage frequently surfaces opponent-filtered metrics in their analysis.
Sources and further reading
- Cleaning the Glass — Ben Falk’s site, the canonical source for opponent-filtered and garbage-time-adjusted NBA stats.
- Basketball-Reference — public historical playoff data with split filtering.
- Zach Lowe at The Ringer — the model for analytical playoff writing that integrates structural projections with film study.
- FiveThirtyEight Sports archive — historical playoff probability work, including Neil Paine’s scaling-focused pieces.
- NBA.com/stats — official stats portal with opponent-quality filtering.
The Tatum Game 2 struggle that opened this article — 6-of-22 against a switching defense that the regular-season data had already flagged as his most challenging matchup — was the kind of result the playoff-scaling framework anticipates. Boston won the championship anyway because their structural profile was deep enough to absorb a star’s struggling series. The math of which numbers travel and which don’t is, in my opinion, the single most useful analytical lens for understanding how the postseason actually works. For the broader frame on reading NBA analytics, our guide to pace and space is the natural companion piece.
