Deep Dive №4 · Preseason 2026–27

The Kawhi Anomaly

The biggest rebound ever recorded by a 34-year-old.

Francesco Freedman · The Gravity Report · the GRAVITY player-value model · published October 6, 2026 · 9-minute read

Independent analysis, not peer reviewed. v1, methodology and figures verified against public sources as of October 6, 2026. This article will be revised.

Format change, October 20, 2026: this Deep Dive is now a regular web article instead of a LaTeX PDF. Same text, numbers, tables and forecasts. The historical model is now called GRAVITY Historical everywhere (it was GRAVITY-RS). The original PDF stays available. Original PDF.

In short. In the 2024–25 season, Kawhi Leonard played 37 games and was graded as a good starter. In the 2025–26 season, at 34 years old, he played 65 games and was graded as the 5th best player in the league per minute. His jump in impact score (+1.31) is the biggest one-year gain ever recorded by a player 34 or older, and he is the only star since 1978 to set career highs in usage and true shooting in the same season after 33. Late surges have an unkind history. Only eight players have surged to star level after 31. None of them kept tfrac34 of that gain, and the median kept almost none of it. Leonard's surge came in two parts, taking on a heavier role and hotter shooting. Both of those parts age differently historically. When we rebuild his season piece by piece, we predict an impact of +1.31 next season, roughly a top-ten level, which means he keeps about a third of his surge.

Two records

The latest edition of The 644 ranks Leonard at sixth with a score of 77.3, the oldest player in the top ten by three years.¹ The season that earned him that score came out of nowhere. We ran every NBA season since 1977–78 through GRAVITY Historical, the version of our model we use for history. We use this version because the full 644 leans on on/off data from Cleaning the Glass, a subscription stats site that strips out garbage time (minutes when the game is already decided). That data only goes back to 2003–04 and can't be shared. GRAVITY Historical is the same formula without that one input, so it works for every season and anyone can check it. It reproduces the historical values in Deep Dives No. 1 and No. 2 to within 0.03.² Two records fall out of that table.

Before going further, a quick note on what the numbers mean. Every rating in this article is an impact score: how much a player helps his team per minute, compared with an average NBA minute. Zero is average. A plus number is better than average, and the bigger it is, the better. To give you a sense of the scale from last season: +0.6 was better than 80% of the league's regular rotation players, +1.3 was roughly a top-ten player, and +2.0 or more was a top-six player. Only five players got there.

The first record is the size of the jump. Leonard went from +0.86, which was better than 84% of rotation players, to +2.17, the fifth-best in the league, a gain of 1.31. Among players 34 or older with 1,500 minutes, nobody has ever improved their impact score by more in one year. P.J. Tucker is the next closest, at +1.27, and he climbed from below average to just average. Only one player aged 32 or older has made a bigger jump, Damian Lillard in 2022–23, and he gave most of it back the next season, keeping only 17% of the improvement.

The second is the shape of the season. Leonard used 33.5% of the Clippers' possessions and shot .629 true shooting, and both numbers were career highs.³ Since 1978, four players have set career highs in these categories in the same season after turning 33. The other three players are Bojan Bogdanović (2022–23), Elden Campbell (2001–02), and John Lucas (1986–87). However, none of them were graded higher than +1.04 that year. Kawhi was graded +2.17.

What happens to late surges

A 34-year-old who suddenly plays better than he did at 33 raises an obvious question: is he reaching a new level, or is it just a good year? Historically, we have an answer for the general case. Since 1978, 70 players aged 31 or older have improved their impact score by 0.6 or more in a single season, about the distance between an average player and one better than 80% of the league. For the 66 of those who played the next year, the median player dropped by 0.57. The median player kept a third of his gain, and 13 of them kept three-quarters or more. Those 13 are mostly role players who climbed towards average; however, the stars tell a much harsher story.

Table 1. Every player since 1977–78 who improved by 0.6 or more after turning 31 and reached a star-level +1.5. “Kept” is the share of that one-season improvement still there the next season.

Player	Season	Age	Before	Surge	Next year	Kept
Clyde Drexler	1994–95	32	+0.81	+1.81	+1.07	26%
Chauncey Billups	2007–08	31	+1.51	+2.21	+1.07	−62%
Kevin Garnett	2007–08	31	+1.51	+2.27	+1.50	−2%
Steve Nash	2009–10	35	+0.85	+1.51	+1.15	45%
Tim Duncan	2012–13	36	+1.00	+1.60	+0.98	−3%
Dirk Nowitzki	2013–14	35	+0.89	+1.57	+0.70	−28%
Jimmy Butler	2022–23	33	+1.82	+2.47	+1.35	−71%
Damian Lillard	2022–23	32	+0.45	+2.15	+0.73	17%
Kawhi Leonard	2025–26	34	+0.86	+2.17	?

Eight players surged to star level after 31. Not one of them kept three-quarters of the improvement they made in their surge season the following year. Nash kept the most with 45%. Only five finished the next season at or below where they started. This record alone would settle the question of whether Leonard's surge would stay next year, but it doesn't. Kawhi's surge is entirely different.

The anatomy of the surge

GRAVITY Historical builds a season from nine measured parts, and some of them repeat from one year to the next more reliably than others. Think of it as a reliability score from 0 to 1. At 1, last season tells you exactly what to expect next season. At 0, it tells you nothing. Among players 31 and older, a player's scoring volume, creation, and offensive load all score above 0.9, and shooting efficiency scores 0.72, the least reliable part of the whole thing.

This table splits Leonard's surge into what he did more of, and what he did better. Last year, he took on far more of the offense. His load and volume each rose by about a full point on that scale, and those are the parts that usually hold. His efficiency also jumped from about average to +1.5, close to elite, and that's the part that's supposed to hold least. The two box-score numbers the model leans on most, Box Plus/Minus and win shares, sit in between: more reliable than shooting, less reliable than role.

Cleaning the Glass, which filters out garbage time, shows where the efficiency came from. Kawhi scored 126.9 points per 100 shot attempts, the best mark of his career and putting him in the 91st percentile at his position. Half of his shots came from mid-range, as they have for a decade, and he made 52% of them, also a career high. He drew shooting fouls on 13.9% of his shots, up from 8.7% a year earlier. However, his three-point percentage fell from 41% to 38%.⁴

That shows us two different kinds of improvements. The foul drawing might glimpse a return to his own prime. He drew fouls on 13% to 15% of his shots every season from 2016 to 2020, and getting to the free-throw line is one of the more reliable parts of an older player's game, scoring 0.86 on that same 0-to-1 scale. The mid-range makes are a career high on the shot he takes most; however, the shooting efficiency is the part that carries over least. If the forecast we make is wrong, the mid-range shot will most likely be where it goes wrong.

Put simply, our forecast expects him to keep about half to 60% of his bigger role in the offense but only about a third of his shooting jump. We predict a 2026–27 impact of +1.31, roughly a top-ten level. A simpler model that only uses his last two overall ratings predicts +1.29. Both of these would put him a little below his 2023–24 level of +1.45. A very good player, but not quite the one from last season.

Table 2. A look at Leonard's season, part by part. Each value shows how far above (+) or below (−) an average NBA minute he was, where +1 is clearly good and +2 is elite. Reliability (0 to 1) is how well each part carries over from one season to the next for players 31 and older (1,009 player seasons). The forecast predicts each part from its last two seasons.

Part of his game	2024–25	2025–26	Reliability	2026–27 forecast
Scoring volume	+1.47	+2.83	0.91	+2.14
Offensive load	+0.93	+1.85	0.93	+1.46
Creation	+0.17	+0.81	0.93	+0.49
Efficiency under load	+0.33	+1.48	0.72	+0.69
Box Plus/Minus	+1.04	+2.62	0.82	+1.55
Win shares per 48	+0.67	+1.87	0.75	+1.01
Steals and blocks	+0.52	+0.78	0.86	+0.52
Rebounding	+0.11	+0.35	0.97	+0.21
Getting to the line	−0.42	+0.59	0.86	+0.15
Impact score	+0.86	+2.17		+1.31

Can he stay in the top eight?

The 644 has a 2-year blend, meaning that part of his 2025–26 season would carry forward. Therefore, Kawhi doesn't have to repeat it to stay near the top. Over 65 games, he would need an impact of +1.28 to be given a GRAVITY score of 69.3 (8th place score from last edition).⁵ That bar is lower than you'd think: +1.28 is roughly the 12th-best per-minute season in the league, because his great 2025–26 still counts toward next year's score. An impact of +1.1 would still keep him in the top 10. Our forecast of +1.31 clears that top-eight bar by just 0.03, which is about as close as it gets. For players 31 and older, forecasts like this one usually miss by about 0.40 in either direction. Taking that into account, his chance of staying top 8 is about 50%, and top 10 about two in three.

In this situation, games matter less than they seem. The availability term gives last season's 65 games 30% of the weight, so a 45-game season would actually lower the bar slightly, from +1.28 to +1.25, because his great 2025–26 counts for more when next season is short. What's going to end up moving the answer is whether the shooting holds or not.

There's also a simpler way to ask the question that doesn't depend on our model at all: will Kawhi finish top eight in Box Plus/Minus (BPM), a public stat on Basketball Reference that estimates how many points per 100 possessions a player adds over an average player? That version is harder for him. It only counts next season, and he has to play at least 1,500 minutes to qualify. Our forecast works out to a BPM of about 4.7. In the last eleven seasons, eighth place has never taken less than 5.1, and it has taken as much as 7.4. He's also only reached 1,500 minutes in six of his last eight seasons.⁶

Toronto

Kawhi's trade finally got completed on September 14th after the investigation of his endorsement deal closed.⁷ The Raptors extended him for 2 years and $115 million.⁸ Toronto's other creator, Scottie Barnes, used 23.4% of the team's possessions last season. If Kawhi's load falls in Toronto, the part of his surge that history says should last is the part that the new role would take away. That's one way that the trade could change our forecast, and it would definitely change it for the worse.

The forecast

Kawhi Leonard ranks top 8 in The 644, July 2027 edition.

50%

#1–8 = Win · #9 or below = Miss. This is our own ranking; we keep one bet on it each article so the model gets graded too.

Kawhi Leonard finishes 2026–27 top 8 in Box Plus/Minus among players with 1,500 regular-season minutes.

30%

Basketball Reference, final regular-season table.

Kawhi Leonard uses 30% or more of Toronto's possessions in 2026–27.

60%

The part of his surge we expect to hold. He used 33.5% last season. Basketball Reference usage rate; void if he plays under 1,000 minutes.

Kawhi Leonard's true shooting falls below .610 in 2026–27.

65%

The part we expect to fade. He shot .629 last season. Basketball Reference; void if he plays under 1,000 minutes.

Public-stat bets are graded on Basketball Reference's final regular-season tables by April 30, 2027. Registered October 6, 2026. Scored by Brier score in the Forecast Ledger.

Relabel note (October 8, 2026)

The series is renamed from Articles to Deep Dives, because The Gravity Report now publishes two formats: Deep Dives (long, with the full method) and Articles (short). Same number, same address. No numbers, findings, or forecasts changed.

Method and sources

GRAVITY Historical (GRAVITY-RS in the code) follows the specification's retrospective equation, 1977–78 to 2025–26. Late surge: age 31 or older, 1,500+ minutes, prior season 1,000+, improvement of 0.6 or more. The part-by-part forecast predicts each part from its current and prior value, fitted on players 31 and older (1,009 cases). Rank and percentile translations use 2025–26 players with 1,000+ minutes (percentiles) and 1,500+ minutes (ranks). Code: papers/kawhi/surge.py, papers/common/gravity_rs.py, papers/common/pipeline.py.

The Gravity Report, The 644, July 2026 edition (CSV). https://thegravityreport.com/the644/the644-2026-07.csv
Basketball Reference, league advanced and per-100 tables, 1977–78 to 2025–26. https://www.basketball-reference.com/leagues/
Basketball Reference, Kawhi Leonard. https://www.basketball-reference.com/players/l/leonaka01.html
Cleaning the Glass, Kawhi Leonard (garbage time filtered; percentiles by position), accessed September 29, 2026. https://cleaningtheglass.com/stats/player/2173
The Gravity Report, The GRAVITY Specification, v3. https://thegravityreport.com/notes/spec/
Basketball Reference, 2025–26 advanced table. https://www.basketball-reference.com/leagues/NBA_2026_advanced.html
NBA.com, Raptors–Clippers Leonard trade. https://www.nba.com/news/raptors-clippers-kawhi-leonard-trade
TSN, Raptors confirm Leonard extension. https://www.tsn.ca/nba/article/raptors-confirm-contract-extension-for-kawhi/

The Forecast Ledger Original PDF (archived)

Cite this article

Freedman, F. (2026). "The Kawhi Anomaly." The Gravity Report, GRAVITY Deep Dive No. 4.
https://thegravityreport.com/notes/kawhi/