Combining Math + Film Study: The Best NBA Players of 2025-26 (and a new #1!)
With the 2025-26 season nearly complete and the Finals now underway, my player evaluations for the year are essentially concluded. This post is my ranking of the best players in the NBA this season, measured by estimated championship value.
The core question is straightforward:
How much does this player increase a good team’s probability of winning the championship?
That framing is the organizing principle behind the project. It weights regular-season impact, playoff translation, opponent-specific resilience, playoff and role portability, scalability next to other high-end talent, and the ability to remain valuable across a range of roster constructions.
I’ve now done this exercise for every NBA season since the merger. The purpose is not only to evaluate the current league, but to place each player-season into a broader historical framework of championship equity. Across the full project, I’m trying to estimate player value in the setting that matters most: high-leverage basketball against elite competition.
My background informs the structure of the project. I’ve played, coached, and scouted basketball at multiple levels, worked in NBA front office environments, and have a professional inferential statistics background. The evaluation process is built around integrating quantitative signals with film-based analysis in a way that accounts for uncertainty, context, and historical comparability.
The evaluation process is a Bayesian synthesis of statistical evidence, film study, and historical comparison. The goal is to estimate each player’s true championship-level value as accurately as possible, using every meaningful signal available.
The statistical side includes RAPM and RAPM-family impact metrics, luck-adjusted on/off data, EPM, DARKO, LEBRON, box-score production, efficiency indicators, play-by-play data, and postseason samples where available. These signals help capture a player’s measurable impact: how teams perform with him on the floor, how resilient that impact is across lineups, how much value shows up outside the box score, and how stable the profile appears across different samples.
The film side is equally central. I’m looking at how the value is actually being created: advantage generation, pressure response, counters, defensive recognition, off-ball value, spacing effects, passing windows, rim pressure, shot quality, matchup dependence, and how opponents try to solve the player over the course of a series. The film helps interpret the data, but it also generates its own evidence. Some skills are visible before they fully show up in the numbers; some statistical signals are real but context-bound; some box-score production is more portable than other box-score production.
The process is iterative. Statistical signals inform what to look for on film, film observations inform how to interpret the statistical profile, and both are checked against historical precedent and playoff translation. I’m trying to update toward the best overall estimate using all available evidence.
The main questions are:
Regular-season value matters, but championship value is ultimately filtered through the playoffs. The players at the top of this list are the players whose impact is most likely to hold up, scale, and remain decisive in the highest-leverage environments.
Each player is listed in the following format:
Player — OFF / DEF / NET
The offensive, defensive, and total impact estimates are measured roughly in points per game of impact contribution to a random team, adjusted to calibrate the proxy to added championship probability.
Each player also has a confidence interval around his net point estimate. These ranges reflect uncertainty from role, health, sample size, statistical noise, team context, and ambiguity in film interpretation. A player with a range (X-Y) means that holding the evaluations of other players constant, I could see an argument for the player to be as high as the X-th best player or as low as the Y-th best player just by making slightly more optimistic or pessimistic assumptions than I have made in my center estimates.
This ranking assumes full health.
General interpretation of NET estimates:
The point estimates should be read as estimated central values, not exact measurements. The confidence intervals are especially important; many players have overlapping ‘probability distributions’ and are difficult to separate cleanly. Not all confidence intervals are of the same width – some players are inherently fuzzier to evaluate than others.
1. Victor Wembanyama — 2.4 / 3.9 / 6.3 (1-2) -- All-time great
2. Nikola Jokic — 5.8 / 0.2 / 6.0 (1-3) -- All-time great // Strong MVP
3. Shai Gilgeous-Alexander — 5.2 / 0.4 / 5.6 (2-3) -- Strong MVP
4. Giannis Antetokounmpo — 3.5 / 1.5 / 5.0 (4) -- Solid MVP
5. Kawhi Leonard — 3.9 / 0.4 / 4.3 (5-8) -- Weak MVP
T6. Luka Doncic — 4.5 / -0.3 / 4.2 (5-8) -- Weak MVP
T6. Stephen Curry — 4.4 / -0.2 / 4.2 (5-8) -- Weak MVP
8. Cade Cunningham — 2.9 / 0.7 / 3.6 (5-13) -- Solid All-NBA
9. Jalen Brunson — 3.6 / -0.4 / 3.2 (8-14) -- Solid All-NBA
10. Anthony Edwards — 2.8 / 0.4 / 3.2 (8-14) -- Solid All-NBA
The following are the players who did not make the final top 10, but have high-end evaluations that, holding all other players constant, would land them in the top 10:
Jaylen Brown
Tyrese Maxey
Donovan Mitchell
Jamal Murray
Because this project covers every season since the merger, I use past season winners as an anchoring mechanism. The league environment changes substantially across eras, but the underlying question remains consistent: who provided the most championship value in that season? Here are my winners for every season since the merger, with Victor Wembanyama ending Nikola Jokic’s 4-year run.
Open to answering any questions people may have!