In March 2022 I wrote Triple Double Russ, which scraped Russell Westbrook’s game log and asked a simple question: when he records a triple double, do his teams win more often? The answer then was yes, and by a wide, statistically significant margin. That post was a snapshot of a career still in motion. Westbrook has since played four more seasons and changed teams three times, moving from a featured star to a reserve. This revisit re-collects the data through the 2025-26 season and asks two things: is the same analysis still reproducible, and does the original thesis still hold?
I am keeping the original post unchanged as a 2022 time capsule. Everything below is recomputed from a fresh scrape.
Yes, with one repair. Basketball-Reference still publishes Westbrook’s per-season game logs at the same URLs the original post used, for instance the 2025-26 log, and a plain request still returns them. What changed is the markup. The 2022 scraper keyed on the old game-log table and its column layout; the site has since renamed the table to player_game_log_reg and moved every statistic behind a data-stat attribute (pts, trb, ast, game_result). A scraper written against the 2022 page returns an empty table today rather than an error, which is the quiet kind of failure worth guarding against. The rewritten scraper lives in fetch_westbrook_data.py and raises loudly if the expected table is missing.
Two practical notes carried over from the original. Basketball-Reference rate-limits scrapers, so the fetch script sleeps between requests. And to keep this page building without touching the network, the scrape is cached to a CSV that the analysis below reads. The numbers here are current as of the 2025-26 season; re-running the fetch script later will collect more games.
The fresh scrape covers every regular-season team game from Westbrook’s 2008-09 rookie year through 2025-26. The original post had 1,093 team games to work with; this one has 1,438.
print(f"Career team games: {career_games}")
print(f"Active for {active_games} of them ({active_pct:.2f}%)")
print(f"Triple doubles: {td_games} ({td_rate:.2f}% of active games)")
print(f"Active games without a triple double: {non_td_games}")Career team games: 1438
Active for 1301 of them (90.47%)
Triple doubles: 209 (16.06% of active games)
Active games without a triple double: 1092The triple-double count is the load-bearing number, so it is worth checking against an authority. Basketball-Reference’s career triple-double leaderboard lists Westbrook first all-time with 209 as of the 2025-26 season. The scrape-and-count here returns 209, an exact match. The definition is unchanged from 2022: a triple double is any game where three of points, rebounds, assists, steals, or blocks reach double digits, which in practice is almost always points, rebounds, and assists.
The 2022 post found that Westbrook’s teams won 73.58% of games in which he recorded a triple double, against 55.79% when he did not. The updated numbers:
pd.DataFrame(
{"Win %": [f"{career_win:.2f}%", f"{active_win:.2f}%",
f"{td_win:.2f}%", f"{non_td_win:.2f}%"]},
index=["Career", "Active", "With a triple double", "Without a triple double"],
)| Win % | |
|---|---|
| Career | 56.33% |
| Active | 56.73% |
| With a triple double | 73.21% |
| Without a triple double | 53.57% |
The thesis holds. Westbrook’s teams still win about 73% of the games in which he records a triple double, a figure that has barely moved in four seasons, and only 54% of the games in which he does not. If anything the gap has widened slightly, from roughly 18 points in 2022 to 19.6 points now, because his non-triple-double win rate has drifted down as his teams have gotten worse. A two-proportion z-test rejects the null that the two win rates are equal.
print(f"z = {z_stat:.3f}, p = {p_value:.2e}")z = 5.249, p = 1.53e-07That the relationship survived four more seasons and three team changes is the strong form of the result. It is also where caution belongs. A triple double is not a treatment applied at random. It marks a game in which Westbrook played heavy minutes and stayed involved on both ends, and those games cluster in the seasons where he was a featured starter on a competitive team, for instance his 2016-17 MVP campaign in Oklahoma City and his 2020-21 season in Washington. The win-percentage gap is real, but it is a statement about the kind of game a triple double signals, not evidence that chasing the tenth assist causes the win.
The late-career data makes that caveat concrete rather than theoretical.
Since 2023, Westbrook has been a reserve, moving from the Clippers to the Nuggets to the Kings and coming off the bench for most of those games. His triple-double rate collapsed accordingly. The seasonal counts tell the story better than any summary statistic.
sns.set_style("white")
fig, ax = plt.subplots(figsize=(12, 6))
labels = [f"'{str(y)[2:]}" for y in td_by_season.index]
bars = ax.bar(labels, td_by_season.values, color="#1f4e99")
ax.set(xlabel="Season (ending year)",
ylabel="Games with a triple double",
title="Russell Westbrook triple doubles by season")
for bar, val in zip(bars, td_by_season.values):
if val:
ax.text(bar.get_x() + bar.get_width() / 2, val + 0.4, str(val),
ha="center", va="bottom", fontsize=9)
sns.despine()
plt.show()
The four seasons added since the original post contributed 15 triple doubles combined, fewer than he recorded in the single 2016-17 season. The thesis, then, is now carried almost entirely by his prime. It describes 2017-era Westbrook accurately and 2026-era Westbrook barely, because the reserve version of him rarely plays the kind of game that produces a triple double in the first place. This is the generalizability caution in practice. A relationship estimated over a career can hold in aggregate while saying little about the player as he exists today.
The original post split triple doubles by location and marked Durant’s 2016 departure from Oklahoma City, on the theory that Westbrook’s triple-double era began once he was the unambiguous focal point of the offense.
fig, ax = plt.subplots(figsize=(13, 6))
years = list(td_by_loc.index)
ax.plot(years, td_by_loc.get("Home", 0), marker="o", label="Home")
ax.plot(years, td_by_loc.get("Away", 0), marker="o", label="Away")
ax.axvline(x=2016.5, color="k", linestyle="--")
ax.set(xlabel="Season (ending year)",
ylabel="Games with a triple double",
title="Russell Westbrook triple doubles by location")
ax.legend(["Home", "Away", "Durant leaves OKC"], loc="upper right")
sns.despine()
plt.show()
The split by era is stark:
print(f"With Durant (through 2015-16): {td_with_kd} triple doubles")
print(f"After Durant (2016-17 onward): {td_without_kd} triple doubles")With Durant (through 2015-16): 37 triple doubles
After Durant (2016-17 onward): 172 triple doublesWestbrook recorded 37 triple doubles across his eight seasons sharing a team with Durant and 172 in the seasons after. Some of that is simply age and tenure, an older Westbrook with the ball in his hands more often, so I would not read the Durant departure as the sole cause. But the timing is hard to ignore, and it lines up with the win-percentage story: the triple-double seasons are the seasons in which Westbrook was the engine of his team.
The sharpest criticism of Westbrook is that he chases triple doubles at his team’s expense. If that were true, we would expect his teams to lose by more in games where he gets a triple double, the empty-stat-line loss. The margins say the opposite.
pd.DataFrame(
{"Avg. margin": [f"{td_win_margin:+.2f}", f"{non_td_win_margin:+.2f}",
f"{td_loss_margin:+.2f}", f"{non_td_loss_margin:+.2f}"]},
index=["Triple-double wins", "Non-triple-double wins",
"Triple-double losses", "Non-triple-double losses"],
)| Avg. margin | |
|---|---|
| Triple-double wins | +11.61 |
| Non-triple-double wins | +11.54 |
| Triple-double losses | -7.45 |
| Non-triple-double losses | -10.74 |
When Westbrook’s teams win, the margin is nearly identical whether or not he records a triple double (11.6 against 11.5). When they lose, they lose by less with a triple double (-7.4) than without (-10.7). That is the same conclusion the 2022 post reached, and four more seasons did not disturb it. Whatever a Westbrook triple double is, it is not a marker of a team quietly losing while he pads his line.
The 2022 result survives replication. Westbrook’s teams win far more with a triple double (73%) than without (54%), the difference is statistically significant, and the margin data shows no evidence that his triple doubles come at his team’s expense. The scrape is still possible, once the table name is repaired, and the count still matches Basketball-Reference to the game.
What the extra four seasons add is a caveat the original post had no way to see. The relationship is anchored in Westbrook’s prime, and his move to the bench has made triple doubles rare enough that the thesis now describes a version of him that no longer takes the floor most nights. The statistic is durable. The player it was measuring has moved on. Both things can be true, and holding them together is the honest reading of the updated data.
import sys
import IPython
import matplotlib as mpl
print("originally published 2022-03-06; revisited 2026-08-12")
print(f"Python: {sys.version.split()[0]}")
print(f"pandas: {pd.__version__}")
print(f"numpy: {np.__version__}")
print(f"seaborn: {sns.__version__}")
print(f"matplotlib: {mpl.__version__}")
print(f"IPython: {IPython.__version__}")originally published 2022-03-06; revisited 2026-08-12
Python: 3.11.14
pandas: 2.3.3
numpy: 2.3.5
seaborn: 0.13.2
matplotlib: 3.10.8
IPython: 9.8.0The scraping code is in fetch_westbrook_data.py, and the cached dataset it produces is data/westbrook_game_logs.csv. The original 2022 post is preserved as-is.