**tl;dr** Used [nba_api](https://github.com/swar/nba_api) to get team stats for 1103 teams starting from the 1983-84 season (data can be found on [github here](https://github.com/zbj1423/NBA-Playoffs-vs-Regular-Season-1983-2022.git) or [google sheets here](https://docs.google.com/spreadsheets/d/1RbduWYtAVmfztuekGE8AGgXlc6ANAivsO6xtIYJFGrg/edit?usp=sharing)). Made some [graphs](https://imgur.com/SJ2AtBh) showing differences in team stats between playoffs and regular season (also [split into eras](https://imgur.com/lPmHkOs)). Overall, a lot of the stats support the eye-test, and it was cool to see some exact values of differences between playoffs and regular season. The major differences: average points and offensive rating go down in the playoffs. Defensive rating goes up. All shooting percentage stats go down. More fouls are called and more free throws attempted. Teams have less turnovers and less total possessions. Assist stats all decrease, so maybe more iso-heavy possessions. Also, an interesting effect of playoff teams in 80s and 90s attempting more 3s in the playoffs vs. regular season compared to teams from 2007-2022.
# Introduction
With the playoffs starting tomorrow, I wanted to share some data I gathered using the [nba_api](https://github.com/swar/nba_api) in Python. A lot of claims are made about playoff basketball: higher intensity, players play harder, defenses are better, more gameplan adjustments, slower pace, etc. Many of these observations feel accurate just from watching games and looking at box scores, but I wondered exactly how much the stats support these claims.
# Results
To start, I looked at per-game team averages across all teams since the 1983-84 season and compared stats between each team’s regular season games and playoff games (detailed steps in Methods section).
The results are shown in these bar graphs: https://imgur.com/SJ2AtBh
Of course, averaging stats across 40 seasons is not an ideal approach since the game has changed so much throughout the years since 1983. So to try to address this issue, I split the team stats into 5 approximate NBA eras (roughly based on dominant playstyles and players of each era): 1983-90, 1990-98, 1998-07, 2007-14, 2014-22. The same analysis was run for teams in each of these eras to see differences between playoffs and the regular season.
The graphs after splitting into eras are much more cluttered but are shown here: https://imgur.com/lPmHkOs
**Note**: the major caveat with splitting the data into 5 eras is reduced sample size (i.e. number of teams) within each era. This means our estimate of the difference between regular season and playoff stats in each era have much wider variance and are less reliable. I removed the error bars in the bar graphs to decrease clutter, but keep in mind that they are pretty large and there are likely no statistically significant differences between eras if they were to be tested.
# Main Takeaways
From just this overview of comparing team stats, most of the differences in playoff games seem to fall in line with popular claims (i.e. better defenses, slower pace, etc.).
My 4 main takeaways:
**1. It’s harder to score in the playoffs.** Points and offensive rating go down in the playoffs. Defensive rating goes up. Teams take less FG overall and all shooting percentage stats go down. Teams actually attempt more 3s in the playoffs but 3PT% goes down. The exact numbers:
| Stat | Difference in Playoffs |
|————|————————-|
| PTS | -3.24 |
| FG% | -1.18 |
| 3PT% | -1.18 |
| eFG% | -2.14 |
| FGA | -2.19 |
| 3PA | +0.45 |
| Off_rtg | -3.00 |
| Def_rtg | +3.17 |
**2. Playoff games are more physical.** More fouls are called and more free throws attempted. FT% goes down slightly, which may be due to more pressure in the playoffs. The exact numbers:
| Stat | Difference in Playoffs |
|————|————————-|
| PF | +1.52 |
| FTA | +1.08 |
| FT% | -0.42 |
**3. Playoff games have slower pace.** Teams have less turnovers and less total possessions. Number of assists decrease and percentage of field goals that are assisted on (AST%) also decrease, which may indicate more iso-heavy possessions. The exact numbers:
| Stat | Difference in Playoffs |
|————|————————-|
| TOV | -0.76 |
| TOV% | -0.17 |
| AST | -1.89 |
| AST% | -3.36 |
| PACE | -2.07 |
**4. Splitting into NBA eras doesn’t change results much except for 3PA and 3P%.** To me, the most interesting difference from regular season to playoffs was in the 3PA and 3P%. Averaging across all seasons since 1983 shows that teams attempted 0.45 more 3s on average, but 3PT% dropped by 1.18%, likely due in large part to better defenses (3.17 better defensive rating in playoffs). However, these stats become more interesting after splitting the data into separate eras: we see that teams in the earlier eras (1983 through 2007) attempted more 3s in the playoffs on average, whereas in the two more “modern” eras (2007 through 2022) teams attempted around the same number of 3s as in the regular season. Differences in 3P%, however, are not drastically different among the 5 eras except for higher in 1990-98, which may be due to interaction effects related to the fact that the 3PT line was shorter in seasons 1994-97. These effects may hint at potential differences in how defenses played and/or offensive gameplans in earlier eras compared to modern ones. For instance, teams in the 80s and 90s were likely not gameplanning as much to protect the 3-point line, which could perhaps lead to teams in the playoffs attempting more 3s. Moreover, it’s interesting to think that maybe teams in earlier eras were, at least in part, more successful in the playoffs because they were “ahead of their time” and knew to attempt more 3s (eg. the ‘94, ‘95 championship Rockets teams come to mind). Even though 3P% decreased in most eras (likely due to better overall defenses), attempting more 3s in earlier eras could possibly still have led to more success simply because of the power of gaining 3 points in a single possession.
# Data and Methods
All data were collected using the [nba_api](https://github.com/swar/nba_api). The dataset consists of per-game stat averages for all teams starting from the 1983-84 season through the 2021-22 season. In total: 1103 regular season teams, 624 of which made the playoffs in their respective years. Traditional stats include: PTS, FGM, FGA, FG_PCT, FG3M, FG3A, FG3_PCT, FTM, FTA, FT_PCT, OREB, DREB, REB, AST, STL, BLK, TOV, PF, PLUS_MINUS. Advanced stats include: NET_RATING, OFF_RATING, DEF_RATING, EFG_PCT, TS_PCT, PACE, OREB_PCT, DREB_PCT, REB_PCT, AST_PCT, AST_TO, AST_RATIO, TM_TOV_PCT. Note: for plus/minus and all of the advanced stats, I could only get data beginning from the 1997-98 season onwards, which resulted in 743 total regular season teams, 400 of which made the playoffs.
**Analysis comparing regular season stats to playoff stats:**
The differences in stats between regular season and playoff games were calculated using the following steps:
1. Only teams that made the playoffs were included in the analysis (624 teams total for traditional stats and 400 teams for advanced stats)
2. To offset the effects of playing against bad teams, regular season games were filtered to only include games played against teams with above .500 records
3. For each of the 624 teams, calculate the mean of all stats across regular season games and across playoff games, respectively
4. Subtract regular season averages from playoff averages (to measure how stats change in playoffs relative to regular season)
5. Calculate the mean of differences across all teams (these values are what’s plotted in the bar graphs above, with error bars showing standard error across all teams)
# Limitations and possible next steps
This project only looks at common traditional and advanced stats, which only gives a small glimpse at how playoff basketball differs from the regular season. For example, I can’t really use these data to investigate questions like “how often do rotations change between playoff games?” or “how do the types of shots teams take change (eg. distance from basket)?” So I might play around some more with the nba_api and try to get more specific stats like shot distance, closest defender distance, fast break points, rotations, etc.
# Links for this project:
Thank you to the developers of the [nba_api](https://github.com/swar/nba_api)
[Github repo](https://github.com/zbj1423/NBA-Playoffs-vs-Regular-Season-1983-2022.git) with all the code and dataframes
[Google sheets](https://docs.google.com/spreadsheets/d/1RbduWYtAVmfztuekGE8AGgXlc6ANAivsO6xtIYJFGrg/edit?usp=sharing) with the extracted per-game averages data for all teams since 1983-84 season
[Bar graph figures](https://imgur.com/gallery/wSwoTZm)
[Basketball-Reference glossary](https://www.basketball-reference.com/about/glossary.html#poss) with abbreviations and brief explanations of stats
Thanks for reading. Any feedback is much appreciated. Hope you all find something interesting in these data. Excited for the playoffs to start!
Edit: added some thoughts and corrected links
by zhubajie747
27 Comments
Wow great post, well done.
Does taking away sweeps and 4-1 wins change the results?
I’m confused. How can the league average offensive rating change by a different amount than the league average defensive rating
Honestly anyone who has watched playoff basketball for at least 3 years knows all of this already. None this stands as “oh that’s interesting!!!”
Defenses tighten up because scouting improves + familiarity with opponent. Rotations shorten + become dynamic as adjustments are made. The same plays are spammed until other team can stop. Mismatches mean a helluva lot more than regular season record indicates. Veterans can handle the pressure better. Physicality roots out a lot of the skinnier players. Foul merchants take a hit.
It’s nice to have the data that confirms the eye test. The game does indeed “slow down” in the post-season. I think you are correct in that you’d see a lot more isolations compared to the regular season if the data was collected on that.
I find it interesting that more fouls are called. I’ve been under the impression that less are called and that FTA’s go down. However, seems the opposite is true, which makes sense because teams are more physical defensively. The change is that FT% goes down, but not attempts.
With the free throw stuff, I wonder if it’s that physicality goes up to a greater degree than the increase in attempts. More fouls are called, but maybe more fouls are also missed. Just a curiosity I had.
Anyway, great work!!!
There is also a bias for playoffs when analizing the regular season average to playoffs average. Regular season has more terrible teama than playoffs. We gotta compare the regular season stats of those 16 teams had between each other to playoffs.
Very well done. Appreciated the read.
I’d like to see this separated by decade or era
I really like this post. Some of it is stuff you’ll notice watching the game (more physical, game pace slows) but seeing the statistical difference (especially across eras) is really cool, so thank you
Does this analysis actually tell us anything? I would not have used league wide averages, instead looking only at regular season games where the playoff teams were playing each other. Otherwise, too many alternate explanations are let in. For instance, the reduction in offensive rating could be down to the playoffs cutting out the bad teams, where teams can boost their numbers.
TL;DR teams up their defense in the playoffs, chuck more 3s and play more hero ball.
I wonder how much roster-shortening in the playoffs impacts these stats. Shorter benches and more minutes per game for individuals. Would make sense that the pace slows down for more on-court “rest”, and also heavier ISO due to the stars playing more minutes as opposed to the bench units that normally rely on other methods to run the offense.
I’m curious on the pacing of the regular season vs playoffs. Plus doesn’t most teams only use 7-8 player rotation? especially the deeper they go into the playoffs.
Ty. Can you show bubble playoffs too?
I love posts like these. Thank you for your efforts in putting this together. Great job
defense really matters in the post season
Ftr, DRTG going up would not imply tougher defenses, the opposite in fact
Not a fan of the graphs you used, mad confusing
This is the data I need to back up my claim that the Gobert trade was worth it. Now if only he would stop punching teammates and play, we could see it in action!
Would be interesting if you could ID the teams that made the playoffs by year and limit the regular season analysis to those teams only. Wondering how much the non playoff teams are altering the results
Is there anyway you can do this for just 2016-2022?
Would be curious to know if teams can turn on a “switch” for the playoffs.
This is some next level stuff. It confirms what we all suspect, but it is very interesting to see the numbers. Thanks!
The FTA is kind of surprising to me. I feel like the playoffs have less of those ticky-tacky calls.
Aside from pace, all of this is explained by playoff teams facing tougher opponents in the playoffs than they faced in the regular season.
I think it would be more interesting to look at tracking stats that get at the actual style of play: transition possessions, shot location, playtype (iso, pick and roll, etc…), and tracking (percent of shots open or wide open, catch and shoot, etc.).
So youre saying I should not bet that Tatum will hit 50 points in at least one playoff game
That’s really interesting that the eras don’t show much difference besides 3PA going up