Filtering and Querying a DataFrame

Most data questions are really filters. Which teams were good on both ends? Which players cleared a usage threshold? Which games were close? In pandas you answer them with a boolean mask — a column of True/False that selects the rows you want — or with the more readable .query() method. Master this one move and a 30-row table or a 30,000-row play-by-play log answers the same way. As a worked example I'll narrow a list of NBA teams down to the genuine two-way contenders — the ones good on both ends — and highlight them on a chart.

This builds on Pandas for Sports Data. The data is the bundled nba_ratings.csv (per-100-possession team ratings, Basketball-Reference, retrieved June 2026), so it runs offline.

1. A condition is a column of True/False

   Comparing a column to a value doesn't return one answer — it returns a True/False value for every row. That boolean Series is the mask.

   import pandas as pd
   
   df = pd.read_csv("nba_ratings.csv")   # Team, W, L, W/L%, ORtg, DRtg, NRtg
   lg_off = df["ORtg"].mean()
   lg_def = df["DRtg"].mean()
   
   above_avg_offense = df["ORtg"] > lg_off   # a True/False Series, one per team

   Put a mask inside df[...] and pandas keeps only the True rows. The league-average offense and defense (about 116 points per 100 on each end) are our thresholds.

2. Combine conditions — carefully

   To demand both above-average offense and above-average defense (a lower defensive rating is better), combine masks with &, and wrap each condition in parentheses.

   mask = (df["ORtg"] > lg_off) & (df["DRtg"] < lg_def)
   contenders = df[mask].sort_values("NRtg", ascending=False)
   
   # The same question reads more like English with .query():
   # contenders = df.query("ORtg > @lg_off and DRtg < @lg_def")
   print(f"{len(contenders)} of {len(df)} teams clear both bars")
   print(contenders[["Team", "ORtg", "DRtg", "NRtg"]].to_string())

   Teams above average on both ends

   Teams above average on BOTH ends: 11 of 30
                         Team    ORtg    DRtg   NRtg
   0           Boston Celtics  124.23  112.51  11.71
   1    Oklahoma City Thunder  120.43  113.12   7.31
   2   Minnesota Timberwolves  116.63  109.98   6.65
   3           Denver Nuggets  119.49  114.05   5.44
   4          New York Knicks  119.11  114.22   4.89
   5     New Orleans Pelicans  118.31  113.70   4.61
   7             Phoenix Suns  118.78  115.69   3.09
   8       Philadelphia 76ers  117.79  114.76   3.02
   10   Golden State Warriors  118.67  115.95   2.72
   12     Cleveland Cavaliers  116.21  113.80   2.41
   15        Sacramento Kings  117.68  116.01   1.67

   Eleven of the thirty teams were above average on both ends — the Celtics leading the group with a +11.7 net rating. Note the two ways to ask: the bracket-and-mask form is explicit, while .query() reads like a sentence (the @ tells it lg_off is a Python variable, not a column).

3. Highlight the filtered set

   Filtering is most persuasive when you can see what was kept. We'll chart every team's net rating and color only the contenders.

   import matplotlib.pyplot as plt
   
   ordered = df.sort_values("NRtg")
   elite = set(contenders["Team"])
   colors = ["#C56A1E" if t in elite else "#C2B7A1" for t in ordered["Team"]]
   fig, ax = plt.subplots(figsize=(8, 9))
   ax.barh(ordered["Team"], ordered["NRtg"], color=colors)
   ax.axvline(0, color="#6C7079")
   fig.savefig("filtered_bar.png", dpi=144, bbox_inches="tight")

   

   The highlighted bars are the rows our mask kept. Building the color list is itself a tiny filter — a comprehension that checks membership in the elite set team by team — which is a handy pattern any time you want to draw attention to a subset without dropping the rest.

Troubleshooting