Filter, GroupBy, Merge

Boolean masks work on DataFrames exactly like on arrays — the condition just reads like a query:

• df[df['acc'] > 0.8] — keep the good models
• df[(df['acc'] > 0.8) & (df['params'] < 1e6)] — & for and, | for or, parentheses mandatory
• df.sort_values('acc', ascending=False) — rank the results
• df['optimizer'].value_counts() — count each category

That first line is the most-typed pandas expression on Earth. Inside the brackets is the same boolean mask you used on NumPy arrays — df['acc'] > 0.8 produces a True/False per row, and df[...] keeps the Trues.