Reading the Numbers
Rate Stats vs Counting Stats
4 min read
TL;DR
A counting stat rewards opportunity. A rate stat measures quality. Confusing the two is the most common mistake in sports analysis, and it is why the league leader in a category is often not the best player in it.
Two different questions
Counting stats add up: home runs, tackles, assists, total yards. They answer *how much did this player produce?*
Rate stats divide: batting average, yards per carry, points per 36 minutes, strikeouts per nine innings. They answer *how good was this player when given the chance?*
Both are real. They just answer different questions, and a lot of sports argument is two people answering different questions at each other.
Why the leaderboard misleads
A running back with 300 carries at 3.9 yards per carry gains 1,170 yards. One with 180 carries at 4.8 gains 864. The first leads the league; the second was better every time he touched the ball.
Neither number is wrong. But if you ask "who was more effective," the counting stat gives you the wrong answer, because it is measuring workload as much as skill. Volume is a coaching decision. Efficiency is closer to the player.
The rule of thumb
Ask what the denominator is. If there isn't one, you're holding a counting stat.
- No denominator (52 receptions, 31 home runs) → depends on opportunity
- Per game → adjusts for availability, not for role
- Per opportunity (per carry, per target, per plate appearance) → closest to quality
- Per opportunity, adjusted for context → closest to truth, and the hardest to compute
Move down that list and you learn more about the player. Move up it and you learn more about the team's usage of him. Both are worth knowing; just know which one you're looking at.
Keep these
- Counting stats measure production; rate stats measure quality per chance
- League leaders in counting stats are often just the players given the most work
- Always find the denominator — if there isn't one, opportunity is doing the talking
Up next
What Sample Size Does to a Batting Average
