K/9 vs Usable K Rate: Why Innings Decide Strikeout Projections
Two starters with identical K/9 can be a full strikeout apart in a fair projection. The number that separates them is not talent. It is workload: how many batters each one will actually face tonight. Here is how PropPrizm turns a rate stat into a strikeout projection, and why the innings column deserves more of your attention than the K/9 column.
What K/9 Actually Measures (and What It Hides)
K/9 answers a simple question: if this pitcher threw nine full innings, how many strikeouts would he collect at his current rate? It is a fine leaderboard stat. But it hides two things that matter enormously for a single-game projection.
First, nobody throws nine innings. The league-average start is a little under five and two-thirds. A 9.5 K/9 arm who leaves after five innings gets credit in the rate stat for four innings he never pitched. When you project a single game, the rate has to be multiplied by real, expected workload, and expected workload varies far more between pitchers than strikeout talent does.
Second, K/9 rewards pitchers who allow traffic. A pitcher who gives up a single before striking out the side records three strikeouts in one inning. A cleaner pitcher who retires the side in order records the same three. Same inning count, same strikeouts, but the first pitcher faced four batters to do it. Innings-based rates quietly inflate for pitchers who put runners on, because more baserunners means more batters faced per inning, and more chances to record a strikeout.
The Cleaner Rate: K%
Strikeout percentage, K%, fixes the second problem by changing the denominator: strikeouts divided by batters faced, not innings. A 25% K% means one strikeout for every four batters, no matter how messy or clean the innings were. It stabilizes faster than almost any pitching stat and travels better across contexts, which is why PropPrizm's strikeout model is built on K% rather than K/9.
That leaves the first problem, and it is the bigger one: how many batters will he face tonight?
Usable K Rate = K% × Expected Batters Faced
A strikeout projection is two estimates multiplied together. The strikeout rate is the stable part. Expected batters faced is where games are decided, and it moves with things that have nothing to do with the pitcher's stuff.
Pitcher A: 9.5 K/9, but his team pulls him at five innings. At roughly 21 batters faced and a 24% strikeout rate, a fair projection is about 5.0 strikeouts. His standard market line will often sit at 5.5, and he needs a good night just to reach it.
Pitcher B: 8.6 K/9, lower on every leaderboard, but he works into the seventh. At roughly 26 batters faced and a 23% strikeout rate, the fair projection is about 6.0 strikeouts. The "worse" strikeout pitcher projects a full strikeout higher, purely on workload.
What Sets the Leash
Expected batters faced is mostly about the leash, and the leash is set by things you can check before first pitch:
- Recent workload pattern. A starter who has been pulled at five innings in four straight outings is not suddenly pitching into the eighth. Recent innings per start is the single most predictive input.
- The third time through the order. Hitters gain roughly 30 points of expected performance the third time they see a starter, and modern teams increasingly pull pitchers before that penalty. Contenders with deep bullpens are the most aggressive about it.
- Baserunner rate. Walks and hits shorten starts two ways: they inflate pitch counts, and they end outings early when they cluster. A high-strikeout, high-walk profile is the classic short-leash combination.
- Roster context. A stretched-out reliever converting to starting, a pitcher returning from injury, or a rookie on an innings plan all carry caps that no rate stat will show you.
Rule of thumb: for a pitcher with a mid-20s strikeout rate, every extra inning of expected work is worth about one strikeout in the projection. Half an inning of leash is half a strikeout, which is the whole distance between a line looking rich and looking fair.
Where This Lives on PropPrizm
The model does this multiplication explicitly. Predicted strikeouts on the matchup dashboard come from strikeout rate applied to an expected-batters-faced estimate, which is itself built from estimated innings, opposing lineup quality, and how much traffic the pitcher tends to allow. Two places on the player card make the workload piece visible:
- The IP/Start tile shows average innings per start across whatever window you have selected. Flip it to the last five games and you will see the current leash, not the season-long average.
- Est IP on the matchup card is the model's forecast for tonight specifically, adjusted for the opponent and context.
When a strikeout projection looks surprisingly high or low against the market line, the innings estimate is the first place to look. Most of the time, the disagreement is not about the pitcher's talent. It is about how long he will be allowed to show it.
Bottom Line
K/9 tells you how good the strikeout stuff is per inning of opportunity. It does not tell you how much opportunity tonight holds. K% times expected batters faced does, and the second term moves the projection more often than the first. Read the rate for talent. Read the innings for the answer.
PropPrizm is a statistical tool for informational and entertainment purposes and does not guarantee outcomes.