← Blog
Back to Blog

Why the First Three Innings Aren't Like the Last Three

A starting pitcher's strikeout total looks like a single number, so it's tempting to treat his outing like a single event: six innings of the same guy doing the same thing. It isn't. The first trip through a lineup and the last one are close to different pitchers facing different hitters, and almost everything that makes a strikeout projection hard lives in that gap.

The Same Arm, Three Different Games

Baseball has one structural quirk no other major sport has: the same nine opponents come back around, in order, over and over. A starter who works six innings sees most of that lineup three separate times. Each pass is a fresh look for the hitter and an increasingly familiar one for the pitcher, and it shows up in the numbers with unusual consistency.

The pattern is known as the times-through-the-order effect, and its direction never really changes. Hitters improve each time they face the same starter. Here's the rough shape of it league-wide — direction and scale, not a live readout:

Time through orderK rateOpp. wOBATypical pitches
1st (batters 1–9)23.7%.30842
2nd (batters 10–18)22.5%.32538
3rd (batters 19–27)20.9%.34235

Roughly three points of strikeout rate evaporate between the first pass and the third, and more than thirty points of opponent wOBA appear. The hitter has now seen the fastball, timed the slider, and watched what the pitcher goes to with two strikes. The pitcher, meanwhile, is 70 pitches deep with a slightly slower fastball and a slightly flatter breaking ball.

The key asymmetry: a strikeout in the second inning and a strikeout in the sixth are worth the same on the stat line, but they are not equally likely to happen. Projecting a full start as "his season K rate times the innings" quietly assumes they are — and that assumption is wrong in a predictable direction.

Why This Breaks Simple Extrapolation

The most common mistake in live research is linear extrapolation. A pitcher has four strikeouts through three innings, someone does the arithmetic, and out comes a projected eight over six. Two separate things make that number too high.

First, the innings he's already thrown were his best ones. He was working through the lineup for the first time, at full velocity, against hitters who hadn't seen a pitch from him all night. The remaining innings are structurally harder. The rate he posted in the first three is not the rate he'll carry into the last three.

Second, and more important, those last three innings might not exist.

The Real Variable Is Whether the Innings Happen

This is where strikeout research quietly becomes workload research. A starter's total is a product of two things: how often he misses bats, and how many batters he gets to face. The second one has far more variance — and it's decided by a person in the dugout, not by the pitcher's arsenal.

A start that ends after five innings and 88 pitches and a start that ends after seven and 105 are separated by roughly eight extra hitters. At a 23% strikeout rate, that's nearly two additional strikeouts of pure opportunity. No change in stuff, no change in matchup, just more chances. Almost nothing about a pitcher's skill profile moves a projection that far.

Example — Two paths to the same standard line

Pitcher A: 27% strikeout rate, but averages 5.1 innings and 84 pitches because his walk rate runs the count up. Expected batters faced: about 21.

Pitcher B: 23% strikeout rate, but averages 6.2 innings and 98 pitches because he pounds the zone. Expected batters faced: about 25.

A projects to roughly 5.7 strikeouts. B projects to roughly 5.8. The higher-strikeout pitcher is the lower projection, because efficiency bought B four extra hitters. Compare only the strikeout rates and you get this exactly backwards.

The Third Time Through Is a Decision, Not a Certainty

Modern bullpen management has made the third trip through the order the most fragile part of a start. Teams know the split in the table above, and many of them act on it aggressively. That creates a cliff rather than a slope: a starter is either allowed to face the top of the order a third time or he is not, and those two worlds produce very different totals.

The things that decide which world you get are mostly knowable before first pitch:

What This Means for a Standard Line

When the market sets a strikeout line, it is pricing the whole distribution: the outcomes where the starter cruises into the seventh and the ones where he walks four in the fourth and is gone by the fifth. That's why a pitcher with elite swing-and-miss stuff and a short average outing often carries a lower line than his rate stats suggest he should. The market isn't underrating his arm. It's rating his innings.

Understanding that saves you from the most common false read in this space — seeing a high strikeout rate next to a modest line and assuming the number is stale. Usually the number is doing exactly what it should, and the missing piece is workload.

How PropPrizm Handles It

The model does not project strikeouts as a rate multiplied by a fixed inning count. It builds the number from expected batters faced first, then applies a matchup-adjusted strikeout rate to those opportunities. That ordering matters, because it puts the highest-variance input where it belongs — at the front.

You can see the same idea on the matchup card. When a projection sits below what a pitcher's strikeout rate would suggest, the explanation is almost always in the innings column, not the stuff column.

Quick rule: before you look at any pitcher's strikeout rate, look at how many batters he's likely to face. Opportunity has more variance than skill in a single start, and it's the part most research skips.

The Takeaway

A start isn't one continuous performance. It's a first pass that flatters the pitcher, a second that's roughly neutral, and a third that may never arrive. Treating those three phases as interchangeable is what turns a reasonable strikeout estimate into a bad one — usually too high, and usually for a pitcher whose stuff genuinely is excellent.

The fix is unglamorous. Start with how deep he's likely to go, and let the rate work on top of that. It's a less exciting question than which slider is nastiest, and it explains far more of the outcome.


Want to see it in the numbers? Open any pitcher on the matchup dashboard and compare his strikeout rate to his recent outing lengths. The gap between those two is where most of the projection lives. PropPrizm is a statistical tool for informational and entertainment purposes and does not guarantee outcomes.