The Perfect (Game) Score 2: The Re-Take

Three years ago I wrote about Sandy Alcantara and The Perfect (Game) Score, where I imagined a version of Don Mattingly with future-vision: a manager who knew exactly when each start was about to go bad, and pulled Sandy at the peak every single time. Clairvoyant Mattingly turned Sandy's already-unanimous 2022 Cy Young season (228.2 IP, 2.28 ERA) into a 208.2 IP, 1.08 ERA season, the lowest ERA by a 200-inning pitcher since Mordecai “3-Finger” Brown back in 1906.

At the end of that piece, I wrote a "Next Steps" section promising to run this for every starter and answer a question from the 2021 Cy Young race: how close to "perfect" was Corbin Burnes pulled in every start, and how does it compare to Zack Wheeler? Then I started playing God of War Ragnarök and forgot all about this.

Picking this back up in 2026, we now have now computed the live Game Score after every plate appearance, of every pitcher’s start from 2015 through last night (roughly 53,000 starts). Every start now has an actual Game Score (what happened) and a Perfect Score (the peak the starter could have exited at, if his manager had the crystal ball). Divide the season's actual game score by the season's perfect game score, and you get a number I'm calling Hook Efficiency: 1.000 means every start ended at its peak, and anything below is game score that was left on the table by leaving the pitcher in too long.

Which finally lets us ask the big question.

Think about everything that's happened to starting pitcher usage since 2015. The third-time-through-the-order penalty went from a research finding to a religion. Bullpens metastasized to eight arms. Entire front office departments now exist to tell managers when a starter's release point has drifted two centimeters. The complete game is functionally extinct. If there was ever a skill the sport has collectively decided to optimize, it's when to take the ball from the starter.

So the league's Hook Efficiency must have climbed steadily toward perfection, right? Here's the league, year by year:

That's a .870 in 2015 and a .873 in 2026. The share of perfect exits crept from about 34% to about 37%, so the easy calls (the "he's clearly done, go get him" calls) are being made slightly more often. But the total damage absorbed after the peak? Unmoved.

Here's the thing though: this isn't a story about managers failing. It's a story about what the Perfect Score is actually made of.

The gap between actual and perfect isn't mostly the 8th-inning Sandy situation from the original article, where a cruising starter visibly runs out of gas and the manager waits one batter too long. Those are the predictable ones, and the slow rise in perfect-exit rate suggests teams have gotten a little better at them. The gap is dominated by ambushes: the pitcher who cruises into the fourth having retired nine straight and doesn’t make it out of the inning. No model sees that coming, because if the pitcher's own arm doesn't know the blowup is coming, the manager's laminated card certainly doesn't.

In other words, Hook Efficiency has a structural ceiling, and the league has been pressed up against it the whole time. The information revolution optimized everything that information can optimize. What's left is the part of baseball that is genuinely, irreducibly unknowable in advance. It turns out that part is worth about 13% of a Game Score, every year, forever. The 2023 article's premise was a manager with future-vision precisely because that's what it would take.

So, Burnes vs. Wheeler, 2021, as promised:

Craig Counsell gave the hook Burnes close to perfection that season, but the real answer is that both numbers are outrageous. The league average hook efficiency that year was at .867; these two were managed at +.920. The great 2021 rate-vs-volume argument, the one where people threw hypothetical 22.2-inning stat lines at each other on Twitter, was secretly a debate between two of the best-managed pitcher seasons in baseball.

Here's my favorite part, run out their full perfect pitching lines and the 2021 debate doesn't get settled. Perfect Wheeler still holds the volume crown at 195.1 innings (with 226 strikeouts), but perfect Burnes is a historic season, and I need you to look directly at these numbers: a 0.91 ERA and a 0.61 WHIP over 148.2 innings. The crystal ball would have given Zack Wheeler a deserving Cy Young season, but it would also give Corbin Burnes the second lowest ERA for someone with +140 IP all time, only behind Ferdie Schupp in 1916.

Which brings us to this season. The best-hooked pitcher in baseball in 2026, at a nearly clairvoyant .971 with perfect exits in 10 of his 14 starts, is Shohei Ohtani.

Of course it is, the one pitcher whose workload is pre-negotiated between the training staff, the front office, and presumably Decoy, is the one being pulled at the peak almost every time. Behind him: Jacob Misiorowski (.968, a 14-of-19 perfect exit rate for the young flamethrower on the organizational leash) and once again, Zack Wheeler at .944. Five years after 2021, a different manager, and Wheeler is still among the best-hooked pitchers alive. At some point you have to wonder whether hook efficiency is partly a pitcher trait: some guys decline visibly, politely, one batter at a time, and some guys go from cruising to crime scene between pitches.

On the other end of the leaderboard lives Michael Lorenzen, who has left 282 points of Game Score on the table this season (most in baseball) with three perfect exits in 21 starts. Somebody in the Rockies dugout is a deeply patient person.

And the man who started all this? Sandy Alcantara, 2026: a .889 Hook Efficiency, eight perfect exits in 21 starts. Perfectly respectable, above the league line.

However here's the wrinkle I didn't see coming. When I ran the engine back over 2022, Cy Young Sandy's Hook Efficiency was .921, which is Burnes/Wheeler territory. The article I wrote three years ago gently roasted Don Mattingly by imagining a version of him with superpowers, and it turns out regular, non-clairvoyant Mattingly was already managing Sandy's workload at a near-elite level. The 20 innings the Perfect Score shaved off that season were basically the structural minimum, the irreducible ambush tax that no manager in history has avoided.

(Methodology notes: Game Score v2 per Tom Tango (start at 40, +2 per out, +1 per K, −2 per BB, −2 per hit, −3 per run, −6 per HR) computed after every plate appearance from Statcast data, 2015-2026. The Perfect Score is the maximum live GSv2 over each start, with ties going to staying in)