Season line
14 starts, 8-2, 1.79 ERA, 0.95 WHIP, 85.2 innings. He's struck out 95 over that span, against 26 walks. This is the baseline the matchup adjusts from — the model starts here and then asks what the specific lineup he's facing does to strikeout arms, how the park plays, and how deep he's likely to go.
He throws right-handed
Shohei throws right-handed, which is the ordinary case and means fewer wild swings in the lineup he faces. Opposing managers aren't restacking against him the way they would a left-hander, so his projection moves more on the quality of the lineup than on its handedness. What does move it: how deep he's been going lately, and whether the lineup in front of him strikes out.
Strikeouts and control
Shohei has 95 strikeouts and 26 walks in 85.7 innings this season — 10.0 K/9 against 2.7 BB/9, a 3.7-to-1 ratio. That's around league-typical control — enough to work deep when the lineup cooperates, not enough to bail him out of a bad matchup.
What his floor looks like
Shohei has made 14 starts this season, averaging 6.1 innings with a 1.79 ERA across them. In the 0 starts we hold detailed lines for, he got through five or more innings 0 times. That ratio is the honest read on his floor, and it's what separates an arm worth projecting on the outs board from one whose good starts are unpredictable enough that the market prices them better than we do.
How the model reads him
The number on Shohei is about tonight, not about him. Strip out the name and the model is asking one question: what does this specific lineup do to this kind of arm, in this park, for as many batters as he's likely to face? That gets scored against every other starter on the slate. The combination math is proprietary and re-tuned every week. The grading isn't — it's all public, including the misses.
What we'd flag
Books don't price every player — coverage tops out around 87% even for confirmed starters, so "no line" here usually means no book posted one rather than a gap in our data. Lineups change late, and an afternoon projection can be stale by first pitch. And the model is tuned across hundreds of graded outcomes, so one game tells you nothing about whether it's working.