Variance is real: even a correct read misses about 1 night in 3. That's what the distribution says, not a sales pitch.
▸Key: what these terms mean
Big night (8+ K)
The chance this pitcher strikes out 8 or more. A genuinely dominant start, and the colour of the ring. The board is ranked by projected strikeouts, highest first. green = strong chance tonightyellow = borderlinered = low chance
✕ next to a name
Early-hook risk: something caps his night, a pitch limit, a quick manager, or an injury-return leash, so the monster strikeout night is blocked. Tap the row for the reason.
Projection
Our single best guess for tonight's strikeouts: the average of all 40,000 simulated starts, adjusted by how accurate our numbers have actually been this season.
Ceiling (top 10%)
His 90th-percentile night: one start in ten reaches this many strikeouts or more. The max-out number.
Tonight's form (0–10)
How good his stuff looks coming into tonight, from observable signs only: last-start velocity vs his norm, recent two-strike whiffs, rest days, how often the other lineup chases, and how long his manager lets him work. 5 is neutral; 7+ is rare (a couple per night across the league) and means the sim expects his best-stuff version. Hover or tap the bar for the ingredients.
Projections and accuracy grades only: we publish numbers, not picks.
The shape of his upside. low ceiling = little room above his typical night; normal upside = typical spread; big upside = real blow-up-the-box-score potential; second gear = when his best stuff shows up, the night jumps a whole tier.
Confidence (0–100)
How much the model trusts its own read. It drops when inputs are missing, the lineup isn't posted yet, or the pitcher's sample is thin. Under ~55 means treat the read loosely.
Big night blocked
Something concrete rules out the monster night no matter how good his stuff is: a pitch limit, a manager who pulls starters early, too many walks, or a lopsided matchup. The reason is on the badge.
Reach the total
Every rung is the plain chance he reaches that many strikeouts. 7+ Ks · 71% means 71 of 100 simulated starts had 7 or more. Tonight's line and our chance to go over it are here too. Advanced adds the market comparison: prices, the de-vigged implied chance, and the gap between it and ours.
Strikeout ladder
The chance he reaches each level, level by level. 4+ Ks, 5+, 6+, all the way up. Tap it open on any pitcher to see the whole staircase.
Lineup badge
lineup ✓ = the official card is posted and the sim used the actual nine hitters. no lineup= using the team's overall strikeout profile until the card drops (usually 2–4 hours before first pitch).
Arsenal matchup
Whether his best swing-and-miss pitch attacks a lineup that already whiffs a lot. A real out-pitch versus chase-happy bats nudges the projection up; a flat arsenal against contact hitters nudges it down.
Recent form
A gentle shade on the rate when his velocity or two-strike whiffs are trending away from his season norm, so an arm whose stuff is fading projects lower before the averages catch up.
Calibrated
The probabilities are re-mapped through a backtest of thousands of past starts, so 40% means what it actually did historically. Not just what the raw sim guessed. See the Results page for the curve.
Risk index (0–100)
How the night can go wrong, scored from six named ingredients: the chance of a dud (3 K or fewer), the early-exit chance, night-to-night swing, blind spots in the data, whether history agrees with the model, and any hard limits tonight. Low / moderate / elevated / high. High big-night % with low risk is the read you want; high % with high risk is a trap.
Scenarios
The three ways tonight actually plays out. His best stuff shows up, an ordinary night, or a rough early exit. Each with its probability and typical strikeout total.
Correlated
You have lines on both starters in the same game. One game script. A blowout, rain, a bullpen day: can sink both at once. Not a prohibition, just exposure you should see.
Missing inputs
Data the model wanted but couldn't get tonight (named exactly). The sim used neutral defaults for those pieces and confidence took a haircut: nothing is ever made up.