By Full Spectrum Sports Intelligence ·
How we tested MLB player-prop legs and same-game parlays against real pregame prices on 30 held-out 2026 game days, and what we found.
About this study
Full Spectrum Sports Intelligence research, October 9, 2026. Summary post: Client Support research post. This is a research result, not a pick list.
Result in one paragraph
We tested our MLB player-prop logic on 30 held-out game days from the 2026 season (random draw, seed 20261009). It picked 775 player legs. They won 55.5% of the time, versus 56.3% implied by Kalshi prices. The logic did not pass our gate, so it stays in research. 2025 props could not be tested because Kalshi listed no MLB player markets in 2025.
What was tested
- Strikeout legs: a projection from each starter's strikeout rate and innings per start, using only games before the game date. A leg leans over or under when the projection is at least 0.6 strikeouts away from the posted line.
- Hits and total bases legs: a simple rule from the same pitcher inputs. Hitters facing a starter with a high (shrunk) ERA lean over; hitters facing a starter with a low ERA lean under.
- Same-game parlays: in each game, legs are ranked by strength and nested tickets of 2 to 15 legs are taken from the top.
How it was tested
- Prices: every leg is priced at the last real pregame Kalshi price, including Kalshi's fee. No prices were filled in. No free source posts same-game parlay prices, so a ticket's price here is the leg prices multiplied together. It is not a sportsbook SGP price.
- Results: settled from MLB Stats API box scores.
- No look-ahead: every input is dated before first pitch.
- Splits: the 30 held-out days were drawn first and never used for tuning. Tuning used June and July 2026; a separate check used August and September 2026.
- Tuning rule, declared in advance: a change is kept only if it improves the check period and does not hurt the tuning days.
Results on the held-out days
- Strikeouts: won 59.4% vs 61.7% implied (320 legs)
- Hits: won 53.2% vs 56.4% implied (235 legs)
- Total bases: won 52.3% vs 48.5% implied (220 legs)
Total bases ran 3.8 points above its price, but the 95% range (−1.2 to +8.7 points) includes zero.
Same-game parlays on the held-out days (hit rate vs the leg prices multiplied together):
- 2 legs: 36.6% vs 35.2% (134 tickets)
- 3 legs: 20.9% vs 17.2% (86 tickets)
- 4 legs: 11.8% vs 9.2% (76 tickets)
- 5 legs: 3.2% vs 5.1% (63 tickets)
- 6 legs: 2.0% vs 2.6% (50 tickets)
- 7 to 12 legs: no hits (84 tickets)
3- and 4-leg tickets hit slightly more often than the leg prices multiplied together predict. That is expected: legs in the same game move together, and books price that in. The tickets also overlap, since each 3-leg ticket contains a 2-leg ticket.
Where the legs lost
- Pitchers pulled before the 4th inning: strikeout legs won 44.1% vs 61.6% implied across all priced legs, April–September (195 legs). This is only known after the game, so the next step is a pregame early-pull warning.
- Hitters with 1 or 2 plate appearances: 46.1% vs 52.4% implied across all priced legs, April–September (115 legs). The next step is to wait for confirmed lineups.
- Hitters batting 7th to 9th: 50.8% vs 51.9% implied across all priced legs, April–September (740 legs). Leaving them out did not hold up on the held-out days, so that change was rejected.
Tuning
49 single-rule changes were tried, one per tuning day. None was kept. Two improved the check period but hurt the tuning days, which points to noise. Swapping FIP for ERA in the hitter rule made no real difference.
Limits
- 30 held-out days is under our 40-day bar for any claim.
- Injury-list status, home-run and stolen-base props, and postseason games were not tested.
- Same-game parlay prices are not published by any free source, so ticket prices here are leg prices multiplied together.
Price inputs
Kalshi (pregame player-prop and game prices) and DraftKings closing moneylines via ESPN. Results from the MLB Stats API.