Who should run more, who should stop, and who do opponents run on?

Every sacrifice costs expected runs. It raises the chance of scoring once only with nobody out and a runner already on second, and only if it works: “Bunt must work” is the success rate at which the chance of one run breaks even, from a failed bunt that leaves everyone in place (the lower number) to one that loses the lead runner (the higher). Bunt hits and errors aren’t counted, which would lower it a little. In extra innings the automatic runner makes it the same call, measured on 1,856 extra-inning plate appearances that began with a runner on second and nobody out. A weak hitter at the plate shifts the math toward the bunt; a good one away from it.

Green light: fast (sprint speed 28.3 ft/s or better, the top quarter), attempting under 75% of the league rate for his speed, four or more official attempts, and both his own record and his regressed success at or above the 72% break-even. Success is regressed toward the league rate for his speed with a 92-attempt prior, measured from how success repeats (2024 → 2025), so a fast runner’s regressed rate clears break-even almost whatever his record. Toss-up: fast and rarely runs, but his own record is under break-even; no call and no run value. Below break-even: six or more attempts with regressed success under break-even by 2 or more standard deviations (one is about 4 points, from the prior plus his attempts); under it by less is a toss-up, with no flag and no run value. Too few attempts to judge: fast and rarely runs, but under four attempts. One-run runner: fast, four or more official attempts in 2024–26, and regressed success at or above the one-run break-even; speed alone doesn’t qualify, since every speed bin’s league rate clears that bar. Worth: a green light’s extra attempts (at the league rate for his speed), an upper end since extra attempts are harder ones, or a flagged runner’s 2026 attempts stopped, valued at his regressed success. Counts are official; Savant’s leaderboard misses about a quarter of attempts, so its tracked rates are used only for how often a runner goes. Steals aren’t in the Lineup Lab’s run model; this is where they live.

The plan’s hitters project to +4.9 baserunning runs over 5,329 planned plate appearances (+0.55 per 600). Regressed the same way, the 2026 club’s hitters come to +4.6 over their 6,262 plate appearances with Pittsburgh (+0.44 per 600). Actual 2026: the club ran for +6.8 runs over 6,268 plate appearances, and the plan’s hitters for +5.6 in their own seasons. Isaac Paredes runs −1.2 per 600 where Brandon Lowe ran −0.3: about −0.9 runs over Paredes’s 562 plate appearances, the projection for this swap. Their 2026 actuals were −1.9 and −0.2; the projection regresses three seasons toward average. The plan values hitters on WAR, which already includes baserunning, so this shows that part; it doesn’t change the plan’s wins. Savant baserunning runs per plate appearance, 2024–26 weighted 5/4/3, shrunk toward average with 267 PA of ballast (how the rate repeats from 2024 to 2025: r = 0.61, 265 hitters).

Pittsburgh’s staff: +1.5 running-game runs prevented in 2026 (12th of 30), each pitcher counted for his share with the club. On the 2027 staff, opponents ran most often on Evan Sisk (2.74 per 100 pitches, 2.1× league): candidates for hold and time-to-plate work (we don’t have time-to-plate data to size it). Runs are Savant’s running-game runs prevented (negative = runs allowed). The catcher half of the running game is on the Catching page.