Home Run Derby Draft The HR Almanac

Home Run Almanac

May 10 – May 16, 2026
« All weeks | | Table view »
42
HRs by top 30
21
Picks that homered
58%
Top-15 capture
58%
Top-30 capture
I

The Featured Half-Dozen

Top six projections of the week
II

The Full Slate

Click any row to reveal the game-by-game outlook
Show:
#v Playerv Clubv Bv Gv Proj HRv Sea HR/Gv 21d HR/Gv 7d HR/Gv Barrel %v Qual xv Tagv Act HRv
III

The Patches

Hottest hands · regression candidates

Hot Hand

Three or more home runs in the last seven days. Empirical +2.6% vs baseline.

    Due to Pop

    Four or more in the last three weeks — but one or zero in the last seven. Empirical +1.9% vs baseline.

      Cold — Avoid

      No HR in the last seven AND under three in the last twenty-one. Empirical −58% vs baseline. Rate is downweighted in their score.

        Under the hood

        Rank hitters by season home runs — the obvious play — and you capture 44.9% of a given week's long balls. We capture 48.8%. That four-point edge is worth roughly 36 extra homers over a 21-week season, and it survives 21 weeks of 2025 backtesting with zero look-ahead. None of it is a magic stat. It's three decisions, each one paid for in the data.

        1. Stabilize the rate. Every projection starts from a weighted HR/G: 0.50·season + 0.35·(21d) + 0.15·(7d). The two recent windows aren't taken raw — they're Bayesian-shrunk toward the hitter's season rate against a 25-game prior, so 3 bombs in 6 games regresses hard instead of extrapolating to a 60-homer pace. The 7-day window is deliberately starved at 15%: ranking by last-7 HR alone captured just 39.5%, barely above noise. Recent power doesn't predict — demonstrated skill does. The blend caps at 0.55 HR/G; nobody sustains peak Bonds.

        2. Adjust for the matchup — at quarter strength. That base rate is scaled by park (split by batter hand), the platoon rate versus each probable's throwing hand, the starter's HR/9 (blended 50/50 with his last three starts, then 55/45 against the bullpen behind him), and weather (+0.2% per °F over 70°, a bump for 15+ mph winds blowing out, nothing under a roof). Then the whole multiplier stack is raised to the 0.25 power. Full-strength adjustments lowered capture in testing — the matchup is already half-baked into a hitter's recent line, so we bank a quarter of the swing and no more.

        3. Tax the slumps. The single biggest lever in the model: any hitter with ≤1 HR over 7 days and ≤2 over 21 gets multiplied by 0.6, because slumping bats come in ~58% under their own rate the following week. That one rule added +2.3 points of capture by itself. The only reprieve is contact — barrel rate (min 15 batted balls) scales the projection from ×0.90 under 6% to ×1.20 above 17%, and a hitter still scorching it at 105+ mph beats the cold tag on the evidence that the homers are coming.

        The line: 48.8% / 48.1% / 55.2% capture at the top 15 / 30 / 100, every week graded against what the league's best bats actually did. The secret was never a stat nobody else has. It's regression applied without flinching — shrink the small samples, tax the slumps, and refuse to pay full freight for a matchup.