← DiamondMind Research idea — not yet executed

Park × Hitter Type: Validation and Exploitability

Does the sim match real-life extremes for contact hitters in extreme parks? And is “building for your park” by hitter type exploitable?

Abstract Bottom line: Real-life and sim disagree sharply for contact/line-drive types in extreme parks. In ~25,000 archived at-bats, Richie Ashburn hit a home run every 70 AB in Baker Bowl and zero in the Astrodome—never one. In a classic coded league he had a 14 HR season in Guaranteed Rate Park. That suggests the engine may over-credit HR for low-power, high-contact hitters in HR-friendly or neutral parks, and that building for your park by hitter type might be exploitable despite the usual advice not to. This page lays out the research idea: validate engine vs. real-life park × hitter-type outcomes and test whether roster construction by park is a real edge. Status: idea stage; analysis not yet run.

1. Context: Additive vs. Hitter-Type-Dependent Effects

Our park effects work established that in Diamond Mind Baseball, park effects are additive: the same park adds or subtracts a fixed number of 1B, 2B, 3B, and HR per 600 AB for every batter. Baker Bowl adds about +10.5 HR per 600 AB; the Astrodome subtracts about −6 to −7. That model is evidence-based (Tom Tippett’s 2004 SABR presentation; see Imagine Sports Downloads) and fits DMB exports: the additive diff is roughly constant across high-power and low-power hitters, while the percentage impact varies wildly.

On message boards, a researcher (Lemayripper) cited external work suggesting that hitter type—not just raw power—drives how park affects HR totals. One often-quoted result: move a 5 HR/600 hitter and a 40 HR/600 hitter to a park 10% more HR-friendly; the weak hitter increases by about 50% (in percentage terms), the slugger by only about 1 additional HR. So the effect is not uniform; it works in the opposite direction of what many assume (weak hitters get the bigger percentage bump). Batted-ball data that would nail this down weren’t available until Statcast—so for historical seasons we rely on archival park × player splits and sim output.

Ashburn anecdote (Lemayripper). In around 25,000 archived ABs in both Baker Bowl and the Astrodome, Ashburn hits a HR every 70 AB in Baker and has never hit a single HR in the Astrodome—0 in 25,000. In a classic coded league he had a 14 HR season in Guaranteed Rate Park. That real-life vs. sim gap is the core of this research idea.

2. Hypothesis and Research Questions

We propose to test the following:

  1. Engine vs. real life: For known contact types (e.g. Ashburn, Gwynn, Pierre) in extreme parks (Astrodome, Baker Bowl, Guaranteed Rate), do sim outputs match archived park-specific HR rates? Or do we see systematic over-counting of HR for low-power, high-contact hitters in HR-friendly or neutral parks?
  2. Hitter-type definition: In DMB and our data, what proxies for “hitter type”—HR/600, HRF, BA/OBP/SLG profile, K rate, share of outcomes that are balls in play? Can we segment players and compare park sensitivity by segment?
  3. Exploitability: In leagues with fixed home parks, does loading up on contact types in a bandbox (or avoiding them in a cavern) improve outcomes beyond what additive-only valuation would predict? If so, under what conditions—park set, roster construction, league format?

Conventional wisdom says “don’t build for your park.” A reported plan to exploit park × hitter-type effects in round two of the Open suggests that at least some players believe the opposite. We want to see what the data say.

3. Research Path

3.1 Data We Would Use

3.2 Method

Compare real-life park-specific HR rates (and, where available, 2B/3B/1B) for a small set of well-defined contact types and power types to sim output for the same players in the same parks. Focus on extreme parks (Astrodome, Baker Bowl, and a HR-friendly modern park such as Guaranteed Rate). If the sim systematically credits contact hitters with more HR in HR-friendly or neutral parks than real life did, that supports the “over-credit” hypothesis and frames the exploit question. We would then test whether roster construction by hitter type and home park beats additive-only valuation in controlled comparisons.

3.3 Possible Deliverables

DeliverableDescription
Validation note One-pager comparing Ashburn (and 1–2 others) real park HR rates vs. sim, with caveats (sample size, era, park definitions).
Exploit memo If building for park by hitter type appears to beat additive-only expectation, document when it matters and how (e.g. contact types in Baker/GRF; avoid contact-only in Astrodome).
Thread follow-up If we find material gaps, optional board post citing the research and linking to Tippett and hitter-type literature.

4. Why This Matters for DiamondMind

Our valuation and MVP tools use additive park adjustment: the same delta per 600 AB for every batter. If the engine in practice rewards or penalizes certain hitter types more than that model predicts, then (a) our valuations could be off for those types in those parks, and (b) league play could have an exploitable angle that “don’t build for your park” understates. This research idea is the first step to testing both.

5. References and Links

TL;DR — Research idea

← Back to DiamondMind