The THR Field Manual
Every claim on this page carries a badge, and every badge is earned: this manual was written against the site's own season-scale backtests — 29,168 graded batter-nights — including the strategies that failed. Nobody else will tell you which of their own features don't pick winners. We just did.
Pregame — what this site actually is
TheHomeRuns.org publishes one thing that matters: a ranked board of every batter's home-run likelihood tonight, graded against real box scores the next morning, hits and misses both, permanently public. Everything else on the site — reports, matchup pages, zone maps, the deck — exists to help you trust, protect, and refine picks that start from that board.
The single most important thing a new user can learn: you do not need to synthesize eight reports to make a good pick. The board already did. Five independent backtests found no simple rule that beats its ordering.
Receipts — why "trust the rank" isn't marketing
The Board Verified
/slate — the ranked list. The edge lives here; everything else orbits it.
Read it top-down. Your card is the top of the board minus vetoes (innings 5–6).
The columns that earn attention: rank, score, edge
(best posted DK line), roster (availability), and the weather flag. The rest is
texture.
Free tier: the top-5 names are open every night, and a logged-in free account can lock one pick a night from them — graded on your locks page the next morning. That one graded pick per night is the entire product in miniature: pick, sleep, get graded.
Row click opens the matchup overlay; the batter name goes to the full profile. Sort, hide, and reorder columns to taste — the board remembers your layout.
What NOT to do here Dead
Don't re-order by any single column — not HR/FB, not form, not park. Don't skip rank 2 for a name you like at rank 9. The composite already weighs eight factors; single-column sorts are how you un-solve a solved problem.
The Clock Verified
The board is alive. Knowing when to read it is a bigger edge than any report.
Yesterday's board graded; recap email; track record updates.
Slate publishes with projected lineups. Emails go out.
Full re-scans. Odds refresh. Board order can move.
Hourly at :45, real batting orders backfill as teams post.
Post-5p re-rank + confirmed lineups. Seal your card here.
Hourly price pulls to first pitch. Shop your legs.
Why 5:30 matters: we snapshotted both boards on the same day. The 1 PM board was running on projected lineups — roughly 1 in 5 of the batters who actually played that night weren't on it yet, and no batting orders existed. The ~5:30p board had 94% confirmed lineups. Draft whenever you like; decide after lineups confirm.
Only have five minutes? Spend them after 5:30, not at breakfast.
Know the Bat
/batter/{name} and /deck — where a rank becomes a person.
The Score Breakdown card on every batter page shows the 8 real factor contributions behind his score — the same math the board runs, not a summary of it. When two bats sit in the same rank band and you have one slot, compare Barrel matchup, Avg EV, and ISO and take the higher. Verified (modest — these are the only factors that showed lift beyond rank in regression)
The Deck (/deck) is the same top of the board as swipeable cards with a
3×3 strike-zone grid. Flip the PITCHER ALLOWED layer on — it defaults off — and the
gold ★ pane marks where his damage zones overlap the arm's leak zones. Treat it as conviction
texture, cell-level accuracy only. Context
BvP history is a story, not a signal — the samples are 5–20 PA. Context
Know the Arm
/pitcher/{name} and /game/{slug} — read the arm for story; pick by the bat.
The Threat Matrix (pitcher page or game page) crosses tonight's lineup against the arm's arsenal families. Its Overall column — the batter's overall barrel rate against the arm's leak family — is the one tab-level number that has shown lift beyond rank within a band. It showed up on both nights we tested… and that's two nights, so it earns exactly tiebreak duty: same band, otherwise equal, take the glow. Plausible
The hard truth about pitcher stats Dead
FIP, xFIP gap, HR/9 — every pitcher-side stat we tested is flat as a selector. The "attack the bad arm" thesis has been killed four separate times in our backtests. A leaky pitcher does not turn a rank-90 bat into a rank-20 bat. Read the Leak Meter and Susceptible lists as scouting narrative and as a funnel for which Matrix to open — never as a reason to pick a lower-ranked bat.
The Supporting Cast
Six tools, each with one honest job.
Weather (/reports/park-weather, game Conditions tab) Veto
One job: kill legs in extreme wind-in games (the −5 Wrigley-class blow-in). Mild +2 boosts do not select — ignore them. Roofed games: skip the check entirely.
Dueness (/dueness) Veto
One narrow verified band: a bat at >3× his own median HR gap is broken rhythm — swap him for the next same-band name, unless he's a genuine quality bat (they're exempt). Everything else on that page — the ≥1.2× "overdue" chip, the due-next rail — is browsing, not signal. "Due" is never a reason to add anyone.
Odds & the EV floor (edge column) Verified
Odds do two jobs, both defensive: drop any leg quoted below the 102 EV floor, and shop the best posted price. Honest framing: at evening prices, no slice of the board has shown positive EV — the edge you keep is the vig you don't pay. The site snapshots each morning's line; whether early prices beat late ones is still being measured.
Also in the cabinet: /hr-explorer (every HR plotted — great for understanding, no pick value), /form (display-only by doctrine; see the 7th), /reports/splits-edge (platoon corroboration for a pick you already made Context).
Combos That Compose
Ways to use two or three surfaces together — each badged with what the data actually supports.
① One More Leg Verified
/slate → lock. Tempted to fine-tune your card? Take the next rank down instead. Hits scale ~linearly with legs (verified: 10→20 picks doubled hits, 1.72→3.38/night). One added leg beats any clever rearrangement. The strongest verified result on the site.
② The Confirmed-Lineup Final Pass Verified
5:30p /slate → game page lineup → roster column. Confirm every leg is on the board, in the posted lineup, and healthy — the roster column can lag IL moves. Removes the largest pure-loss category: legs that never bat.
③ The Threshold Stack Verified shape only
Chasing a 3+ HR night (parlays, bragging rights)? Concentrate same-band legs into fewer games — the feasible 3 bats × 3 games shape adds ~5% to P(3+) and nothing to average hits. Choosing which game to stack by arm quality is a coin-flip tiebreak, not evidence Context. Iron rule: never demote a better-ranked bat to make a stack.
④ Susceptible → Matrix → Splits Plausible
Use the Susceptible list as a funnel: find tonight's leaky arms, open the Threat Matrix, and where two same-band bats are equal, take the one glowing in the Overall column; cross-check the platoon edge on Splits-Edge as corroboration. Two-night evidence — tiebreaks only.
⑤ The Bat-Side Tiebreak Verified
Same band, one slot: compare Barrel matchup / Avg EV / ISO on the two batter pages, take the higher. Modest but real, and it's the only tiebreak with season-scale evidence.
Plus the two standing vetoes (weather, broken rhythm — 5th inning) and price discipline (5th). That's the complete kit. Anything else you're tempted to combine is probably in the next inning.
The Traps Dead
Four strategies users reach for that our own backtests killed. The site shows you these numbers because they're interesting — not because they pick winners.
| The move | Feels like | The data says |
|---|---|---|
| Hot-hand chasing /form bands, Last-5, streaks |
"He's locked in" | P(HR | HR yesterday) = 21.1% vs 19.9% base — a rounding error |
| Game shopping park factor, game verdict card |
"Coors tonight!" | Park factor sign-flips as a selector; game-verdict ↔ actual HRs correlation ≈ 0 |
| Arm-fishing susceptible/meltdown lists as pickers |
"Anyone vs this guy" | FIP / xFIP-gap / HR-9 all flat in regression — killed 4× |
| "He's due" dueness as a buy signal |
"Overdue = tonight" | 1.5–3× overdue is noise in both directions; only >3× matters, as a fade |
Also tested and dead: batting-order vetoes (+1 hit per 65 nights), rank reshuffles (the 16–30 band looked hot for two straight nights, then lost to 1–15 over the season — two-night patterns lie).
Build the Card Verified
Construction math, from our own correlation measurements.
- Size first. Card size is the single strongest lever you control. Decide how many legs your format/bankroll supports, then fill top-down by rank. Adding a 10th leg to a 9-leg card raises P(3+ hits) by ~23% — more than any rearrangement of the nine.
- Spread by default. Chasing "at least one hit"? One bat per game maximizes it.
- Stack only for thresholds. Chasing 3+? Fewer games, more bats per game (③ above). It reshapes the tails; it never raises the average.
- Never trade quality for correlation. A worse bat in the "right" game is a net loss. Every time.
Receipts — the 9-for-9 math, since you'll wonder
The Grade
/results, /track-record, your locks page — the habit that makes you better.
Grading closes at ~5:00 AM: yesterday's board, your locks, the season record — misses included, forever. Two habits worth forming:
- Morning, 2 minutes: your locks page. Pending ≠ miss — postponed games grade later.
- Weekly, 5 minutes (/model-history for weight drift, case studies for pattern literacy): the track record. You're calibrating trust, not picking.
Locked picks are frozen at lock time and graded as-locked. Post-seal line moves and late news are data for tomorrow, not reasons to re-pick tonight.
Tonight, in Seven Steps
- After 5:30p ET, open /slate. The confirmed-lineup board is the real one.
- Set your card size — then fill top-down by rank. Size beats cleverness.
- Veto pass: extreme wind-in games, >3× broken-rhythm bats (quality bats exempt), anything under 102 EV.
- Tiebreaks only: same band → higher Barrel/EV/ISO; Matrix Overall glow.
- Confirm every leg is in the posted lineup and the game isn't postponed.
- Lock it. Free tier: lock your one nightly pick from the top-5.
- Tomorrow, check the grade. Hits and misses. That's how this works.