Beta · pedigree-v3How the Program Battle ranking works
The Program Battle Arena ranks Division I baseball programs by their all-time pedigree and lets any two of them go head-to-head. This page states everything behind the number: every weight, every formula, every source, and what this beta does not yet do. Nothing here is a black box, and nothing is invented — each input is a public-record fact with a citation.
The Program Power Index
Each program gets a 0–100 score: a weighted blend of six all-time facts. The weights are fixed and published — they do not change from one program to the next.
| Input | Weight | Why it carries that weight |
|---|
| National championships | 34% | The rarest, least ambiguous proof of an all-time peak — the heaviest input by design. |
| College World Series appearances | 18% | A sustained elite ceiling — repeat trips to Omaha, less single-title luck. |
| NCAA tournament appearances | 18% | Decades of postseason access — résumé depth beyond Omaha. |
| All-time win percentage | 10% | Era-neutral program quality — down-weighted because it carries a schedule confound we will not fabricate around (see below). |
| All-time wins | 10% | Longevity and scale — capped so an old program is not automatically a great one. |
| Recency (last Omaha trip) | 10% | A tiebreaker in an all-time ranking, not a co-primary signal. Want the current-form view? Sort the leaderboard by Recent. |
| Conference championships | reserved | Reserved — not yet sourced; weighted 0 until it is. |
The six active inputs sum to 100%. One further input — conference championships — is built into the engine but held at 0% until its data is sourced and cited. This version is labeled pedigree-v3: an all-time ranking. Its weight sits on the schedule-independent résumé — championships, Omaha, and tournament access — while recency is a tiebreaker, not a co-primary signal. If you want to see who is hot right now, the leaderboard sorts by any pillar, including Recent.
The formulas
- Counts (titles, Omaha trips, tournament appearances, wins) use a concave curve (γ = 0.62) against a pinned all-time reference: 12 titles, 38 CWS appearances, 60 NCAA tournament appearances, 3,600 wins. A program at the reference scores 100 on that input; beyond it, the score holds at 100. The curve means each additional title is worth a little less than the last, but it is deliberately steeper than a plain square root so that a stack of titles (eight, twelve) reads as clearly more than one — the rarest signal should not be flattened.
- All-time win % maps straight from 0.400 (→ 0) to 0.700 (→ 100). It is intentionally one of the lighter inputs: a raw win rate cannot tell a powerhouse schedule from a soft one, and we will not invent an opponent-strength adjustment from records we do not have. So we weight it modestly and say so here rather than dress it up as something it is not.
- Recency scores how recently a program last reached the College World Series — a trip to Omaha this decade scores high, a last appearance generations ago scores low. In an all-time ranking it is a tiebreaker, not a co-primary signal: it nudges programs of otherwise similar pedigree, but it does not bury a multi-title blue-blood for being dormant. To foreground current form instead, sort the leaderboard by Recent.
- The reference caps are pinned, not set to whoever leads today. A program's score depends only on its own record — adding or removing another program never changes anyone else's number.
- Head-to-head probability runs the two scores through a logistic curve (k = 4). A 25-point gap reads about 73/27; a 50-point gap about 88/12. Equal scores are a coin flip. No home-field bump — this is a neutral, all-time matchup.
Confidence, not false precision
Every rating carries a band and a letter grade. A program with all six active facts on record earns the tightest band and an A. A program missing a fact — say, a champion whose all-time win-loss record we haven't sourced yet — gets a wider band and a lower grade, computed from exactly what's missing. We never fill a gap with a guess; we widen the band and say so.
Fact and model, kept apart
In a head-to-head, if a real all-time series record exists between two programs, that record is the headline — reported exactly, with its source. The win probability sits next to it, clearly labeled as our model. When the two disagree — the underdog owns the series but the model favors the other side — we show both and say they disagree. The model never overwrites the record.
Reading the leaderboard
- Every rank shows its reasoning. Beside each program is a six-bar strip — one bar per pillar (titles, Omaha, NCAA tourneys, win%, wins, recent), each as tall as that program's 0–100 score on it. The single number is the verdict; the strip is the argument behind it.
- Sort by any pillar. The default order is the Program Power Index. Re-sort by titles, Omaha trips, tournament appearances, win%, wins, or recent form and the board re-cuts instantly — the underlying scores never change, you are just looking from a different angle.
- Partial records are labeled, not hidden. A program whose all-time win-loss we have not sourced from a cited total yet is marked partial, ranked on the facts we do have, and carries a wider band. The "complete records only" filter hides them when you want just the fully-sourced set. A program missing the pillar you are sorting by sinks to the bottom with a dash — never a guessed value.
Where the numbers come from
- National titles — NCAA year-by-year champions (1947–2026), corroborated against the Wikipedia CWS-appearances table. Where two sources disagreed on a count, we recorded the discrepancy and used the year-by-year list.
- College World Series appearances — Wikipedia's CWS-appearances-by-team list, current through the 2026 tournament.
- All-time win-loss records — Wikipedia's NCAA Division I team-statistics table, current through the 2024 season.
The full provider list is on the data sources page.
What this beta does not do yet
- It rates effectively every tracked Division I program, not just the ones with Omaha pedigree. Programs with real championship or College World Series history but an unsourced all-time record are partial; programs that have never reached Omaha rate stub — a thin résumé, honestly reported, never a guessed number.
- Win % is not schedule-adjusted, so it is weighted lightly: a record cannot, on its own, tell a power-conference slate from a soft one, and we will not fabricate an adjustment.
- Conference championships are weighted at 0 until their records are sourced and cited — that single résumé input is the next to come online.
- Recency is measured only by a program's most recent Omaha trip, not yet by full last-decade results or live RPI. A finer recent-form signal is the next refinement.
- There is no all-time Elo. That needs a complete, ordered game log we don't have; building one would mean inventing the path, so we don't.
- Reference caps are pinned to current all-time leaders. If a program passes a cap, its score holds at 100.
What changed in pedigree-v3
The first version spent nearly a fifth of every score on how recently a program reached Omaha. For an all-time ranking that was too much — it pushed the sport's most-decorated programs down for being dormant and lifted recent one-offs above them. pedigree-v3 cuts recency to a tiebreaker, trims the un-schedule-adjusted win rate, and moves that weight onto championships, Omaha trips, and tournament access — the parts of the record that actually describe all-time pedigree. The title curve was steepened so a stack of titles separates from a single one. The current-form view didn't disappear: it moved to the leaderboard's Recent sort.
This is a prototype. The ranking logic, scope, and accuracy are still being improved — that is what the Beta label means. Found a number that looks wrong? It traces to a citation; tell us which one and we'll reconcile it.
Methodology version pedigree-v3.
