The market
I blend three public scouting boards — MLB Pipeline's Top 250, the FanGraphs Board, and Perfect Game's Top 400 — into a consensus rank for each college hitter. That consensus stands in for "what the market thinks."
Research project · Baseball analytics
The 2026 Draft, Called in Advance
The day before the 2026 draft's first pick, I locked a ranking: a transparent, baseball-only model's read on 71 drafted college hitters, and where it sits higher or lower than the consensus of public scouting boards. The lock is cryptographically timestamped, so the call can't be quietly changed once the picks are in. It is the baseline for the real question I'm building toward — whether public information about a player's makeup adds anything beyond the numbers.
The question
Can a plain baseball model beat the board?
Draft coverage runs on confident calls that are never written down before the outcome is known. I wanted the opposite: to state, in advance and in public, which drafted college hitters a simple model rates higher or lower than the consensus scouting board, then let the picks and the careers that follow grade it. This page is that record. It is a baseball-only baseline — college performance, age, position, conference — with no makeup or character signals. Those come later; this is the baseline they will have to beat.
Pre-registered & timestamped
Why you can trust the date
A pre-registration is only worth something if it cannot move after the fact. The locked ledger's fingerprint is stamped into the Bitcoin blockchain through OpenTimestamps and tagged in version control, and the proof is downloadable below: anyone can check that this exact ranking existed on July 10, 2026, before a single pick was made.
Locked2026-07-10, one day before the draft's first pick
Git tag
track-b-2026-lockCommit
427aacd3d05ea372c711f2bc5f704f676a2d1a4eLedger hash
sha256: aeb2e46a9c40f13b3d3e… (projections)Proof filesmanifest.json · OpenTimestamps proof (.ots) · methodology
The manifest hash-pins the full per-player ledger and the raw scouting-board snapshots without exposing them. The complete ranking (in both directions) is committed privately and released on the schedule described below — the published hash guarantees it can't be swapped after the outcomes are in.
Where the model disagrees
College hitters the model rates above the board
Of the 71 hitters, these are the ones the model rates most clearly higher than the consensus board. It is a model's opinion, pre-registered and uncalibrated — a ranking disagreement, not a scouting verdict about the players. I'm publishing the higher-than-market side here; the full ledger, in both directions, is timestamped and held for release with the results.
- Cam KozealSS · Arkansas
- Brayden MartinIF · Maryland
- Dalton Wentz3B · Wake Forest
- Rylan LujoOF · Georgia
- Daniel Cuvet3B · Miami
- Kollin RitchieOF · Oklahoma State
- Nu'u Contrades2B · Arizona State
- Alex SosaC · Miami
- Brayden DowdOF · Florida State
- Gavin Gallaher2B · North Carolina
- Carson TinneyC · Texas
- Carter BeckOF · Indiana State
- Carson KerceSS · Georgia Tech
- Tyson LeBlancSS · Kansas
- Owen HullOF · North Carolina
How it works
Three boards, one model, a fixed rule
The model
A two-part model estimates each player's chance of reaching the majors and the value they'd produce if they do, from college hitting, age, position, and conference. It is trained on 2012–2018 drafted college hitters and deliberately knows nothing about where a player was drafted.
The call
For each player I compare the model's rank to the consensus rank. A gap past a fixed, pre-declared threshold flags the player as rated above or below the market. The rule was set before the numbers were seen.
What this is not
The caveats
Four limits worth stating before anyone reads too much into a flag.
- It's a weak instrument. On historical data this kind of test reliably detects only large edges; a modest one would be invisible. Treat the flags as provisional, not proven.
- It's baseball-only. No makeup, character, or intangibles — those are the next phase, and this is the baseline they must beat.
- The flags are uncalibrated opinions of a model, not scouting judgments about people. The "above the board" framing is a ranking disagreement, nothing more.
- Real grades take years. Draft-slot agreement is a first look; the true test is six-season Major League value, which won't mature until roughly 2032.
What happens next
Grading the call
Right after the draft I'll score the ranking against where these players actually go, and I'll release the full two-directional ledger — its hash is already published above, so nothing can be revised. The deeper question stays open: whether public information about a player's makeup adds real predictive value beyond the numbers. This baseline is what that work has to improve on. A result of "it doesn't help" is a real, publishable answer too.