Projections | CCAA Baseball 2025-26

—

2027 Team Projections ADV

Returning production model · CCAA-calibrated

These are returning-production projections at the team level: the model estimates each program's 2026 talent, works out how much of that production graduates, applies a growth curve to the players who come back, and assumes departing production is partially replaced. It reads one season of aggregate output, which is a different and coarser thing from the player model in the Players tab.

The growth factors in step 3 below are still the reasoned values this page launched with. The four-season archive can now measure them directly, and the Players tab does. Reconciling the two is the next job on this page.

Every input is CCAA-calibrated. WAR comes from the same oWAR and pWAR figures used across the site, which recalibrate to league average on every page load. A projection is a central estimate, not a prediction of what will happen.

⚠

—

Team Projections

Projected 2027 Finish

Ranked by projected winning percentage

Returning % is share of 2026 playing time
coming back, not share of WAR

Top Returning Hitters

Top 5 Returning Hitters ADV

Ranked by 2026 oWAR

Top 5 Returning Pitchers ADV

Ranked by 2026 pWAR

Top Returner by Team

Best Bat and Best Arm Coming Back

One hitter and one pitcher per program

A two-way player can lead both columns
and is marked as such

Methodology

How These Are Built

Five steps · all constants shown

Estimate 2026 talent. Each team gets two independent reads: its actual winning percentage (ties count as half a win, NCAA convention) and a WAR-based estimate that starts from a replacement-level team and adds the total oWAR and pWAR its players produced.

WAR win% = .300 + (total WAR × scale ÷ games)

The scale factor is not assumed. It is solved for on every page load so that league-wide WAR reproduces league-wide wins above replacement, which keeps the WAR estimate on the same footing as actual results. Current value: —. The two are blended, then regressed toward .500. Regression is the honest admission that a 20-game high school season is a small sample.

Split returning from departing. Seniors leave; freshmen, sophomores and juniors return. Players with no listed class year are treated as departing, which is the conservative assumption.

Grow the returners. High school players improve year over year, and younger ones improve fastest. Returning WAR is multiplied by a class-based growth factor: freshmen ×1.35, sophomores ×1.25, juniors ×1.10. These are reasoned starting values, not empirically derived, and are the single most replaceable assumption in the model.

Replace what graduated. Programs reload at different rates. The model assumes 45% of departed WAR is replaced, scaled by program strength so stronger programs backfill better.

Convert to a record. Projected WAR becomes a winning percentage, blended once more with the regressed 2026 talent estimate, then applied across a season of —, the league median in 2026. A final zero-sum correction shifts every team so the conference averages .500, since the CCAA mostly plays itself and cannot collectively improve.

⚠

Known limitations. There are no park factors, no strength-of-schedule adjustment across the three leagues, no defensive component in WAR, and no accounting for incoming freshmen or transfers. Growth and reload rates are assumed rather than measured. All of these improve once a second season of data exists.

2027 Player Projections ADV

Frozen forecast · CCAA-calibrated

Bands are 20th to 80th percentile
drawn from comparable players

Every returning player in the conference, projected forward one season. Each component of a hitter's line is projected separately and regressed by its own measured reliability, then aged on a curve derived from the archive rather than assumed. This is a different animal from the team model in the other tab, which works off a single season of aggregate production.

These numbers are frozen. They will not move when a stat update is pushed. That is the point: a forecast that quietly rewrites itself cannot be checked, and this one is meant to be checked in June against what actually happens.

⚠

—

Projected 2027 Players

Projected Hitters

—

Class '27

—

Tap a hitter to see the comparable players his band was built from.

Does It Work?

Backtest Scorecard

Held-out seasons · lower error is better

The model was rebuilt without
2025 and 2026, then scored against them

→

—

Methodology

How the Player Model Works

Six steps · every constant measured

Break the line into components. A hitter is not projected as one number. Walk rate, strikeout rate, home run rate, batting average on balls in play and extra-base share are each projected separately, then reassembled into a full stat line. This matters because the components are not equally trustworthy: strikeout rate is half signal at — plate appearances, BABIP not until — balls in play. Averaging them into one figure and regressing them all the same amount throws that away.

Weight four seasons. Each prior season is expressed relative to its own league environment, since the CCAA run environment moves year to year, then weighted — from most recent to oldest and by playing time within each season.

rate = Σ(weight × PA × rate/league) ÷ Σ(weight × PA)

Regress each component on its own constant. Using the harmonic form n/(n+k), which never reaches full credibility. That is deliberate and it is the opposite of what the leaderboards do: the stats pages describe what a player did, so playing time earns full credit, while a projection estimates what he is, and no high school sample is ever fully convincing.

Age the result. Measured from every matched player-season in the archive, pooled rather than averaged per player so a single sophomore who hits three homers cannot write the curve. Shown below in units of league average, after the calibration scalar of —.

Non-seniors return at —, so survivorship is a mild problem here rather than a fatal one.

Project playing time separately. Rate and opportunity are forecast independently, as they should be. Fit across every returning player in the archive:

—

Correlation is —, so this is the loosest part of the model. A player who wins a job takes a jump nothing here can see coming.

Build bands from comparables, not from a bell curve. Each hitter is matched to the 25 most similar historical player-seasons on strikeout rate, walk rate, power, BABIP, playing time and class. The 20th and 80th percentiles come from what those players actually did the following year. A profile the model has historically struggled with gets a wide band automatically, with no one deciding it should.

⚠

Known limitations. No park factors yet. No incoming freshmen or transfers, so a program's real 2027 roster will be larger than what appears here. Class year stands in for age, which misses anyone who repeated a year. Pitcher projections are weak enough that the scorecard above says so outright. And a player who does not return at all is projected as though he will, because nothing in season totals predicts quitting.