College Basketball Roster Turnover: Freshmen, the Portal, and the November Discount

The transfer window is now fifteen days long. Rosters turn over by a quarter or more every year. The market prices teams in November that have never played a game together, and it prices them off names.

Post 69 ended on the largest thing an efficiency rating cannot see: whether the team it describes still exists. This post is about that problem.

College basketball roster construction has changed more in the past five years than in the previous fifty. Annual turnover above 25 percent is now unremarkable, and many programs replace the majority of their rotation. Revenue sharing has redistributed which schools can retain talent. And the transfer window has been compressed to a degree that changes how rosters get built at all.

For a bettor, this is not a lament. It is an opportunity with a calendar attached, because roster uncertainty peaks in exactly the months when market pricing is loosest.

01The Scale of Modern Turnover

Start with the magnitude. A typical Division I program now replaces at least a quarter of its rotation every offseason, and it is common for a team to return two starters or fewer. Some rosters are rebuilt almost entirely.

This has three sources operating at once: normal graduation and eligibility exhaustion, NBA departures at the top of the sport, and transfer movement at every level. The third is the one that changed, and it changed the character of the other two by making roster construction an annual free-agency exercise rather than a four-year development project.

The handicapping consequence is that the unit you are betting on is newly assembled. Continuity, chemistry, and system familiarity, which used to be background assumptions, are now variables that differ enormously between programs and that the market prices poorly.

02The Compressed Window and What It Changed

The transfer window has shrunk dramatically and quickly. It ran 60 days in 2023, 45 in 2024, and 30 in 2025. In January 2026 the NCAA compressed it again, and the 2026 men's basketball window ran just fifteen days, from April 7 to April 21, opening the day after the national championship game.

Separate windows exist for coaching changes: a fifteen-day period opening five days after a new head coach is hired or announced. And a rule worth knowing is that athletes who enroll at one school for the first academic term and then transfer mid-year are not eligible to compete at the second school that season.

Why a bettor should care about administrative calendar detail:

• Roster construction is faster and less deliberate. Fifteen days to evaluate, recruit, and sign replacements produces more mistakes, more fits that do not work, and more variance in outcomes.
• Rosters are settled earlier. By late April most movement is done, which means preseason evaluation has a stable target sooner than in prior years.
• Coaching changes cascade. A late hire triggers a second window and can gut or rebuild a roster weeks after everyone else has finished.
• Uncertainty persists around eligibility. Ongoing litigation over eligibility rules has created a genuinely unsettled environment, and roster availability can shift on legal outcomes rather than basketball ones.

03Revenue Sharing and Talent Retention

The revenue-sharing framework that followed the House settlement allows schools to pay athletes directly within a capped pool of roughly $20.5 million per school, with third-party name, image, and likeness deals operating outside that cap. We mapped the football version in Post 59.

Basketball's position within that structure is specific and worth stating plainly. At most schools, football consumes the large majority of the pool. Basketball's share varies enormously by institution, and at basketball-first programs it can be a much larger proportion than at football schools.

Three consequences that matter for handicapping:

Retention is now a resource question. A program that can pay competitively keeps its returning production. One that cannot watches it leave. This is a durable, program-level property that you can learn once and carry forward.

Basketball-first mid-majors have gained ground in specific cases. A program that allocates heavily to basketball can out-compete a football school for a player. This is one reason the traditional high-major and mid-major hierarchy has become less predictive than it was.

None of it is public. Distribution figures are not disclosed. You are inferring from observed retention, not reading a payroll. Treat your inferences accordingly.

04Returning Production as a Predictor

The most useful single roster metric is returning production: the share of last season's minutes, scoring, or possession usage that returns to the team.

It predicts because it captures several things at once. Familiarity with the system, established roles, chemistry, and the simple fact that returning players are known quantities rather than projections.

The practical reads:

• High returning production is a November asset. These teams start faster than the market expects, because everything that takes new teams two months is already in place.
• Low returning production is a November liability and a February question mark. The team may be better by January. It will usually be worse in November than its talent suggests.
• The metric matters most at the extremes. Middling continuity tells you little. Near-total continuity and near-total replacement both tell you a great deal.
• Quality of what returns matters. Returning four bench players is not the same as returning two starters and a primary ball handler.

05The Transfer Integration Curve

A roster assembled from five experienced transfers has more raw talent than a roster of returning role players and usually plays worse in November. The reason is not mysterious, and the shape of the recovery is predictable enough to bet.

What integrates fast. Individual scoring. A player who averaged 18 points at his previous school will generally score in his new system reasonably quickly, because scoring is largely an individual skill.

What integrates slowly. Defense. Defensive rotations depend on communication, anticipation, and habit. Five players who have never defended together will blow coverages in November that they will not blow in February.

What integrates last. Late-game execution and role clarity. Who takes the shot, who gets the ball, and who defers. Teams frequently look disorganized in close games well into December.

The curve produces a repeatable seasonal arc: worse than expected in November, roughly market-neutral in December and January, and better than the season-long rating by February because the rating still contains the bad early games.

06Freshman Impact Is Bimodal

Freshmen do not perform on a smooth curve from good to bad. The distribution has two clusters, and treating it as continuous produces bad projections.

Elite freshmen contribute immediately. The small group of genuinely NBA-caliber first-year players produce from November. Physical readiness and skill overcome inexperience.

Ordinary freshmen struggle more than rankings suggest. The gap between a top-100 recruit and a top-10 recruit is far larger in first-year impact than the ranking distance implies. Most freshmen are inconsistent, foul-prone, and significantly worse on the road.

The specific freshman failure modes are worth naming because they show up in spread outcomes: turnovers against pressure, fouling on defense because college physicality is a step up, and disappearing in hostile road environments. All three are worse in the first two months.

The market's error is systematic. Recruiting rankings are public, they generate preseason narrative, and teams built on highly ranked freshmen are consistently priced closer to their ceiling than their November floor.

07Roster Construction Archetypes

08The November Discount and the February Correction

Pulling the threads together produces the most useful calendar structure in this section.

November and December. Maximum roster uncertainty, minimum sample, softest pricing. Continuity teams are underrated, portal rebuilds and freshman-heavy teams are overrated. This is where roster knowledge pays most and where you have the least confirmation that you are right.

January. Ratings stabilize, the market catches up, and roster-based edges shrink. The season moves to conference play, and the structural angles from Post 67 take over as the primary source of edge.

February. The correction. Season-long ratings now include two months of games that no longer represent the team, which systematically understates improving rosters. This is the mirror image of November and it is the more comfortable side to bet, because you have evidence.

09What the Market Prices and What It Misses

Priced accurately. Recruiting rankings, transfer portal rankings, NBA departures, and named-player reputation. All of this is public, widely covered, and absorbed into preseason numbers immediately.

Priced poorly. Integration timelines, defensive cohesion, role clarity, whether a specific transfer's production travels to a new system, and whether a returning roster's chemistry is an asset or a ceiling.

The pattern is consistent. The market prices inputs well and fit badly. Who a team acquired is public information. Whether those pieces work together is a judgment that requires watching, and watching does not scale to 350 teams.

This is the same asymmetry we identified in Post 69 with ratings, appearing in a different form. Public information is priced. Observation is not.

10Injury Exposure and Roster Depth

Roster construction determines injury exposure, and college basketball reporting is far less rigorous than professional reporting. There is no mandated injury report. Availability news often surfaces at shootaround or in a beat writer's post an hour before tip.

Two structural points from earlier in this section apply directly. A short rotation from Post 68 concentrates value in fewer players, which raises the cost of any single absence. And the foul trouble variable from Post 64 functions as a partial in-game injury, removing a starter for stretches you cannot forecast.

The practical discipline: check availability last, immediately before betting, and treat any pre-confirmation wager on a thin roster as carrying extra risk you are not being paid for.

11Building a Preseason Roster Sheet

For each program in the leagues you follow, one page, built in October and updated through the season.

1. Returning production. Share of minutes and usage returning, and who specifically.
2. Departures. Who left, to where, and what they did.
3. Incoming transfers. Prior program, prior role, and whether the production is likely to travel to this system.
4. Freshman class. Ranked and unranked, with an honest guess at who plays.
5. Archetype. Continuity, portal rebuild, freshman core, or hybrid.
6. Rotation projection. How many players will actually play, and who cannot afford foul trouble.
7. Coaching change. New staff means the installation curve from Post 68 stacks on top of everything here.
8. Your November expectation. Write down whether you expect this team to start above or below its preseason rating, and why. Then check yourself in January.

That last step is what turns the sheet from a summary into a calibration tool. Most bettors never write down a prediction they can be wrong about, which is precisely why they never find out whether their roster reads are any good.

12Common Roster Mistakes

• Adding up individual production and calling it a team. Five 15-point scorers do not produce 75 points. They produce a role conflict.
• Trusting transfer portal rankings as team projections. They rank acquisitions, not fit.
• Treating freshman impact as continuous. It is bimodal. Elite is different in kind, not degree.
• Ignoring continuity because it is boring. The least exciting roster profile is the best November bet.
• Assuming production travels. A high-usage mid-major scorer moving to a high-major usually sees both usage and efficiency fall.
• Forgetting the coaching change multiplier. A new staff plus a new roster is two installation curves at once.
• Betting November bigger because it is softer. Softer numbers and wider personal error are the same phenomenon.
• Carrying last year's roster read forward. Twenty-five percent turnover means last season's team is meaningfully gone.

13Your Roster Workflow

1. Build the sheets in October. Before any numbers exist, when the information is public and the market has not yet been tested.
2. Classify every team by archetype. This alone sets your November lean.
3. Flag the extremes. Highest and lowest returning production in your leagues are your first two watch lists.
4. Watch the portal rebuilds specifically. Defensive rotations are the tell. When they tighten, the team has arrived.
5. Bet small in November. Roster edges are real and your confidence is not yet earned.
6. Reassess in late December. Ten to twelve games in, you can tell integration from talent.
7. Hunt the February correction. Improving teams whose season-long ratings still contain November.
8. Confirm availability before every bet. No mandated injury reports means this is your job, not the league's.

14The Bigger Picture

Roster mobility is usually discussed as a problem for the sport. Coaches dislike it, continuity has suffered, and the annual rebuild has replaced the four-year arc that used to define college basketball.

For a bettor it is straightforwardly good news, and it is worth being honest about why. Markets are efficient where information is stable and public. Roster turnover makes information unstable and makes the important part of it, fit, unobservable without watching. Every year the sport hands you two months in which nobody, including the book, knows what these teams are.

That window is narrow and it is not free. It requires October work before there is anything to bet, and it requires the discipline to size down during exactly the period when the numbers look most beatable. Most bettors do neither, which is why the window stays open.