How HoopAtlas works
Method and live numbers from this server's database. Last read: 2026-10-31 15:30 UTC.
The board
5,556 players on 366 Division I teams, from ESPN's 2026-27 season rosters. Each player carries his latest season line from ESPN, advanced stats computed from ESPN season totals and team totals (usage, true shooting, assist and turnover rates, per-40, shares of team games and minutes), and his season history since 2021-22 (37,396 player-season rows). A value ESPN does not give is left empty and shown as "Unknown".
Pro outcomes
To learn what predicts a pro career, we recorded where former D1 players went. NBA seasons come from each player's ESPN record (ESPN uses one id for college and pro). European rosters come from the leagues' official sites or APIs and are matched to the college player by name and birth date (or, when ESPN has no birth date, a single same-name college player of the same height).
| League | Seasons | Roster rows | Former D1 players found |
|---|---|---|---|
| ABA League | 2022-23 to 2026-27 | 1,377 | 39 |
| Basketball Champions League | 2022-23 to 2026-27 | 2,482 | 155 |
| EuroCup | 2022-23 to 2026-27 | 1,868 | 143 |
| EuroLeague | 2022-23 to 2026-27 | 1,688 | 66 |
| FIBA Europe Cup | 2022-23 to 2026-27 | 3,292 | 193 |
| Israeli Basketball Super League | 2022-23 to 2026-27 | 1,344 | 111 |
| NBA | 2022-23 to 2025-26 | 757 | 306 |
Not covered: Spain (ACB), Italy (LBA), Germany (BBL), France (LNB), Greece, Turkey, and second divisions. Their terms don't allow reuse, their data sits behind a private API or a bot check, or we haven't added them yet. So the Europe rates below are lower bounds.
OPP v2: Overseas Pro Potential
An L2-regularised logistic regression on a player's final college season: advanced stats, height, years in college, his team's point differential and whether the team is in a high-major conference. It is fitted on real outcomes and checked by leaving one final season out at a time, so every number below is on players the model did not see.
opp-v2.1
| Outcome | NBA games, EuroLeague or EuroCup roster within 2 seasons after the last college season |
|---|---|
| Training players | 4,193 D1 leavers (341 reached the outcome), last college season 2021-22, 2022-23, 2023-24, 2024-25, at least 200 minutes |
| Validation | leave one final season out |
| AUC (held out) | 0.952 (old hand-tuned OPP v1.1 on the same players: 0.789) |
| Brier score | 0.0396 (base rate alone: 0.0747) |
| AUC, high-major | 0.898 on 913 players (v1.1: 0.656) |
| AUC, other D1 | 0.934 on 3,280 players (v1.1: 0.840) |
opp-v2.1-europe
| Outcome | European first-division roster (EuroLeague, EuroCup, Basketball Champions League, FIBA Europe Cup, ABA League, Israeli Basketball Super League) within 3 seasons after the last college season, among players with no NBA games in those seasons |
|---|---|
| Training players | 2,647 D1 leavers (197 reached the outcome), last college season 2021-22, 2022-23, 2023-24, at least 200 minutes |
| Validation | leave one final season out |
| AUC (held out) | 0.839 (old hand-tuned OPP v1.1 on the same players: 0.771) |
| Brier score | 0.0596 (base rate alone: 0.0689) |
| AUC, high-major | 0.718 on 433 players (v1.1: 0.684) |
| AUC, other D1 | 0.809 on 2,214 players (v1.1: 0.768) |
Calibration of the Europe model (held out): predicted against what happened.
| Predicted band | Players | Predicted | Actual |
|---|---|---|---|
| 0–1% | 694 | 0.4% | 0.7% |
| 1–3% | 624 | 1.9% | 1.8% |
| 3–10% | 733 | 5.7% | 5.7% |
| 10–30% | 442 | 17.1% | 18.6% |
| 30–100% | 154 | 43.5% | 37.0% |
Europe tiers
A tier is a band of the Europe probability. Next to each band is how many past players in it actually reached a European first division within three seasons. That measured rate is what the app quotes, not the model's own estimate.
| Tier | Band | Past players reached Europe | Current players |
|---|---|---|---|
| tier 1 high | 30%–100% | 37% (57 of 154) | 117 |
| tier 2 medium | 10%–30% | 19% (82 of 442) | 347 |
| tier 3 low | 3%–10% | 6% (42 of 733) | 643 |
| tier 4 remote | 0%–3% | 1% (16 of 1,318) | 1,082 |
There are no salary or NIL figures: we have no source for them.
Ground rules
- No invented values. Every stored number comes from a named source or is calculated from one by versioned code.
- Birthplace is not a passport. Bosman and Cotonou status stay "Unverified" until a person confirms them.
- AI-written text is labelled as AI, names the model, and is generated only from sourced inputs.
- We skip any source whose robots.txt blocks our crawler, whose terms forbid reuse, or that is behind a login, paywall or bot check.