Penn at Columbia
A pregame research note built from the stored 2026–27 forecast, its stored factors and the questions still waiting for film or roster confirmation.
Forecast edition basketball-efficiency-v2-ab7f181f3f19 · captured Sep 23, 2026
Penn 74.5 · Columbia 77.2
Primary preseason estimate from the published efficiency model. The range describes held-out model error, not a promise about the final score.
What this game analysis uses.
Read the forecast, historical workload, shot map and roster signals in separate columns. A model edition identifies the registered estimate; an internal receipt identifies the retained data edition. Missing or mismatched evidence stays visible as unavailable.
| Evidence | Coverage attached to this game | Edition / receipt | Captured / reviewed | How to learn from it |
|---|---|---|---|---|
| Published game scheduleUnavailable | Start time unconfirmed | Unavailable | — | The forecast remains available, but market checks stay withheld until the start is confirmed. Keep the canonical date separate from a confirmed source clock. |
| Forecast editionRegistered edition | basketball-efficiency-v2-ab7f181f3f19 | basketball-efficiency-v2-ab7f181f3f19 | Sep 23, 2026 | Use this ID and capture clock when comparing the forecast with a later result. |
| Team + player archiveRegistered edition | 6 exact player profiles · two team profiles | 2026-09-23T03:50:32.300000Z | Sep 17, 2026 | Historical workload describes the old sample; it does not establish current availability. |
| Shot coordinate archiveVerified source release | 6 exact joins · 1,559 matched attempts · 1,552 located | 411d4ea11dba7d6688d717e26b3a6807640ae7f3b615b054a775e6f814d5c672 | Sep 5, 2026 | Located attempts are a subset of matched attempts; missing coordinates stay missing. |
| Recruiting releaseVerified source release | 0/2 exact programs · 0 people · 0 reviewed sources | 6534ecd70b2b87befb22c2a5781c8b53dd9069e57a668d0409687c9f60d28a06 | Sep 21, 2026 | A dated announcement is evidence of an event, not proof of eligibility or a projected role. |
| Roster observationVerified source release | 36 listed rows · 2027 snapshot | ecf68bc15c3cb083dafa996b55b522bb578de7bf1f7aa5f1d99a726059da4774 | Sep 17, 2026 | Roster status is an observed listing label; it is not a lineup or eligibility ruling. |
This table documents provenance and joins. It does not turn a source listing into an availability decision, and it does not add a second forecast to the stored model.
How the projected score is built.
This table reconstructs the stored ridge model from its published coefficient arrays. Each efficiency term is points per 100 possessions; the final score is efficiency × projected pace ÷ 100. It is an explanation of the estimate, not an additional forecast.
| Term | Penn | Columbia |
|---|---|---|
| League baseline | 107.03 | 107.03 |
| Own offense effect | -0.75 | -0.58 |
| Opponent defense effect | +3.42 | +2.27 |
| Venue effect | -2.46 | +2.46 |
| Projected efficiency | 107.25 | 111.18 |
| Projected pace | 69.46 | 69.46 |
| Score from equation | 74.49 | 77.23 |
A neutral-site game has zero venue effect. A missing or edition-mismatched coefficient set suppresses this table rather than presenting stale arithmetic.
How much has this model missed?
These held-out results describe the forecast system behind this notebook. They are a calibration reference for the range, not a guarantee about this game and not evidence of a market edge.
Training seasons: 2023, 2024, 2025, 2026 · constant home-margin baseline MAE: 12.0 pts. The test season remains outside model fitting and calibration.
Turn the forecast into questions.
Read these in order before opening the deeper tables. Each row is arithmetic over the stored forecast, exact team ratings, matchup factors or matched roster scenario. It does not add a new model.
| Read | Finding | Stored evidence | Question to take to film |
|---|---|---|---|
| Step 1Start with the range | Both teams remain inside the 80% margin range | Columbia is the baseline side by 2.7 points; the stored home-margin range is -13.2 to +18.7. | Which early matchup signal would make you update away from the baseline? |
| Step 2Find the pressure point | Penn owns the largest stored factor edge | turnover rate · 2.0 percentage points toward Penn. | Which actions and lineups are creating that turnover rate edge, and does it hold against the expected rotation? |
| Step 3Check the possession count | 69.5 projected possessions | +0.6 versus the two teams’ 68.9-possession historical tempo midpoint. | Which team benefits if the game reaches that faster projected count, and how will it create the extra trips? |
| Step 4Test roster sensitivity | 0.2-point movement toward Penn | Baseline +2.7 home · roster lens +2.6 home. | Which source-listed players account for that movement, and is there dated evidence for their current availability and role? |
Missing rows mean the exact supporting record did not pass the notebook’s identity or numeric checks; absence is not scored as a neutral signal.
The numbers behind the estimate.
Latest completed-season team rates carried into this forecast edition. They describe the inputs and context; they are not a claim about a current lineup.
| Metric | Penn | Columbia |
|---|---|---|
| Adj O | 106.3 | 106.5 |
| Adj D | 109.3 | 110.5 |
| NET | -3.0 | -4.0 |
| PACE | 68.7 | 69.1 |
| SOS | 1.3 | -2.0 |
| eFG% | 50.8% | 54.1% |
| TO% | 15.4% | 19.1% |
| ORB% | 30.9% | 37.5% |
| FTR | 35.3% | 28.2% |
What do the last five show?
Season and last-five observations come from the same retained team profiles. They are descriptive samples for film review; they are not a trend claim, a cause, or a replacement for the stored forecast baseline.
| Program / exact ID | Sample | Record | Pace | Offensive efficiency | Defensive efficiency | Effective FG% | Turnover rate | Offensive rebound rate | Free-throw attempt rate |
|---|---|---|---|---|---|---|---|---|---|
| PennID 219 | Season30 games | 18-12 | 69.3 | 108.6 | 106.4 | 50.8% | 15.4% | 30.9% | 35.3% |
| PennID 219 | Last five5 games | 4-1 | 67.3 | 103.5 | 105.0 | 50.3% | 15.0% | 25.1% | 17.1% |
| ColumbiaID 171 | Season28 games | 16-12 | 67.6 | 112.4 | 105.2 | 54.1% | 19.1% | 37.5% | 28.2% |
| ColumbiaID 171 | Last five5 games | 2-3 | 63.9 | 107.3 | 103.9 | 50.7% | 16.3% | 34.4% | 20.9% |
Scouting edition 2026-09-23T03:50:32.300000Z · model basketball-efficiency-v2-ab7f181f3f19 · captured Sep 17, 2026.
Who carried the old possessions?
The three highest-minute contributors in the latest completed season. These rows preserve recorded production for preparation; they are not a projected rotation or an availability decision.
What should the staff test first?
These role signals come from the same completed-season player records. Each question is attached to a recorded workload or rate field; missing denominators remain unavailable.
| Program / player | Recorded role signals | Film question |
|---|---|---|
| TJ PowerPenn · F | Usage 23.7%Team minutes 87.0%A/TO 1.133PA 183.0 | How does the defense force this high-usage role into lower-value decisions? |
| Michael ZanoniPenn · F | Usage 19.1%Team minutes 71.5%A/TO 1.333PA 142.0 | How does the defense force this high-usage role into lower-value decisions? |
| AJ LevinePenn · G | Usage 21.2%Team minutes 60.2%A/TO 2.023PA 69.0 | How does the defense force this high-usage role into lower-value decisions? |
| Kenny NolandColumbia · G | Usage 26.0%Team minutes 82.0%A/TO 1.493PA 192.0 | How does the defense force this high-usage role into lower-value decisions? |
| Blair ThompsonColumbia · F | Usage 16.5%Team minutes 76.7%A/TO 1.393PA 110.0 | How does the defense force this high-usage role into lower-value decisions? |
| Mason RitterColumbia · F | Usage 16.5%Team minutes 60.4%A/TO 0.923PA 14.0 | How does the defense force this high-usage role into lower-value decisions? |
Penn
| Player | MIN | MPG | PPG | APG | TS% |
|---|---|---|---|---|---|
| TJ PowerF | 1,053 | 35.1 | 16.4 | 2.4 | 56.6% |
| Michael ZanoniF | 865 | 28.8 | 11.8 | 0.7 | 56.4% |
| AJ LevineG | 729 | 24.3 | 8.2 | 4.0 | 49.4% |
Columbia
| Player | MIN | MPG | PPG | APG | TS% |
|---|---|---|---|---|---|
| Kenny NolandG | 927 | 33.1 | 17.1 | 3.7 | 58.2% |
| Blair ThompsonF | 867 | 31.0 | 11.4 | 1.1 | 60.7% |
| Mason RitterF | 682 | 24.4 | 7.2 | 1.6 | 57.7% |
Turn the factor edge into a film drill.
These links join the historical workload table to the retained national coordinate archive by exact player and team ID. The questions use recorded attempt denominators to choose film; they do not predict a current rotation or treat an archived shot map as availability.
| Program / player | Recorded shot evidence | Preparation question | Drill-down |
|---|---|---|---|
| TJ PowerPenn · F | 373 matched attempts · 369 with retained coordinates · 30 gamesExact archive name | Where does the located sample cluster, and which coverage or transition rule should the staff test against those looks? | View court map ↗ |
| Michael ZanoniPenn · F | 284 matched attempts · 284 with retained coordinates · 30 gamesExact archive name | Where does the located sample cluster, and which coverage or transition rule should the staff test against those looks? | View court map ↗ |
| AJ LevinePenn · G | 204 matched attempts · 203 with retained coordinates · 28 gamesExact archive name | Where does the located sample cluster, and which coverage or transition rule should the staff test against those looks? | View court map ↗ |
| Kenny NolandColumbia · G | 337 matched attempts · 336 with retained coordinates · 28 gamesExact archive name | 51% of recorded attempts are threes. How will the defense change the closeout without giving up the next action? | View court map ↗ |
| Blair ThompsonColumbia · F | 217 matched attempts · 216 with retained coordinates · 28 gamesExact archive name | Where does the located sample cluster, and which coverage or transition rule should the staff test against those looks? | View court map ↗ |
| Mason RitterColumbia · F | 144 matched attempts · 144 with retained coordinates · 28 gamesExact archive name | Where does the located sample cluster, and which coverage or transition rule should the staff test against those looks? | View court map ↗ |
Located coordinates are a subset of matched attempts; missing locations remain in the attempt denominator and are not imputed.
Who is listed today?
Exact-ID rows from the current roster snapshot, ordered by prior recorded minutes. Status is an observed listing label; it does not establish eligibility, availability, a commitment or a projected rotation.
| Program | Listed | Same program | Different program | New to dataset | Ambiguous / other | Prior stat profiles | Recorded prior minutes |
|---|---|---|---|---|---|---|---|
| PennOpen exact roster observations → | 19 | 17 | 0 | 2 | 0 | 17 / 19 | 6,050 |
| ColumbiaOpen exact roster observations → | 17 | 14 | 0 | 3 | 0 | 14 / 17 | 5,650 |
Same-program, different-program and new-to-dataset are exact retained roster observations. Recorded prior minutes sum only valid attached stat profiles; they do not establish current availability, eligibility or role.
Where is prior experience attached?
Source positions group each current listing as guard, forward, center or unreported. Minutes belong to retained prior-season stat profiles and stay separated by same-program, different-program and other observation status. They are context for film review, not a projected rotation.
| Program | Source role | Listed | Prior profiles | Prior minutes | Same-program minutes | Different-program minutes | Other-status minutes |
|---|---|---|---|---|---|---|---|
| Penn | Guard | 5 | 5 | 1,674 | 1,674 | 0 | 0 |
| Penn | Forward | 11 | 10 | 3,874 | 3,874 | 0 | 0 |
| Penn | Center | 3 | 2 | 502 | 502 | 0 | 0 |
| Penn | Unreported | 0 | 0 | 0 | 0 | 0 | 0 |
| Columbia | Guard | 9 | 7 | 2,646 | 2,646 | 0 | 0 |
| Columbia | Forward | 8 | 7 | 3,004 | 3,004 | 0 | 0 |
| Columbia | Center | 0 | 0 | 0 | 0 | 0 | 0 |
| Columbia | Unreported | 0 | 0 | 0 | 0 | 0 | 0 |
Roster season 2027 · receipt ecf68bc15c3cb083dafa996b55b522bb578de7bf1f7aa5f1d99a726059da4774 · captured Sep 17, 2026.
Penn
| Player | Status | MIN | MPG | PPG | TS% |
|---|---|---|---|---|---|
| TJ PowerF | same program | 1,053 | 35.1 | 16.4 | 56.6% |
| Michael ZanoniF | same program | 865 | 28.8 | 11.8 | 56.4% |
| AJ LevineG | same program | 729 | 24.3 | 8.2 | 49.4% |
| Ethan RobertsF | same program | 701 | 30.5 | 16.9 | 54.0% |
| Augustus GerhartF | same program | 595 | 19.8 | 5.5 | 63.1% |
| Lucas LuethF | same program | 467 | 16.7 | 3.3 | 43.1% |
| Cam ThrowerG | same program | 458 | 17.0 | 6.0 | 52.9% |
| Dalton ScantleburyC | same program | 449 | 15.5 | 5.9 | 54.2% |
Columbia
| Player | Status | MIN | MPG | PPG | TS% |
|---|---|---|---|---|---|
| Kenny NolandG | same program | 927 | 33.1 | 17.1 | 58.2% |
| Blair ThompsonF | same program | 867 | 31.0 | 11.4 | 60.7% |
| Mason RitterF | same program | 682 | 24.4 | 7.2 | 57.7% |
| Miles FranklinG | same program | 622 | 23.9 | 9.5 | 50.2% |
| Gerard O'KeefeG | same program | 583 | 21.6 | 8.5 | 61.5% |
| Connor IgoeF | same program | 553 | 19.8 | 8.3 | 56.0% |
| Ryan SoulisF | same program | 520 | 20.0 | 5.8 | 52.7% |
| Hampton SandersG | same program | 428 | 15.3 | 4.2 | 53.9% |
A one-possession question.
The model has Columbia within 2.7 points of Penn; the nominal range spans -13.2 to 18.7. Start the preview with the late-game possessions the baseline cannot separate.
Questions for the next check
- Which team can create a clean look in the final four minutes?
- Which handler faces the most pressure when the pace slows?
- What dated availability or roster evidence could move this outside one possession?
Why the baseline leans.
| Factor | Home attack | Away defense | Away attack | Home defense | Edge |
|---|---|---|---|---|---|
| Shot quality | 51.9% | 51.8% | 51.1% | 52.5% | Home 1.6 pts |
| Ball security | 18.4% | 15.6% | 16.8% | 15.9% | Away 2.0 pts |
| Second chances | 30.8% | 30.1% | 29.5% | 29.0% | Home 0.3 pts |
| Free-throw pressure | 31.5% | 29.8% | 33.9% | 32.5% | Home 0.3 pts |
These are the stored matchup-factor estimates for this edition. They identify a film starting point; they do not establish a tactical result or a player availability decision.
Check the line beside this forecast.
The notebook refreshes the exact game and current forecast edition before showing any qualifying market comparison. A missing quote stays unavailable; no line or edge is inferred.
Checking the current forecast edition and its qualifying market observations…
Keep the evidence chain open.
- Confirm current availability and the expected rotation from a dated update.
- Use the roster and player archives to identify the personnel behind each factor.
- Read the interval before treating a close projection as a decisive edge.
Follow the record.
The schedule ID, model edition and captured forecast remain attached to this notebook. A missing market observation is unavailable evidence, not a zero.
No verified pregame market line is attached to this edition.