Research Lab
The testable claims behind Total Creation Value — each with a defined test plan and an observable signal that would confirm or kill it. A research program, not a results page.
Correlate IIB scores with TCV percentile rankings across all seasons. Control for position.
r > 0.80 between IIB > 4.5 threshold and top-10% TCV rank.
No formal test logged yet.
Regress team assist rate on PVA vs AST/TO for all players with 500+ possessions.
PVA explains more variance in team AST% than AST/TO ratio.
No formal test logged yet.
Compare DPC and steal rate as predictors of opponent HalfCourt PPP over 30-game windows.
DPC beta coefficient > steal rate beta in multivariate regression.
No formal test logged yet.
Run SGV against corner-3% and plot residuals. High-SGV players should show corner-3 lift for teammates.
SGV predicts teammate corner-3% improvement better than own corner-3% baseline.
No formal test logged yet.
Compare DSV and DRPM as predictors of opponent PPP in playoff matchups vs regular season performance.
DSV-to-playoff-defense correlation > DRPM-to-playoff-defense correlation.
No formal test logged yet.
Identify players with COV > PVA percentile. Measure their teammate scoring uplift on cuts and rolls.
COV > PVA gap predicts teammate basket creation at rim above expectation.
No formal test logged yet.
Measure SAV before and after key initiator trades/departures. Require 40+ game sample on both sides.
SAV drops by >= 15% within 20 games of losing primary initiator.
No formal test logged yet.
Flag players where MIV percentile > (IIB + PVA) composite percentile by 20+. Validate via film.
MIV-high players show documented cut activity and positioning lift in film review.
No formal test logged yet.
Split RPV by game quarter. Compare Q4 rim assignment frequency vs Q1-Q3 baseline.
Q4 RPV exceeds Q1-Q3 average for players with positive game-leverage correlation.
No formal test logged yet.
Calculate component variance for all players with <200 playoff possessions. Rank by component.
PTV coefficient of variation > all other components in <200-possession sample.
No formal test logged yet.
Regress UP on IIB rolling standard deviation (20-game window) vs total games played.
IIB variance explains more UP variance than games-played alone.
No formal test logged yet.
Calculate play-type concentration for all players. Correlate with CFP scores.
Players where one action type accounts for >50% of scoring have CFP in top quartile.
No formal test logged yet.
Segment players by primary action type. Compare IIB x PVA interaction term by segment.
PnR-initiator segment shows higher IIB+PVA correlation than catch-and-shoot segment.
No formal test logged yet.
Bin players by position and DSV threshold. Compare live-ball turnover generation rates.
Guard DSV > 3.0 group outperforms big DSV > 3.0 group in live-ball turnovers per 100.
No formal test logged yet.
Tag transition possessions by initiator vs finisher role. Calculate COV by role.
Initiator-role COV median exceeds finisher-role COV median by >0.5 points.
No formal test logged yet.
Plot SGV against usage rate. Fit piecewise regression with breakpoint at 28% usage.
Negative SGV slope in usage > 28% segment; positive or flat below threshold.
No formal test logged yet.
Use PVA and AST/TO as predictors of team wins in playoff series. Sample: 2015-2024.
PVA model RMSE < AST/TO model RMSE in predicting playoff win totals.
No formal test logged yet.
Calculate year-over-year correlation for each component. Require 600+ possessions in both years.
IIB YoY r > all other components for qualifying players.
No formal test logged yet.
Partial-out each component's independent contribution to opponent HalfCourt PPP suppression.
DPC partial r-squared exceeds all other components in the multivariate model.
No formal test logged yet.
Identify players with high MIV and low IIB+PVA. Compare to team performance with/without them.
High-MIV, low-IIB+PVA players show positive team performance impact not explained by composite.
No formal test logged yet.
Identify switching bigs with low block/steal rates. Compare DSV to defensive matchup outcome data.
Switching big DSV residuals are systematically negative vs matchup-outcome ground truth.
No formal test logged yet.
Filter players with PVA below median. Measure COV distribution and team-level assist lift.
High-COV, low-PVA players show >=10% teammate assist rate lift above expectation.
No formal test logged yet.
Track defensive performance change for traded players. Compare RPV and PER as predictors.
RPV pre-trade explains post-trade defensive performance better than PER pre-trade.
No formal test logged yet.
Segment players by team playoff seed. Measure PTV vs regular-season TCV correlation by segment.
Playoff-seeded team context shows higher PTV vs TCV correlation than non-playoff context.
No formal test logged yet.
Identify SAV leaders. Track teammate IIB change when they are on vs off court.
Top-SAV quartile shows teammate IIB +0.4 on/off split with flat own IIB.
No formal test logged yet.
Compare component volatility metrics (rolling SD, range) across experience levels.
Rookie/sophomore player UP rolling SD > all other components in the same cohort.
No formal test logged yet.
Cluster players by action type and shot creation origin. Compare within-cluster SGV variance.
Gravity creator and floor spacer clusters overlap significantly on raw SGV score.
No formal test logged yet.
Calculate catch-and-shoot share for all players. Correlate with CFP scores.
Catch-and-shoot share > 65% predicts CFP top quartile with 70%+ accuracy.
No formal test logged yet.
Regress IIB on team offensive rating. Test for non-trivial team-quality beta coefficient.
Team offensive rating explains >=10% of IIB variance after controlling for individual volume.
No formal test logged yet.
Build regression models using (a) PVA, (b) SAV, (c) PVA+SAV, (d) AST/TO. Compare R-squared on team AST%.
PVA+SAV model R-squared > any single-predictor model including traditional AST/TO.
No formal test logged yet.
Untested is not the same as false. These hypotheses define what TCV is built to prove — each carries a specific test plan and an observable signal before its status can change. Nothing is marked confirmed without a verifiable result. This is a living research program: statuses update as formal tests are run, not as the model ships.