Decision standard
2025-26 our play-by-play · 15 teamsStatistical validation and game-day decisions are different questions
A coach cannot wait for statistical significance. The useful question is rarely "can we prove this tendency" — it is "given the evidence available right now, what should we believe, how uncertain should we be, what might they do, and what is our best response given the consequences of being wrong." Coach Ready answers that without lowering the bar for anything it calls a validated metric.
Two evidence standards
Discovery standard
Can we prove this is a real, repeatable phenomenon?
Proprietary metrics
Claims about stable player traits
General basketball findings
Product claims
Split-half reliability, sample floors, replication across a cohort, and rejection when the test fails.
Metric Lab
Nothing in the decision standard lowers this bar. Every rejected metric stays rejected.
Decision standard
Is the evidence available right now useful for making this decision?
Opponent preparation
Game planning
Matchup decisions
Scouting
Rotation considerations
Tactical preparation
Evidence strength, the cost of the adjustment, the consequence of being wrong, the opponent's available counters, and whether a response holds up across several plausible opponent behaviors.
Game Plan Intelligence
A decision-useful signal is never restated as a proven tendency or a player trait.
How a recommendation is now reasoned
The old chain — metric, threshold, program rule, recommendation — is still inside this one. What is new is that the recommendation now depends on what happens if we are wrong, and on whether a response holds up across more than one thing they might do.
Where a signal can go
We believe this strongly enough, and the response is affordable enough, to attack it.
The evidence does not settle it, so we take the response that limits our downside if the read is wrong.
Not strong enough to move our base plan, but worth having a counter rehearsed and ready.
Not enough evidence to act before tip. Here is the early-game signal that would decide it.
This is what gives a weak signal somewhere honest to land. It is not surfaced as a fact and it is not thrown away.
Evidence categories
Enough volume, repeated across the season, that we are comfortable planning against it.
Real and consistent, but not enough to treat as settled.
The sample is small. It points somewhere. It is a working hypothesis, not a tendency we have established.
A human saw it. Valuable and unquantified — we hold it as an observation, not a measurement.
Two credible sources of evidence disagree. The disagreement itself is scouting intelligence.
We do not have the evidence. Saying so is the honest output.
These sit alongside sample sizes, never in place of them. Every assessment carries the real sample it came from.
Asymmetric risk
Cheap, reversible, consistent with what we already teach — a directional signal is enough. A weak read can justify action when being wrong costs us almost nothing.
A response that creates a real vulnerability, asks us to abandon our identity, or costs significant preparation needs a strong repeated signal — even a good read may not be worth it if their counter is dangerous.
What it costs us to be wrong · set by our staff
program-definedA switch a veteran group runs without thinking is a week of practice for a young one, so the price of an adjustment is our call, not the data's. Change one here and the confidence bar moves with it: the same evidence can land on exploit instead of protect, or the other way round. It changes nothing about what was measured.
"2" versus high pick-and-roll
A situational automatic that has to be recognized live.
A screen is no excuse
Principles we already hold. No new preparation cost.
Blitz
A blitz puts us in a 4-on-3 behind the pressure and requires every back-line rotation to be right. If the read is wrong we have handed them an advantage possession.
Drop
Drop is our lowest-risk coverage and it is close to our base. Being wrong costs us pull-up looks, not broken rotations.
Pistol
Pistol is already one of our base entries. Emphasizing it is not an adjustment.
Show hi-lo
Show asks the screener's defender to commit and recover. Cost is real but recoverable.
Soft
Our own document treats Soft like a blitz on the back line, so it is not free — but the front is conservative.
Sprint pick-and-roll
Changing how we start possessions costs practice time and moves us off our normal entries.
Switch same size
Switching creates a mismatch we then have to scram out of. Wrong read means a post-up or a slip, and it asks the whole group to rehearse the exception.
Versus roll and slide
A rotation rule inside a coverage we already run.
Versus spread pick-and-roll
A structural adjustment to how we cover spacing.
Weak
Weak is an angle rule we already teach. Applying it against the wrong profile costs almost nothing.
Rejected metrics, re-examined
Nothing here is restored as a validated metric. The audit asks only whether the underlying observation has a legitimate game-day use, and records the sentence Coach Ready is still forbidden from saying.
Personal Effective Minutes threshold (fatigue limit)
A player has a stable personal minutes threshold past which his efficiency declines within a stint.
For a specific player, involvement declined after a specific rotation pattern in a handful of recent games. That is a pattern in this season's rotations, not a fatigue limit.
"He fades after seven minutes." The cohort test found the opposite of a fatigue signature — involvement rose modestly as stints ran longer.
Legitimate as a rotation thing to watch in the first half. Illegitimate as a player attribute, a scouting bullet, or a Fit Engine input.
Performance Stability (game-to-game consistency)
This player is reliably himself game to game — consistency is a measurable trait.
Nothing. Once residualized on production, minutes and usage, the reliability disappears entirely — there is no signal left to be directional about.
Any framing that presents variance as temperament or as a projectable trait.
This is the important negative case: the decision standard does not rescue a metric that measures noise. Useless under both standards.
Opponent-Quality Translation
This specific player's production translates upward against better defenses.
The cohort-wide effect is real even though the player-specific part is noise: everyone scores more efficiently against weaker defenses. Used as schedule-strength context on an opponent's season numbers, it changes how much we discount them.
A per-player "translates up" or "translates down" badge. Split-half reliability for the player-specific slope is essentially zero.
Survives only as opponent context on other numbers, which is exactly where the Metric Lab already put it.
Usage Elasticity / Role Elasticity
His efficiency holds up when his usage rises — measurable within one season.
For game planning only: if we take their primary creator's touches away, someone else has to take them. Which player absorbs usage is observable from lineup data even though how well he handles it is not.
Any claim about how efficiently the second creator will handle the extra load. The within-season slope did not clear our reliability bar.
Turns a rejected player rating into a legitimate question about who our pressure hands the ball to.
First-half versus second-half behavior
Half splits identify a program-specific or player-specific second-half attribute.
Fouling rises clearly in second halves across the whole cohort. That is a general basketball fact, useful for foul-risk and rotation planning, and it is not about any individual.
Attributing the cohort-wide pattern to a specific player or program as a tendency.
General, not proprietary. Fine as planning context, never as a scouting finding about them.