A sim that only spits out a number is a black box with a scoreboard. This one plays every possession then shows the exact identity clashes that produced the total, which scheme beats which, which archetype the coverage frees, which matchup decides the night, one game or a whole season. You watch it play, then you watch it explain.

Case 01 · the trace, not the score

The score is the answer. The trace is the proof.

The engine plays a matchup possession by possession and hands back not just the result but the interaction trace: the scheme-versus-scheme and archetype-versus-coverage clashes that produced it. Here is Duke's identity against a switch-heavy composite opponent.

Duke Blue DevilsTeam KR 90.6

71%

win prob

avg 79 · 73

Switch-Heavy Opponentcomposite · ~85

29%

Interaction trace · offense

Point-forward hub vs switch defenseswitching puts a smaller defender on the hub; he punishes the mismatch in the post and from the elbow, the possession the whole game bends aroundDuke +4.1decides

Movement shooter vs top-lock coveragethe opponent chases him off the line, cutting his looks but bending the help the hub then exploitseven

Interior big vs drop-and-dumpan efficient rim finisher against a defense that sinks; clean looks at the rim in the half courtDuke +1.8

Freshman guards vs perimeter pressureball pressure targets the young creators, the one place the Duke floor dipsOpp +2.3

Interaction trace · defense

Positional size vs opponent wingsDuke switches one through four without giving up size, smothering the opponent's first actionDuke +2.2

Drop-oriented big vs pick-and-popa stretch five pulls the anchor out; a live seam the opponent attacks all nightOpp +1.6

Each clash is a real entry in the interaction library, matched to this pairing. The totals are not assumed; they are the sum of these matchups, possession by possession. That is the chain that produced the score.

Projected box score · this matchup

Duke · key linesPTSREBASTeFG

Cameron Boozerhub2611558%

Isaiah Evansshooter123149%

Patrick Ngongba IIanchor117162%

Duke total79341554%

Opponent total73311351%

~67 possessions · every line projected from the player read, then shifted by the interaction library for this pairing

What drives Duke's edge

The hub against switches: every switch is a mismatch he scores over.

Positional size erasing the opponent's first action on defense.

What compresses the margin

If the freshman guards foul early, the pressure edge tilts to the opponent.

If the stretch five gets the anchor moving, the pick-and-pop seam opens the game up.

The engine does not just say Duke wins 71%. It shows the whole game: the clashes on both ends, the box score they add up to, and the two or three things that would flip it. That is a scouting report falling out of a score, not a number handed down from a black box.

Real Duke engine read vs an illustrative composite opponent, run through the real possession-engine, interaction-trace, and projected-box-score structure. Demonstration projection.

Case 02 · a distribution, not a prediction

It does not predict one score. It plays the game thousands of times.

A single projected score pretends to a certainty no one has. The engine runs the matchup thousands of times and hands back the whole distribution: the likely margin, the spread around it, and a confidence set by the least-informed team in the game.

Opponent by 15evenDuke by 15

Most likely Duke by 6, but the middle 80% of outcomes runs from a 4-point loss to an 18-point win. The result is a range, and the engine trades in the range, not the point.

Confidence 66% · a simulation is only as confident as its least-informed team; the composite opponent takes it down

A 71% favorite is not a promise. It is a bet, and the engine sizes it honestly: a coin-flip swing game and a near-lock both get played thousands of times, and the width of the distribution tells you which one you are looking at. Believable over impressive, applied to a scoreboard.

Illustrative distribution on the real Monte-Carlo simulation structure. Demonstration figures.

Case 03 · one game, or the whole season

The same engine runs the schedule, and the bracket.

Scale the possession engine up and every game on the schedule gets played thousands of times. The record becomes a band, the season's coin-flips get named, and a full bracket resolves into advancement odds. Same Duke read.

28-4 to 31-1

projected record band · middle 80% of season runs

at Rival A52% · coin-flip

vs Rival B57% · swing

at Rival C48% · coin-flip

Bracket: to Final Four34% odds

The record band is the range across thousands of seasons; the swing games are the three coin-flips that actually decide where the band lands. This is how a season stops being a hunch and becomes a set of odds you can prepare against, game by game, with the trace behind every one.

Real Duke engine read run through the real season and bracket simulation structure. Demonstration projection; opponents shown generically.

The law underneath

A final score is an answer. The trace is the proof.

Any model can hand you a number and ask you to trust it. This one refuses the black box. It plays every possession, shows the exact scheme and archetype clashes that produced the total, and hands the result back as a distribution with its confidence attached, not a single score pretending to certainty. One game or a whole season, you watch it play, then you watch it explain. The number is where it ends. The trace is why you believe it.

You watch it play. Then you watch it explain.

Simulation produces the projection and the trace; turning that into a scout, a game plan, and live adjustments is the next surface.

Intelligence

Team Intelligence

The team read that feeds every simulation.

See the team →

Intelligence

Game Ops

Turn the projection into a scout, a game plan, and live halftime adjustments.

See game ops →

Intelligence

Roster Construction

Change the roster, then simulate the team you could build.

See roster →

Get access