FADE IN: A dimly lit sports bar. Multiple screens showing pre-game coverage. SAL (60s, reading glasses, three phones, old-school bookie who survived the digital age) sits in a corner booth. DANNY (20s, sharp but green, thinks he's figured out sports betting) slides in across from him. DANNY Sal, I've got a lock. Titans are playing the Wolves tonight. Titans' star point guard just sprained his ankle in warmups. The line hasn't moved yet. I'm hammering the Wolves moneyline. SAL (not looking up from his phones) You think you're the only one who saw that tweet? DANNY The line hasn't moved yet! SAL The line will move. But that's not your problem. Your problem is you're thinking in straight lines. One event, one outcome. The real money is in the web.
SAL When Marcus Reed sprains his ankle, you see one thing: Titans are weaker, Wolves win. Simple probability shift. Maybe Titans' win probability drops from 60% to 40%. DANNY Right. So I bet the Wolves. SAL And so does everyone else. The line adjusts. Your edge disappears in minutes. But here's what the amateurs miss: Reed's ankle doesn't just change the game outcome. It changes EVERYTHING connected to that game. And those connections? That's where the real value hides. He pulls out a small notebook filled with grids. SAL (CONT'D) This is Cross Impact Analysis. Instead of looking at one event in isolation, you map how every event affects the probability of every OTHER event. It's a matrix of conditional probabilities.
Sal draws a grid on a napkin. Rows and columns labeled with events. SAL Here are the events connected to tonight's game. Event A: Reed is out. Event B: Titans win. Event C: Total points over 210.5. Event D: Titans' backup guard scores over 15 points. Event E: Wolves' center gets over 12 rebounds. He fills in the matrix: SAL (CONT'D) The matrix shows: if Event A happens (Reed out), what's the new probability of each other event? SAL (V.O.) A→B: Titans win drops from 60% to 40% A→C: Over 210.5 drops from 55% to 42% (less scoring without Reed) A→D: Backup over 15 pts rises from 20% to 65% (more minutes, more shots) A→E: Wolves center over 12 rebounds rises from 35% to 55% (weaker interior defense) SAL (CONT'D) One event. Four probability shifts. And the market is only pricing in the first one.
DANNY Wait. So the over/under changes too? SAL Of course it does. Reed averages 24 points a game. His backup averages 11. That's a 13-point expected scoring drop from one position. Even if the backup plays more minutes, the efficiency drops. The total is likely to go UNDER. DANNY But the total line hasn't moved either. SAL Because the market is slow on cross impacts. The moneyline adjusts fast — that's the obvious bet. The total adjusts slower. And the prop bets? The player props? Those are the last to move. That's your window. He circles Event D on the matrix. SAL (CONT'D) The backup guard's point total. The market still has his over/under at 12.5 based on his season average. But with Reed out, he's getting 35 minutes instead of 15. His conditional probability of hitting 15+ just tripled. THAT'S the bet.
SAL But we're not done. Cross Impact Analysis doesn't stop at first-order effects. Event A affects Event D. But Event D affects OTHER events too. DANNY Like what? SAL If the backup guard is taking more shots (Event D), he's taking shots AWAY from someone else. The Titans' power forward, who normally gets 18 shots, might only get 12. His point total prop is now overvalued. He adds to the matrix: SAL (CONT'D) D→F: If backup takes more shots, power forward's points drop. His over/under at 22.5 is now probably a strong under. And it cascades further. If the power forward is getting fewer touches, he's less engaged defensively. Which means the Wolves' small forward has easier driving lanes. His assist numbers go up. DANNY (eyes widening) It's like dominoes. SAL It's like a WEB. Every node connects to every other node. Pull one thread and the whole structure shifts. The amateurs see the thread. The professionals see the web.
Sal flips to a fresh page and draws a larger matrix — 6 events by 6 events. SAL A proper Cross Impact Matrix is N-by-N. Every event gets a row and a column. The cell at row i, column j tells you: "If event i occurs, what is the new probability of event j?" He fills in numbers: SAL (V.O.) B(Win) C(Over) D(Backup) E(Reb) F(PF pts) A(Reed) 0.40 0.42 0.65 0.55 0.35 B(Win) — 0.60 0.45 0.30 0.55 C(Over) 0.55 — 0.50 0.45 0.50 D(Backup) 0.38 0.48 — 0.50 0.30 E(Reb) 0.42 0.52 0.50 — 0.48 SAL (CONT'D) The diagonal is empty — an event doesn't impact itself. But every other cell is a conditional probability. This matrix IS the game. Everything else is just commentary.
SAL Now here's how you make money. You compare the cross-impact adjusted probabilities to the market's implied probabilities. He writes two columns: SAL (V.O.) Market implied probability vs. Cross-impact adjusted: Wolves moneyline: Market 55% → Adjusted 60% → Small edge Under 210.5: Market 48% → Adjusted 58% → Moderate edge Backup over 12.5 pts: Market 45% → Adjusted 65% → LARGE edge Wolves C over 12 reb: Market 38% → Adjusted 55% → Large edge PF under 22.5 pts: Market 50% → Adjusted 65% → Moderate edge SAL (CONT'D) The moneyline? Tiny edge. Everyone's already on it. The under? Decent. But the backup's points and the center's rebounds? Those are 15-20 point probability gaps. The market hasn't priced in the cross impacts yet. DANNY So I bet those instead of the moneyline? SAL You bet where the EDGE is largest. The moneyline is the obvious play with the smallest edge. The prop bets are the hidden plays with the biggest edges. Cross Impact Analysis shows you exactly where the market is wrong.
DANNY Can I parlay all of them? Backup points AND center rebounds AND the under? SAL Careful. That's where correlation kills you. These events aren't independent — they're connected through the same causal web. A parlay assumes independence. If the events are positively correlated, the parlay overestimates your edge. DANNY What do you mean? SAL If the backup scores a lot, it probably means the game is close, which means more possessions, which means the total might go OVER, not under. Your "under" bet and your "backup over" bet are partially contradictory. He draws arrows between the events. SAL (CONT'D) Cross Impact Analysis shows you the correlations explicitly. Cell D→C tells you: if the backup scores big, the over probability shifts to 48%. That's HIGHER than the baseline. So combining "backup over" with "game under" in a parlay is fighting against the correlation structure. DANNY So I pick the bets that are POSITIVELY correlated for parlays? SAL Or you bet them individually and size each one by the edge. Don't let the parlay payout seduce you into ignoring the math.
The game tips off on the screen above them. Sal watches intently. SAL Now watch. The matrix isn't static. As the game unfolds, new information updates the conditional probabilities in real time. First quarter ends. The backup guard has 8 points. SAL (CONT'D) He's on pace for 32. The market is adjusting his live line, but it's still lagging. More importantly, his hot shooting is pulling the defense toward him, which means... DANNY The center is getting easier rebounds? SAL (pointing at him) The cross impact is playing out in real time. Center already has 5 rebounds in the first quarter. His live over/under just moved from 12.5 to 14.5, but our model had him at 55% for over 12 — he's tracking even higher. He checks his phone. SAL (CONT'D) And the total? Currently on pace for 198. Under is looking strong. The market had it at 48% under. We had it at 58%. The first quarter data is confirming our cross-impact model. DANNY This is insane. SAL This is math. The insane part is that most people bet without it.
Final buzzer. Wolves win. Total: 201 (under). Backup guard: 22 points. Center: 14 rebounds. Power forward: 16 points (under his line). Sal checks his notebook. Every cross-impact prediction landed. DANNY (stunned) You called all of it. SAL I didn't call anything. The matrix called it. I just read the matrix. He closes the notebook. SAL (CONT'D) ISO 31010, Section B.6.2. Cross Impact Analysis. Intelligence agencies use it to predict geopolitical events. Epidemiologists use it to model disease spread. I use it to find value in a basketball game. DANNY Can you teach me to build the matrix? SAL (standing up) I just did. One event changes everything it touches. Everything it touches changes everything THEY touch. Map the web, quantify the connections, and you'll see what the market can't. He drops a twenty on the table for the drinks. SAL (CONT'D) The straight-line thinkers bet the moneyline and break even. The web thinkers bet the cross impacts and build an edge. Choose your geometry. Sal walks out. Danny stares at the napkin matrix, then pulls out his own notebook and starts building his first grid. FADE OUT. — END —