Correlation is the one concept that separates SGP bettors who find value from SGP bettors who donate money.
If you understand which legs are genuinely correlated and which ones just feel connected, you have an edge most bettors — recreational and sharp — don't fully exploit. This guide covers how correlation works in same-game parlays, how sportsbooks price it (and where they get it wrong), and how to build a repeatable framework for finding +EV SGP opportunities across NFL, NBA, MLB, and NHL.
1. [What Is Correlation in Sports Betting?](#what-is-correlation-in-sports-betting)
2. [Why Correlation Is the Key to SGP Value](#why-correlation-is-the-key-to-sgp-value)
3. [Positive vs. Negative Correlation](#positive-vs-negative-correlation)
4. [The Correlation Spectrum](#the-correlation-spectrum)
5. [Sport-by-Sport Correlation Maps](#sport-by-sport-correlation-maps)
6. [How Sportsbooks Price Correlation (And Where They Get It Wrong)](#how-sportsbooks-price-correlation)
7. [Building Your Own Correlation Framework](#building-your-own-correlation-framework)
8. [Real-World SGP Correlation Examples](#real-world-sgp-correlation-examples)
9. [Common Correlation Mistakes](#common-correlation-mistakes)
10. [FAQ](#faq)
If one event happening makes another event more or less likely, those two events are correlated. That's it. Most people overcomplicate this.
Think about it outside of sports first. If it's raining, you're more likely to see people carrying umbrellas. Rain and umbrellas are positively correlated — one makes the other more likely. If it's 95 degrees and sunny, you're less likely to see people in winter coats. Heat and heavy coats are negatively correlated — one makes the other less likely.
Now apply that to a football game. If Patrick Mahomes throws for 320 yards, what else is probably true about that game? The Chiefs are probably winning. The game is probably high-scoring. Travis Kelce probably had a decent receiving day. Those outcomes are all positively correlated with Mahomes having a big passing game — they tend to happen together, not independently.
That's correlation. It's the relationship between outcomes. And in same-game parlays, it's everything.
When you build a traditional multi-game parlay, each leg is from a different game. The Celtics covering has nothing to do with the Cowboys winning. Those outcomes are independent — the result of one doesn't influence the other.
But in a same-game parlay, every leg comes from the same game. And in a single game, outcomes are deeply interconnected. The quarterback's performance affects the spread. The game script affects rushing volume. The pace affects the total. Nothing happens in isolation.
This interconnectedness — correlation — is what makes SGPs fundamentally different from traditional parlays. And it's what creates the opportunity for bettors who understand it.
Most SGP content dances around this: sportsbooks know about correlation, they adjust for it, and they still don't get it perfectly right.
In a standard multi-game parlay, the math is straightforward. Each leg is independent, so you multiply the implied probabilities together. A 3-leg parlay with three -110 legs:
In an SGP, the legs aren't independent. If you parlay "Shai Gilgeous-Alexander over 29.5 points" with "Thunder -5.5," those outcomes are connected. When SGA goes off for 35+, the Thunder are far more likely to be winning by 6 or more.
If the book priced these as independent — just multiplying the probabilities — the parlay would pay too much, because the combined probability of both hitting is higher than the product of the individual probabilities suggests. The book would be giving away value.
So sportsbooks apply a correlation adjustment — sometimes called the "SGP tax" or "correlation tax." They reduce the payout to account for the fact that the legs move together.
The correlation tax is a blunt instrument. Sportsbooks use mathematical models (more on this in Section 6) to estimate how correlated each combination of legs is, but these models are approximations built on historical data — and they have blind spots.
The value in SGP betting lives in the gap between:
1. How correlated the legs actually are (based on the specific game context, matchup, and conditions)
2. How much the sportsbook charges for that correlation (the adjustment baked into the SGP price)
When the book underestimates the correlation — charges too little tax — you're getting a better price than you should. When the book overestimates it — charges too much tax — you're getting ripped off.
Your job as an SGP bettor is to find the spots where the book's correlation model is wrong. That's the entire game.
Not all correlations point in the same direction. Understanding whether legs are positively or negatively correlated is the foundation of every SGP you build.
Positively correlated legs tend to hit at the same time. When one wins, the other becomes more likely to win too.
Example 1: Patrick Mahomes 280+ Passing Yards + Chiefs -3.5
This is one of the most intuitive positive correlations in sports betting. When Mahomes is throwing for 280+ yards, the Chiefs offense is humming. They're sustaining drives, scoring points, and controlling the game. In games where Mahomes has thrown for 280+, the Chiefs have covered a 3.5-point spread at a significantly higher rate than their baseline.
The mechanism is clear: elite QB performance drives team success. The passing yards aren't just a stat — they're a proxy for offensive dominance.
Example 2: Shai Gilgeous-Alexander 30+ Points + Thunder -5.5
SGA is the Thunder's engine. When he scores 30+, it usually means he's in rhythm, the Thunder offense is clicking, and the game isn't close enough for him to sit in the fourth quarter (or he's the reason it got out of hand in the first place). Over the last two seasons, when SGA has scored 30+, the Thunder have covered a 5.5-point spread roughly 65% of the time — well above the ~50% baseline.
Example 3: Josh Allen 2+ Passing TDs + Bills Over 24.5 Team Total
Passing touchdowns are literally points. If Allen throws two or more TDs, that's at least 14 points from those scores alone (assuming extra points). Add in any rushing or field goal production, and 24.5 becomes much easier to clear. The correlation here is almost mechanical — TDs are a direct input to the team total.
Negatively correlated legs work against each other. When one hits, the other becomes less likely. These are the traps that catch most recreational SGP bettors.
Example 1: Both QBs Over Passing Yards + Game Under
This is the classic negative correlation trap. If both quarterbacks are throwing for big yardage, the game is almost certainly high-scoring. Passing yards and points are tightly linked — you can't have two QBs combining for 550+ passing yards in a 17-13 game. It happens, but it's rare enough that parlaying these legs is fighting basic math.
Yet you see this in recreational SGP builds constantly. People think "I like both QBs" and "I like the under" without realizing those views are contradictory.
Example 2: Player Rebounds Over + Blowout Spread
If you're taking a player's rebounds over in an NBA game and also taking a team to win by 15+, you're creating a tension. In blowouts, starters sit in the fourth quarter — sometimes earlier. Your rebounds guy might be on pace for 12 boards through three quarters, then sit the entire fourth and finish with 9. The blowout you needed for the spread is the same thing that kills the rebounds prop.
This is especially dangerous with role players whose minutes are most sensitive to game flow.
Example 3: Running Back Rushing Yards Over + Team Losing
If a team is down two scores in the second half, they're throwing the ball. The running back's rushing volume dries up. Taking a RB's rushing yards over alongside the opposing team's spread is a negative correlation — the game script that makes one leg hit actively undermines the other.
Some correlations flip depending on context:
Correlation isn't binary — it's a spectrum. Not all positive correlations are equally strong, and understanding the degree of correlation is what separates good SGP builders from great ones.
Statisticians measure correlation on a scale from -1 to +1 (the Pearson correlation coefficient, or r). You don't need to calculate these yourself — but knowing the ranges helps you think about which leg combinations are tight, loose, or working against you.
These legs are tightly linked. When one hits, the other hits most of the time.
| Leg Combination | Typical r | Why It's Strong |
|---|---|---|
| QB passing TDs + team win | 0.65–0.75 | TDs are points; points win games |
| Star player 30+ pts + team covering large spread | 0.60–0.70 | Star dominance = team dominance |
| Game total over + combined passing yards over | 0.70–0.80 | Passing yards are the primary scoring mechanism |
| Pitcher 7+ Ks + team win (ace pitchers) | 0.55–0.65 | Dominant pitching = opponent can't score |
Translation: these legs want to hit together. When one lands, the other usually follows.
SGP implication: Sportsbooks usually price these correctly because the correlation is obvious. The value here is smaller, but it exists on the margins — especially with alternate lines.
These legs are connected but not locked together. There's a meaningful relationship, but plenty of games where one hits and the other doesn't.
| Leg Combination | Typical r | Why It's Moderate |
|---|---|---|
| Star scoring + team spread (moderate spread) | 0.35–0.50 | Star can score 30 and team still doesn't cover |
| WR receiving yards + team win | 0.30–0.45 | One WR's production doesn't determine the game |
| NBA assists leader + game over | 0.30–0.40 | High assists suggest pace, but not always scoring |
| RB rushing yards + team win (run-first team) | 0.35–0.50 | Correlation is game-script dependent |
Translation: these legs are friends, not soulmates. They trend together, but one can easily hit without the other.
SGP implication: This is where the most value lives. The correlation is real but hard to quantify precisely, and sportsbook models often under- or over-adjust. These are your bread-and-butter SGP legs.
These legs are essentially independent — one outcome tells you almost nothing about the other.
| Leg Combination | Typical r | Why It's Weak |
|---|---|---|
| Player assists + game total | ~0.05–0.15 | Assists don't drive scoring volume in a meaningful way |
| Defensive player tackles + spread | ~0.05–0.10 | Individual defensive stats are noisy |
| First basket scorer + game total | ~0.00 | Who scores first has no bearing on final total |
| Player steals + team win | ~0.05–0.15 | Steals are too random to predict game outcomes |
Translation: these legs don't know each other. One hitting tells you nothing about the other.
SGP implication: These legs are fine to include in SGPs because they add odds without creating correlation issues. But don't fool yourself into thinking they're correlated just because they're in the same game.
These legs actively work against each other.
| Leg Combination | Typical r | Why It's Negative |
|---|---|---|
| Bench player props + blowout spread | -0.30 to -0.45 | Blowouts mean starters sit, but bench players play garbage time (mixed) |
| RB rushing over + team losing by 10+ | -0.35 to -0.50 | Losing teams abandon the run |
| Goalie saves over + team winning big | -0.30 to -0.45 | Dominant teams face fewer shots |
Translation: these legs are pulling in opposite directions. One hitting makes the other harder to land.
These are contradictory outcomes. Parlaying them is almost always a mistake.
| Leg Combination | Typical r | Why It's Contradictory |
|---|---|---|
| Both QBs over passing yards + under | -0.60 to -0.75 | High passing = high scoring |
| Pitcher over strikeouts + over runs (same team) | -0.50 to -0.65 | Dominant pitching suppresses runs |
| Player scoring under + team covering large spread | -0.55 to -0.70 | Team blowouts require someone scoring |
Translation: these legs are contradictions. Parlaying them is arguing with yourself.
SGP implication: Avoid these. Period. If you find yourself building an SGP with strongly negatively correlated legs, stop and rethink your game thesis.
Correlation patterns differ by sport because the underlying game mechanics differ. Here's a practical map for the four major North American sports.
Football has the strongest and most exploitable correlations because game script is so deterministic. Once a team falls behind, the entire offensive approach changes — and that change cascades through every prop.
High-Value Correlations:
Basketball correlations are driven by pace, star usage, and blowout risk. The biggest factor most bettors ignore: minute distribution in blowouts.
High-Value Correlations:
Baseball correlations are pitcher-centric. The starting pitcher's performance is the single biggest driver of game outcomes, and most SGP value flows from there.
High-Value Correlations:
Hockey correlations are goalie-driven and power-play dependent. The sport's inherent randomness (puck luck, deflections) makes correlations weaker overall, but they still exist.
High-Value Correlations:
You can't spot mispricing if you don't understand how the pricing works. So let's look under the hood — briefly.
Sportsbooks don't just multiply individual leg probabilities together. They use copula models — statistical tools that model how multiple variables depend on each other. The name sounds academic, but the process is intuitive once you see the steps:
1. Individual leg pricing: The book prices each leg independently first (Mahomes over 279.5 passing yards at -115, Chiefs -3.5 at -110, etc.)
2. Correlation matrix: The book's model estimates the pairwise correlation between every possible combination of legs. Mahomes passing yards and Chiefs spread might have a correlation of 0.55. Mahomes passing yards and the game total might have a correlation of 0.40.
3. Joint probability calculation: Using the copula model and the correlation matrix, the book calculates the joint probability of all selected legs hitting simultaneously. This is always different from simply multiplying individual probabilities.
4. Margin application: The book adds its margin (vig) on top of the joint probability to arrive at the final SGP price.
The result: an SGP with positively correlated legs pays less than a traditional parlay with the same individual leg odds, because the true probability of all legs hitting is higher than the independent calculation suggests.
These models are built by quant teams with PhDs and massive datasets. They're very good. But "very good" still leaves cracks, and those cracks are where your edge lives.
1. New Player Combinations
Correlation models are trained on historical data. When a player changes teams, gets a new role, or plays with a new lineup, the historical correlations may not apply. Early in the season, or after a major trade, the model is working with stale data.
Example: When a star player gets traded mid-season, the correlation between their performance and the new team's spread is essentially unknown to the model. It might default to league-average estimates, which could be significantly wrong.
2. Injury-Driven Correlation Shifts
When a key player is out, every correlation in the game shifts. If the starting RB is injured and the backup is in, the correlation between rushing yards and team success changes. The model might adjust for the player being out, but it often doesn't fully recalibrate all the downstream correlations.
This is especially exploitable in the NFL, where a single injury (starting QB, #1 WR, lead RB) can reshape the entire game's correlation structure.
3. Weather and Venue Factors
Models incorporate weather, but often as a simple adjustment rather than a full correlation recalibration. A game played in 25 mph winds doesn't just reduce passing yards — it changes the correlation between passing and winning. In extreme wind, the team that runs the ball better wins, and the correlation between rushing props and the spread strengthens dramatically.
Dome vs. outdoor, altitude (Denver), extreme cold — these factors shift correlations in ways that models may underweight.
4. Cross-Market Correlations the Model Underweights
Some correlations span market types that the model treats as loosely connected. For example:
Sportsbooks have the most data and the best models for high-volume markets (spreads, totals, major player props). For niche props — blocks, steals, pitcher outs, hockey assists — the correlation estimates are less precise. These markets are where you're most likely to find mispricing.
You're not going to out-model DraftKings' quant team. That's not the game. The game is finding specific situations where the model's assumptions don't match reality — new player contexts, injury cascades, weather shifts, and cross-market blind spots. Those are the edges.
You don't need a PhD in statistics to build correlated SGPs. You need a systematic way of thinking about games. Five steps.
Before you open a sportsbook app, ask yourself: What's the story of this game?
Every game has a likely narrative. Maybe it's "the Thunder are going to blow out a bad Wizards team." Maybe it's "this is a low-scoring, grind-it-out NFC North rivalry game in December." Maybe it's "both offenses are elite but the defenses are suspect, so this is going to be a shootout."
The narrative is your hypothesis. It tells you which direction the correlations should flow.
Good narratives are specific and falsifiable:
Every game has a primary driver — the player or factor that most determines the outcome. Find that driver and build your SGP around them.
Once you have your driver and narrative, map the cascade: If my primary driver performs as expected, what else becomes more likely?
Example cascade for "Josh Allen has a big game":
Now open the sportsbook and build the SGP. Compare the SGP price to what you'd expect based on your correlation analysis.
Quick pricing check method:
1. Note the individual odds of each leg
2. Calculate the "independent" parlay price (multiply implied probabilities)
3. Compare to the SGP price the book offers
4. The difference is the correlation adjustment
If the book is charging a 15% correlation tax on legs you believe are 30% correlated, you might be getting value. If the book is charging a 25% tax on legs you think are only 10% correlated, you're getting ripped off.
This isn't exact — you're estimating, not calculating precise copulas. But it gives you a directional sense of whether the book's pricing is in the right ballpark.
Different sportsbooks use different correlation models. DraftKings, FanDuel, BetMGM, and Caesars will all price the same SGP differently — sometimes significantly.
Always check at least 2-3 books before placing an SGP. The same 3-leg SGP might be +350 on DraftKings and +400 on FanDuel. That's a massive difference, and it tells you one book's correlation model is more aggressive than the other's.
This is one of the most underutilized edges in SGP betting. Most recreational bettors build their SGP on one book and never check if they're getting the best price.
> Pro tip: Some books are systematically more generous on certain correlation types. Track which books give you the best prices on QB + spread combos, player props + totals, etc. Over time, you'll develop a mental map of which book to use for which SGP type.
Theory only matters if you can apply it. Here are four SGP builds — three smart, one trap — that show how correlation analysis works on real slates.
Game: Thunder (-8.5) vs. Wizards | Total: 228.5
Narrative: The Thunder are significantly better than the Wizards. SGA should dominate, and the Thunder should control this game from start to finish. The Wizards don't have the defensive personnel to slow SGA, and their offense is too inconsistent to keep pace.
Primary driver: Shai Gilgeous-Alexander
Correlation map:
Why this works: I'm taking the alt spread at -4.5 instead of -8.5 to reduce the blowout/minutes risk. SGA scoring 30+ and the Thunder winning by 5+ are strongly correlated. Adding Williams gives me a secondary scorer who benefits from the same offensive environment. All three legs are positively correlated through the same mechanism: Thunder offensive dominance.
What I avoided: Taking SGA points over with the full -8.5 spread. That's a trap — the correlation between SGA scoring and the Thunder covering 8.5 is actually weaker than it looks because of the minutes issue. In games the Thunder win by 15+, SGA often plays 30-32 minutes instead of 36. That costs you 4-6 points of production.
Game: Bills (-7) vs. Dolphins | Total: 49.5 | Weather: 15°F, 10 mph wind
Narrative: Cold weather in Buffalo. The Bills are the better team and should control this game. Cold weather suppresses passing efficiency for both teams, but especially the Dolphins, whose speed-based offense is less effective in the cold. The Bills' ground game should be a factor. Game script should favor the Bills building a lead and running the clock.
Primary driver: Weather + Bills offensive balance
Correlation map:
Why this works: All three legs are connected through the same game narrative: cold weather, Bills control, ground-game dominance. The Bills covering means they're running the ball (Cook yards over). Running the ball and controlling the clock means fewer total possessions (under). Cold weather independently suppresses scoring (under). Every leg reinforces the others.
The correlation the book might miss: Weather-driven correlation shifts. The model knows cold weather reduces passing, but it may not fully adjust the correlation between rushing props and the spread in extreme cold. The relationship between Cook's rushing and the Bills covering is stronger in 15-degree weather than in September — and the model might not fully capture that.
Game: Lakers vs. Celtics | Total: 220.5
A recreational bettor's SGP:
1. LeBron James over 25.5 points
2. Jayson Tatum over 27.5 points
3. Game over 220.5
Why this looks correlated: "If both stars go off, the game should be high-scoring." Two superstars battling, points flying everywhere. It feels like a story.
Why it's actually a trap:
Stop and think about what you're actually betting. You're betting that LeBron scores 26+, Tatum scores 28+, and the game goes over — and you're telling yourself those three things are connected. They're not. Here's the part that stings:
A better version: If you like the over, pair it with something that actually drives scoring — total assists over, a team total, or pace-related props. Those legs share a mechanism. "Both stars go off" is a fan's fantasy, not a bettor's thesis.
Game: Braves (-155) vs. Marlins | Total: 7.5 | SP: Chris Sale vs. Marlins' #5 starter
Narrative: Chris Sale is one of the best pitchers in baseball facing a weak Marlins lineup. This should be a low-scoring game dominated by Sale's performance. The Braves should win comfortably behind Sale's pitching.
SGP build:
1. Chris Sale over 6.5 strikeouts (-130)
2. Braves -1.5 run line (-120)
3. Game under 7.5 (-105)
Why this works: Sale striking out 7+ means he's dominating. Dominant Sale = Marlins aren't scoring = Braves are winning = low-scoring game. All three legs flow from the same source: Sale dealing. The strikeouts are positively correlated with the Braves covering the run line (Sale's dominance suppresses Marlins offense), and both are positively correlated with the under (fewer Marlins runs = lower total).
This is a textbook positively correlated SGP where every leg reinforces the same narrative.
After building correlation models and analyzing thousands of SGPs, these are the mistakes I see most often — from casual bettors and experienced ones alike.
Just because legs are in the same game doesn't mean they're correlated. "Player A assists over" and "game total over" in an NBA game have a correlation close to zero. Including uncorrelated legs in your SGP isn't inherently bad — they add odds without correlation issues — but don't pretend they're part of your correlated thesis.
The fix: For every leg in your SGP, ask: "Does this leg winning make my other legs more likely to win?" If the answer is no, the leg is independent. That's fine — just don't count it as part of your correlation edge.
This is the most expensive mistake in SGP betting. Recreational bettors routinely parlay legs that actively fight each other because each leg individually looks good.
"I like the under AND I like both QBs to throw for 250+." Those views are contradictory. Pick one narrative and commit to it.
The fix: Before submitting any SGP, scan for negative correlations. Ask: "Is there any scenario where one of my legs hitting makes another leg less likely?" If yes, you have a problem.
"Mahomes passing yards and the Chiefs winning are correlated" is true. But the correlation isn't 1.0. Mahomes has thrown for 300+ yards and lost. He's thrown for 180 yards and won. The correlation is real but imperfect, and bettors who treat moderate correlations as certainties are overexposing themselves.
The fix: Think in probabilities, not certainties. A correlation of 0.5 means the legs move together more often than not, but there's still a huge range of outcomes where they diverge.
Finding correlated legs is only half the battle. The sportsbook knows they're correlated too, and they're charging you for it. If you build a perfectly correlated SGP but the book's correlation tax fully accounts for (or overcharges for) the correlation, you have no edge.
The fix: Always compare the SGP price to what you'd expect from independent pricing. If the discount is larger than you think the correlation warrants, you're paying too much.
"It's a rivalry game, both teams will be fired up, so I'll take both teams' defensive props over." That's a narrative, not a correlation. Being "fired up" doesn't have a measurable, consistent statistical relationship with defensive prop outcomes.
The best SGP bettors distinguish between statistical correlations (measurable, repeatable relationships in the data) and narrative correlations (stories that feel logical but don't show up in the numbers).
The fix: If you can't point to a specific mechanism — "when X happens, Y becomes more likely because of Z" — the correlation might be a narrative, not a statistical reality. Test your assumptions against historical data when possible.
Every leg you add to an SGP increases the correlation tax and reduces your edge. A 2-3 leg SGP with strong correlations is almost always better value than a 6-leg SGP where only some legs are correlated.
Sportsbooks love long SGPs because the correlation tax compounds. By the time you have 6+ legs, the book's margin is enormous — even if some of the legs are well-correlated.
The fix: Keep SGPs to 2-4 legs. Every leg should earn its spot by being part of your correlated thesis. If a leg is just "I like this prop," it doesn't belong in a correlated SGP — bet it straight.
SGP correlation refers to the statistical relationship between different legs (bets) within a same-game parlay. When two legs are positively correlated, one outcome happening makes the other more likely — like a quarterback throwing for 300+ yards and his team winning. When legs are negatively correlated, one outcome makes the other less likely — like both teams' quarterbacks going over their passing yards while the game goes under the total. Understanding these relationships is the foundation of building profitable same-game parlays.
Sportsbooks use mathematical models called copulas along with correlation matrices to estimate how connected each combination of SGP legs is. They calculate the joint probability of all legs hitting simultaneously — which differs from simply multiplying individual probabilities — and then add their margin. Positively correlated legs pay less than an equivalent traditional parlay because the true probability of all legs hitting is higher than the independent calculation suggests. This adjustment is often called the "correlation tax" or "SGP tax."
Yes, but it's harder than most betting content suggests. The edge comes from finding specific situations where the sportsbook's correlation model doesn't match reality — new player combinations the model hasn't seen, injury-driven correlation shifts, weather impacts, or cross-market correlations the model underweights. You won't find +EV SGPs on every game, and the edges are typically small. But with a systematic approach and discipline, it's possible to identify spots where the book's correlation pricing is off.
Compare the SGP price to the "independent" parlay price. Calculate what the parlay would pay if you multiplied the individual leg probabilities together (as if the legs were independent). The difference between that number and the actual SGP price is the correlation adjustment. If the adjustment seems larger than the actual correlation warrants — for example, a 20% discount on legs you believe are only weakly correlated — the book is likely overcharging. Checking the same SGP across multiple sportsbooks also helps; if one book offers significantly better odds, the others may be overpricing the correlation.
Most bettors don't have access to the same copula models sportsbooks use, but you can approximate correlation analysis using historical data and basic statistics. [ParlayIQ's Correlation Engine](/tools/correlation-engine) analyzes the statistical relationship between SGP legs and identifies which combinations are positively correlated, negatively correlated, or independent — helping you build smarter same-game parlays and avoid common correlation traps. You can also build your own correlation estimates using free sports reference sites and a spreadsheet, though it requires more manual effort.
This guide is part of ParlayIQ's [SGP Strategy Series](/guides/same-game-parlays).
Last updated: February 2026