SGP Correlation Matrix: Which NFL Prop Combinations Actually Reinforce Each Other

The four-leg ticket I would not bet today

Years ago, when same-game parlays were the shiny new thing on UK books, I built a four-leg ticket for a Sunday primetime fixture. QB passing yards over, that same QB’s anytime TD, his WR1’s receiving yards over, and the team’s lead RB’s anytime TD. The logic felt airtight – they were all on the same team, all riding the same expected high-scoring script. The ticket lost on the fourth leg. The QB had thrown four touchdowns, the RB had not scored, and I had not noticed that the construction of the ticket was actively working against itself.

What I had built, without realising it, was a ticket where two legs were positively correlated and two were negatively correlated. The QB-throwing-TDs leg and the RB-scoring leg were directly competing for the same red zone opportunities. Every TD the QB threw was a TD the RB did not rush in. Building parlays that fight themselves is the single most expensive mistake in the SGP market, and I want to lay out the framework that has stopped me from making it.

What correlation actually means in a parlay context

Correlation in betting terms is the degree to which one outcome predicts another. Two legs are positively correlated if the success of one makes the success of the other more likely. They are negatively correlated if success on one makes success on the other less likely. They are independent if knowing the outcome of one tells you nothing about the other.

The interesting wrinkle, and the one most punters do not internalise, is that bookmakers know about correlation. They build SGP pricing engines that adjust the parlay price for the correlation between the legs you have selected. A positively correlated SGP pays out less than a true parlay of independent legs would. A negatively correlated SGP – to the extent the book will even let you build one – pays out more. The price reflects the maths.

What this means for an SGP bettor is that the right correlation question is not “are my legs positively correlated, in which case I should build the ticket”. The right question is “are my legs positively correlated by more than the book has priced them to be”. That is a harder question and a more useful one. Most casual SGP construction is built on the first question, which is why most casual SGPs lose money over time.

The cleanest positive pairings

The pairings that compound their probability cleanly are the ones built on a shared underlying driver. A QB throwing for over 280 yards is a strong predictor that his WR1 will go over his receiving-yards line – same offence, same passing game, same volume environment. A team’s lead RB going over his rushing yards is a strong predictor that the team will cover the spread, because rushing volume is downstream of positive game script.

The QB passing-yards over and his WR1 receiving-yards over have an empirical correlation around 0.55 to 0.65 depending on the team and the matchup. That is a strong positive relationship. The bookmaker’s pricing engine knows this and discounts the SGP price accordingly. The question is whether the actual correlation in the specific game you are betting is higher than 0.55 – for instance, in a script where the QB is throwing more deep balls than usual because the team is trailing. If you can identify spots where the in-game correlation exceeds the book’s baseline, you have value.

The cleanest secondary pairings: a team’s lead RB rushing-yards over with a team total over, in a script where the favoured team is expected to run out the clock. A QB anytime TD pass with the team total over the touchdown threshold, because anytime TD requires a TD to happen at all. Pickswise analysts running their season-long prop ledger to 59 winning player props and +7.7 units used variations of these structures repeatedly – not because the parlay format is generous but because the underlying logic of the legs was sound.

The pairings that quietly cannibalise each other

The negative-correlation pairings are the expensive ones. The most common: a QB’s anytime TD pass leg combined with the RB on the same team’s anytime TD rush leg. Both legs compete for the same red zone trips. A team has, on average, three red zone visits per game. Each visit produces somewhere between zero and one touchdown. If the touchdown is a passing TD, it cannot also be a rushing TD by the same offence on the same play. Adding both legs to your SGP is paying for two outcomes that are, in a meaningful sense, mutually exclusive.

A second common trap is the team total over combined with a player’s reception over for a slot receiver. The team total over usually implies aggressive scoring, which often means more deep shots and longer plays. The slot receiver’s reception count, conversely, is highest in ball-control offences where the team is moving the chains with short throws. Pace and scoring volume help, but the route distribution can shift away from the slot in high-scoring spots, which is why these legs have a slight negative correlation more often than punters assume.

The most pernicious negative correlation is the late-game-prop combination – an alt-line under on QB passing yards paired with the team’s underdog moneyline. The underdog moneyline cashing requires a comeback. The comeback requires the QB to throw a lot, which puts the alt-line under in jeopardy. Building both legs is betting against your own ticket.

Why cross-team correlation works differently

Same-game parlays force you to draw your legs from a single game. That is the structural advantage and the structural constraint at once. You can build correlated tickets cleanly within one game; you cannot mix in legs from other fixtures. Cross-game parlays do not benefit from SGP correlation pricing – each leg is priced as an independent event, and the parlay multiplier compounds at the book’s standard rate.

Within a single game, you have two teams and roughly twenty potential prop targets between them. Correlations on the offensive side of one team are generally positive. Correlations between players on opposing teams are generally weaker, sometimes positive, sometimes negative depending on the structural relationship.

The interesting case is “both quarterbacks over passing yards”. The legs sound independent – two different QBs, two different offences – but they are not independent. Both legs are leveraged on a high-scoring, pass-heavy game environment. The total going over is a strong predictor that both QBs hit their passing-yard overs. That cross-team correlation is genuinely positive, around 0.35 to 0.40, and most book pricing engines do account for it but with imperfect precision. There is occasional value in finding spots where the actual correlation exceeds the book’s adjustment.

How the matrix actually maps in practice

If you were to draw the correlation matrix for a generic NFL game, the strongest positive pairings would cluster around shared-offence yardage legs and shared-script outcomes. QB passing yards with WR1 receiving yards. Lead RB rushing yards with team-total over. QB anytime TD pass with team total over. The strongest negative pairings cluster around competing red-zone outcomes – QB rushing TD with WR anytime TD on the same team in a tight game, or RB rushing TD with QB passing TD when the offence’s red-zone trips are limited.

The matrix that matters most is the one for the specific game you are betting. A game projected as a 51-point shootout has a different correlation structure from a game projected as a 38-point grinder. The shootout produces more compounded volume – both teams’ pass-game props correlate more tightly, because the volume environment is the dominant variable. The grinder produces tighter individual outcomes – single drives matter more, and the negative correlation between competing red-zone roles is sharper.

What I do before constructing an SGP: I write down the implied script (favourite, total, projected possessions) and then list the legs I am considering. For each pair of legs, I ask whether they move in the same direction or different directions under the implied script. If three of three legs move together, the ticket is structurally clean. If one leg moves against the others, I drop it. The discipline costs me payouts, because more legs equals bigger parlay numbers. The discipline also saves me from buying tickets that were beaten before they were placed.

The cross-team SGP that genuinely works

The one cross-team SGP construction I consistently like is the high-total game built around both quarterbacks’ over passing yards. Both legs lean on the same underlying variable – game flow producing volume. The negative correlation that exists between most opposing-team props (one team’s success usually means the other team’s failure) is dampened in the QB passing case because both QBs benefit from a high-volume environment.

The same logic does not extend to opposing receivers. WR1 of team A and WR1 of team B both going over their lines is a weaker correlated pair than the QB version, because individual receivers depend on QB quality, defensive coverage and game script in ways that vary more across teams than across positions. The QB version aggregates across the offence; the WR version is a single thread inside the offence.

What I would not do is build a long SGP – five or more legs – even with strong correlation logic. The math against parlays gets steeper with each leg added, and the bookmaker’s edge expands proportionally. Three legs is comfortable. Four legs is reaching. Five legs and beyond is, in my honest view, a recreational product wearing the costume of a strategic one.

If the SGP framework has you thinking about market structure more broadly, the next read is the comparison of why UK and US sportsbooks look so different for NFL props – including how their SGP and parlay engines differ in construction.