GPP vs Cash: Which Players to Target in Each Contest Type

Cash games pay you for the left side of a player's distribution. GPPs pay you for the right side, divided by how many people already own it. Every difference below follows from that one sentence. If you take nothing else away, take this: you are not picking "good players" in either format. You are picking players whose shape matches the way the contest pays.

The payout curve decides which statistic matters

A 50/50 pays roughly the top 44% of the field the same amount. That flat structure has a consequence people underrate: every point you score above the cut line is worth exactly nothing. You are not maximising your score. You are maximising the probability that your score clears a line.

A tournament inverts it. The top 1% takes most of the prize pool, so beating the median wins you approximately nothing — min-cashing a GPP frequently returns less than the buy-in after you account for the entries that missed. You are maximising the probability of a score three or four standard deviations above your projection.

Same slate. Same projections. Two completely different questions asked of the same distribution. A player with a 14-point projection who scores between 11 and 17 every week is enormously valuable in the first question and nearly worthless in the second. A player projected for the same 14 who scores 4 most weeks and 38 occasionally is the reverse.

The single most common mistake in DFS is building one lineup and entering it in both formats. It cannot be right in both. The distributions that win a double-up are the distributions that lose a Milly Maker.

Leverage only exists in one of the two formats

In a 50/50 you are paid for being right. In a tournament you are paid for being right when other people are wrong, because the prize is divided among everyone who was also right. That is the entire concept of leverage, and it has a workable definition:

leverage = your lineup's chance of a top-1% score ÷ the field's ownership of it

Consider a running back owned by 55% of the field who goes for 30 points. In cash, that is a clean win — he cleared his salary and you made the cut. In a tournament, he did almost nothing for you. Every one of the thousands of entries that beat you also had him. He didn't differentiate you; he just kept you from falling behind. We go deeper on this in ownership and leverage explained and on when to fade in chalk vs contrarian.

Cash archetypes: what to target

Everything here is downstream of one idea — you are buying predictability, and you are willing to pay for it in upside you will never be paid for anyway.

• Volume over efficiency. Touches, targets and minutes are dramatically more stable week to week than yards per carry or three-point percentage. Usage is the most autocorrelated input in fantasy sports; efficiency is mostly noise wearing a number. A back with 18 carries and 4 targets has a floor. A back averaging 6.1 yards a carry on 9 carries has a coin flip.
• Confirmed roles and confirmed starters. A cash lineup has no room for a player who might sit. Late scratches are the single largest source of preventable cash-game losses, and the fix is simply refusing to roster ambiguity.
• Pass-catching backs. In PPR, a reception is the highest-floor scoring event in football: it requires nothing but a target, no broken tackle, no goal-line role, no touchdown variance. Five catches is 5 points before a single yard.
• Game-script-neutral players. Someone who gets his work whether his team leads by 14 or trails by 14. A bell-cow back on a heavy favourite has a great median and a poor floor, because the version of the game where his team trails is the version where he sees eight carries.
• Value that clears the threshold. Roughly 2.5x salary per $1,000 on DraftKings, about 2.0x on FanDuel. Cash lineups are built by finding the cheapest players who clear that bar reliably, then spending every dollar saved on the safest expensive players available.

What to avoid in cash: touchdown-dependent players, because a touchdown is a binary event inside a continuous payout structure. Deep-threat receivers with four targets and a 16-yard average depth. Pitchers with walk problems, whose bad outcomes are catastrophic rather than merely poor. And generally, anyone whose median outcome sits well below their mean — that gap is the tell that a projection is being propped up by rare outcomes.

GPP archetypes: what to target

Here you are buying the right tail, and you are willing to accept a much worse median to get it.

• Right-skewed players. You want a large gap between the 90th percentile and the median. A receiver with four targets who might catch two of them for 90 yards and two scores is a better tournament play than a possession receiver with an identical projection, because the projection is an average of outcomes you are not being paid for.
• Correlated groups, not individual players. GPP lineups win as units, not as nine independent bets. Stack the passing game, bring one opposing player back, take the game you think goes to 55 points. The lineup that wins a large tournament is almost always one where several players had their ceiling simultaneously — which only happens if their ceilings are causally linked. See stacking and correlation.
• Players whose ownership is suppressed by a story rather than by their projection. A bad-weather narrative that the wind readings don't actually support. A tough-on-paper matchup that the underlying defensive numbers contradict. A national-broadcast pivot that everyone makes at the same time, leaving the obvious play underowned.
• Tournament punts that unlock the top of the board. The sub-$4,000 player with a genuine path to 15 points, rostered specifically so you can pair two studs nobody else can afford together. The punt does not need to be good in expectation. It needs to be occasionally viable and structurally enabling.
• Contrarian quarterbacks and pitchers. These are the highest-leverage roster spots in football and baseball because ownership concentrates hardest at the top of those positions — and because both positions anchor stacks, so being different there makes the whole lineup different.

What to avoid in GPPs: the chalk value play whose entire appeal is that he is safe. He is in 60% of lineups. He wins you nothing you did not already have, and he costs you the roster spot you needed to be different.

Position by position

The NHL row deserves a note, because it is the most counterintuitive line in the table. Rostering the goalie who is facing your own stack sounds insane — you are betting on goals being scored and on goals not being scored at once. But the outcome that wins a tournament is the one where your stack scores four and the goalie still wins 4-3 on 38 saves, and essentially nobody else in the field has constructed that. That is leverage expressed as correlation, and it only makes sense where the payout is top-heavy.

The overlap rule

Build both from the same slate and compare them. If your cash and GPP lineups share more than about a third of a roster, one of them is wrong.

Take a nine-slot NFL build. A reasonable outcome is three shared players: the pass-catching back whose floor is real and whose ownership happens to be moderate, one receiver you like in both contexts, and a value lineman-adjacent play that both builds need for salary reasons. The other six differ — the cash build takes the safe QB and the volume receivers, and the GPP build takes the shootout QB, two of his pass-catchers, a bring-back, and a punt that pays for a stud.

That is what a correctly-differentiated pair looks like: same projections, same slate, two different bets.

Reading this off the simulator

This is precisely what the contest-type toggle at the top of the projections page controls. Set it to Cash and the simulator optimises the probability of clearing the cut line, ignores ownership entirely, and stops warning you that your entries are similar — because in a 50/50 they should be. Set it to GPP and it optimises the tail, applies the leverage term, and starts caring very much about how many distinct bets your portfolio actually represents.

On the results page, the numbers to read are chance to cash in cash mode and top 1% in GPP mode. Both are drawn on each lineup's distribution curve as a shaded region, so you can see the probability rather than take it on trust — and the visual gap between the two shaded areas is the whole argument of this article in one picture. In GPP mode also watch distinct bets: if you entered 50 lineups and that number reads 6, you have bought fifty entry fees and six wagers.

For more on how the underlying simulation works, see what Monte Carlo simulation actually tells you and how to read simulation output. For the bankroll side of the decision — how much to put in each format in the first place — see cash game strategy, contest selection, and how many lineups to enter in a GPP.

The rule, restated

Cash games ask how often you are good. Tournaments ask how big you can get when nobody else did. Build for the question you are actually being asked.