What is the difference between DFS mean, median, and ceiling?

Mean, median, and ceiling are three readings of the same set of simulated games. Mean is the average. The median is the middle game. The ceiling is a strong game, not the single best draw in the set. They come apart because DraftKings scores are right-skewed: most weeks cluster, and a few explosive weeks pull the average up.

If you only remember one line, remember this. The mean tells you what the player is worth if you could play the week many times. The median tells you the week you should expect to actually see. The ceiling tells you how good the less common week is. Cash lineups live closer to the median. Large tournaments are built to catch the ceiling.

Mean is the expected score

The mean column is the expected DraftKings score across the simulations. Yardage bonuses are inside it. DraftKings adds 3 points at 300 passing yards, at 100 rushing yards, and at 100 receiving yards. The mean counts how often those bonuses show up. It does not pretend every player gets one, and it does not pretend nobody does.

That bonus is rare for most players and ordinary for a few. When it lands, it adds a lump of points that a smooth per-yard formula never shows. A receiver who finishes at 99 yards and a receiver who finishes at 100 are a full 3 points apart on DraftKings, plus the tenth of a point for the extra yard. Averaging many games is how that cliff becomes part of the expectation instead of a coin flip you have to call by hand.

The Base column is the same expected stats scored without those bonuses. It is off by default. Turn it on from the Columns menu when you want the volume story with the spikes taken out. Use Base when you are asking how much of the score is just the carries, targets, and yards. Use mean when you want the score you would expect over many repeats of the week, bonuses included.

Lineup builds on this site add up the mean. The stacks board ranks on the mean as well, and prints p90 next to it so the ceiling of that stack is visible. If you are comparing a projection to a salary, mean is the numerator.

The median is the typical game

p50 is the median of the same draws. Half the simulations score that or less, and half score more. It is not a second model. It is the middle of the list the mean already averaged.

The median sits below the mean when the distribution has a long right tail. A receiver can pile up ordinary games and then have a handful of games with a long touchdown and a 100-yard bonus. Those games raise the average more than they raise the middle. If you only look at the mean, you will think the typical week is better than it is. The player is not projected to score the mean. The player is projected to score a range whose average is the mean.

That distinction matters in cash games. Double-ups and 50/50s pay you for finishing near the middle of the field, or just above it. You need the week that happens often. A player whose mean is carried by a thin tail can miss that bar more often than the mean suggests. p50 is the more honest number for that format, because it refuses to let two spike weeks speak for the other eighteen.

p50 is on the projections table, and it is off by default. Open Columns and turn it on. mean and p90 are already showing.

Ceiling is a strong game, not a fantasy

p90 is the ceiling on this board. About 10% of simulations score that or higher. It is where the explosives and the bonuses show up. It is not the maximum the model can draw, and it is not a promise that the player scores that number this week. Ten percent is a real rate. In a 20-week season it is about two games. In a single Sunday it is an outcome you are hoping to land on, not an outcome you are owed.

p10 is the other end. It is a down game: the score that about 10% of simulations fall at or below. The gap from p10 to p90 is the range you are actually betting. A narrow gap means the week is relatively steady. A wide gap means the score depends on which version of the game shows up.

Large-field tournaments pay the right tail. Two players can share a mean and still be different tournament plays if one has a much higher p90. The gap between mean and p90 is the part of the projection a tournament lineup is reaching for. A high mean with a tight range is a cash profile: you are mostly getting what you paid for. A similar mean with a tall p90 is a tournament profile, if the salary and the rest of the roster can survive the weeks when that ceiling misses.

Ceiling without a role is not a plan. p90 still has to come from the plays, the snaps, and the share. A player with a tiny target share does not grow a ceiling because the format is a tournament. The path from volume to a big score is the subject of how these DraftKings projections are built.

A short example

Picture twenty simulated games for one player. This is an illustration, not a live projection. Most of the games land between 10 and 16 points. Two of them land near 28 because a touchdown and a yardage bonus both hit.

Line those twenty scores up and pick the middle one. That is the median, and it stays in the teens. Average all twenty and the mean is higher, because the two spikes count fully in the average and barely move the middle. The 90th percentile sits up with the spikes. Nothing about the player's salary changed between those three numbers. The question changed.

Roster that player in a double-up and you are mostly buying the teens. Roster that player in a large tournament and you are buying the chance that one of the spikes is this week. Both can be reasonable. They are not the same bet, and a single projection cannot tell them apart.

How to read the board

Read mean, p50, and p90 together. If p50 sits close to the mean, the week is relatively steady and the average is a fair picture of the typical game. If p90 runs well ahead of the mean, the score depends on a less common game, and the mean is being pulled up by that tail.

• Cash and double-ups: start with p50, then check that the mean is not far above it. A large gap means the average is leaning on spike weeks.
• Large tournaments: start with the gap from mean to p90. A tall ceiling only helps if you can roster it without the rest of the lineup needing that same spike.
• Either format: check the volume next to the score. Targets, rushes, and snap share are what make the range believable.

None of the three numbers is the player's projection in isolation. The projection is the distribution. mean, p50, and p90 are three labels on it.