Start or sit without vibes: the four numbers we check every Thursday

Why Thursday

A start or sit question has a deadline, and the deadline is the earlier of the two players' kickoffs. Most weeks that is Thursday night, and by Thursday the two things the call depends on have both arrived: sportsbooks have posted their player prop markets for the week, and the first practice reports of the week have named who is limited and who is out. Before Thursday we are guessing about both. After Thursday one of the players may already be playing.

This season the first week is the exception. Week 1 opens on Wednesday, September 9, with the Patriots at the Seahawks, so anyone with a player in that game has a Wednesday deadline. Every call we publish locks at the earlier kickoff exactly as it stood and is never edited afterwards. That is why the four numbers below are checked on a schedule and not when someone asks.

Number 1: what the books say he will score

The first input is a projected fantasy points figure built from the player's own prop markets. A receiver has a receiving yards line, a receptions line and an anytime touchdown price. Each of those markets carries the sportsbook's fee, so the first step is to strip it out: the paired over and under prices are devigged with the power method, and the yes or no touchdown market is inverted through a Poisson rate. What is left is the market's expected value for each stat, which converts to fantasy points in your scoring.

This is the number nobody else in fantasy uses, and it is the best one available, for one reason. The people who set it lose money when it is wrong. A projection from an analyst costs the analyst nothing if it misses. A line at a sportsbook that misses gets bet into the ground within the hour.

Number 2: how much of him is priced

A market projection is only as good as its coverage, so the second number is completeness: of the markets that would have to exist for the projection to be the whole picture at his position, how many actually do?

Source: the scoring model's position weights.

A receiver with all three markets posted is complete, weight 1.0. A receiver with only a touchdown price is at 0.3, and the call says so. Completeness matters because a missing market is never interpolated, never borrowed from a similar player and never dressed up as a projection. It is reported as missing, and the market side of the call carries less weight as a result.

Number 3: where drafters took him

The third number is consensus ADP, the same price that runs the rankings board, converted to a percentile within his position for the week. It carries whatever weight the market does not.

The blend is fixed and public. The market projection can carry at most 65 percent of the call, and only when completeness is 1.0. At completeness 0.5 it carries 32.5 percent. ADP carries the rest. With no market at all, the call is pure ADP and the page labels it "ADP and news based, no player market posted."

The two inputs live on different scales, per game points against a season draft slot, so they are never added as raw numbers. Each is converted to a within position percentile first, and the blend happens between percentiles.

Number 4: whether he is playing

The fourth number is a gate, not a weight. If the player's official status on Sleeper lists him as out, on injured reserve, doubtful, suspended, inactive, on the exempt list, or any of the other designations that mean he will not take the field, he cannot be the start, whatever his other three numbers say. The call then names the reason: "listed out on the injury report, so this is not a close call." A headline on our news wire never sets this gate on its own, because a headline can be about a different player with the same name.

Questionable is not on that list. A questionable player is scored on his numbers, because a questionable tag is a coin the team has not flipped yet, and the call says the tag is there.

How the four become one sentence

The reason line on every call is built from these four numbers and nothing else, in a fixed order. If one side is gated out, that is the whole reason. Otherwise the two market projections and their gap lead, and if the gap is under 1.5 points the call says it is inside the noise of a single catch and should be treated as close to a coin flip. Then the ADP figures for both. Then, if the market coverage on the start is under 75 percent, a sentence saying so.

Right now, before Week 1 props post, most calls are ADP and news based and labelled that way. The live Week 1 call on Bijan Robinson or Jahmyr Gibbs on September 3 reads "Start Jahmyr Gibbs," and beneath it "ADP and news based, no player market posted." The two are one pick apart on the board. That is an honest coin flip with a lean, and the label is there so nobody mistakes it for a priced call.

What the record will show

As of September 3 the record page lists 577 calls published and none graded, because grading runs each morning off final box scores and the first games have not been played. A call is right when the player we said to start outscores the other in your format. A tie counts as neither. The hit rate reads "nothing graded yet" rather than zero percent, because a percentage of nothing is not a number.

What to do with this

1. Check your lineup on Thursday, not Sunday morning. That is when props and practice reports are both in, and it is before the first kickoff takes a player off the table.
2. Read the basis label before the verdict. "Priced against the betting market" is a different quality of call from "ADP and news based," and you should weigh your own information more heavily against the second.
3. When the gap is inside 1.5 points, the call is telling you it is close. Use the coin flip on something the numbers cannot see, like weather or your league's scoring quirks, and stop looking for a stronger opinion.