How to read a start or sit call that carries a number

A number is a claim, so read it like one

A start or sit call that says "start Gibbs" is an opinion. A call that says "start Gibbs, 91 to 84" is a claim with a number attached, and a number invites a question the opinion never had to answer: where did it come from? If you cannot answer that for a call, you should not act on it, whoever published it. This guide is how to answer it for the calls on the Parlae board.

Every published call has four parts: the score, the basis label, the gap between the two players, and the lock. Read them in that order.

The score: a percentile, not a projection

The score runs from 0 to 100 and is comparable only against players at the same position in the same week. It is not projected points. It is a blend of two percentiles, each computed within the position.

The first percentile is the market side: the player's own prop lines, devigged and converted to fantasy points. The build job pairs each over and under, removes the bookmaker's margin by the power method, and turns the fair probability into a projected mean. Yes or no markets, the anytime touchdown price, are converted by Poisson inversion, so a 40 percent chance of at least one touchdown becomes an expected count of about 0.51 rather than a guessed fraction. Those means become points, and the points become a within position percentile.

The second percentile is the baseline: consensus ADP from live drafts, again converted to a within position percentile. The two never get added as raw numbers, because points per game and a draft slot live on different scales.

The basis label: how much of the score is market

The blend weights the market side by 0.65 times a completeness figure, and gives ADP the remainder. Completeness answers one question: of the markets this position needs for a full picture, how many are actually priced?

Source: Parlae start or sit scoring model, completeness weights.

A receiver with receiving yards, receptions and a touchdown price is complete: the market carries the full 0.65 and ADP carries 0.35. A receiver with only a touchdown price has completeness 0.30, so the market carries just under 0.20 and ADP carries the rest. The label on the call tells you which case you are reading. Market means completeness of at least 0.75. Market partial means some lines exist but not enough. ADP means no lines at all, and the score is pure draft consensus.

That last label matters right now more than any other. As of Sep 3, the Gibbs player page read "No lines posted. Books open Week 1 props in early September," and the Gibbs and Bijan Robinson call carried the note "ADP and news based, no player market posted." A call with the ADP label is telling you it is the draft market's opinion, and the draft market is a season long price, not a Week 1 price. It is the honest fallback, and it should be read as the weakest form of the call.

The gap: when close means close

The call between Gibbs and Robinson reads "Start Jahmyr Gibbs" and the pairing is labelled close. On the compare page the two are one spot apart on the board and 0.9 picks apart in ADP, both in tier 1. That is not a decision the number can make for you. A gap that small, on any basis label, is inside the noise of the method, and the honest reading is that either player is fine and your lineup slot, your bye situation or your other starters should decide.

A useful rule: a gap under about five score points on a market basis, or a pairing labelled close on any basis, is a coin flip, and you should treat the call as a tiebreaker rather than an instruction. A gap of fifteen or more on a market basis is a real difference in what the books expect, and you need a specific reason to go against it. In between, the call is a lean.

The gate: availability overrides everything

One input is not a weight at all. A player who is Out, Doubtful, on IR, on PUP, suspended or inactive cannot be the start, whatever his score. The build reads the designation feed and applies it as a gate before the blend. A Questionable player is not gated; he is scored normally, and the call assumes he plays.

So when a call says start a Questionable player, it is a call for the version of him that is active. If the inactive list says otherwise 90 minutes before kickoff, the call is void, and the page, which computes its state from the clock and the kickoff times at render time, will show it locked as it stood.

The lock: what the call is graded on

Every call locks when the earlier of the two games kicks off and is never edited after. Once both games are final, it is graded: right if the player we said to start outscored the other one in your scoring format, wrong if he did not, and a tie counts as neither. As of Sep 3 the record page showed 577 calls published and nothing graded yet, with the first results due the day after Week 1. That zero is the honest number until games are played, and any site quoting a start or sit accuracy before Week 1 is quoting a different season.

What to do with this

1. On any call you plan to act on, read the basis label first. Market means the books are behind it; ADP means the draft market is, and you should weight it accordingly.
2. If the pair is labelled close or the gap is under five points, stop treating the call as an instruction and decide on byes, your other starters and the kickoff window, using the start or sit page as the tiebreaker.
3. Check the record page after Week 1. If the graded hit rate on market basis calls is not clearly better than the ADP basis calls, the number has not earned your trust yet, and you should say so out loud.