Analysis
DFS research and lineup-construction notes from GameScript: game environment, correlation, ownership, and what the numbers do and do not support.
-
One snap, one benching: the two cleanest calls this slate's research made both lost
Week 8's research produced exactly two claims that cleared every bar this site holds itself to — a leverage call four independent ownership reads agreed on, and a lopsided-script read four independent projection methods converged on. Lamar Jackson's day ended on the first offensive snap of the game. Arizona's did not go the way a blowout-favored total said it would, either, once its backup quarterback got benched for a different backup in the third quarter. Neither failure says the research was wrong. What each one actually says is different, and worth separating.
-
53 to 54.5, 7.5 to 10: one Week 8 number resolved itself three days out, and the other moved the wrong way
A four-day research log on the Week 8 2026 main slate shows BAL@BUF's real total climbing for three straight days while Lamar Jackson's own status went from OUT to Questionable to completely unverifiable — and only leveled off once a practice report finally gave the market something to confirm. On the same slate, IND@JAX's book-to-book spread gap did the opposite: it grew every day there was a second reading, with no injury news big enough to explain it. What the direction of each gap says about which numbers to trust three days from kickoff.
-
Three points apart on paper, fourteen apart on the real board
Week 8's own projection-sum math rates BAL@BUF and CLE@PIT almost identically. The real market total splits them by fourteen points, in opposite directions, for two different and traceable reasons. A worked example of why a bottom-up projection ranking is not a substitute for the total.
-
62, then 22: the leverage score that didn't get weaker — the market just caught up
A Week 5 2026 quarterback carried this slate's single highest leverage score for two straight research passes while his ownership sat frozen at the published floor — even after his injury designation cleared and a named reporter said the team planned to start him. Then, with the beat record no further along than it had been for two days, the number closed on its own. The arc, pass by pass, and the one question that actually separates a real edge from a number waiting to be corrected.
-
PHI -1.5 to CAR -3.5 and back to PHI -1.5: the line this research solved, then un-solved
A seven-pass research thread on Week 6's CAR@PHI line ended by flagging GameScript's own house number as a possible internal bug — a backwards-sign ingest error named in this product's own sourcing docs. Two passes earlier, the same research had already found the real explanation and confirmed the market agreed with the house number. Carolina won 25-14. What happened in between, and why a resolved question reopened itself.
-
18 carries to 6: the Miami backfield split this site called 'reopened' never moved
A week of real usage data crowned Ollie Gordon II as Miami's clear lead back over Jaylen Wright, and this site's own research agreed, repeatedly, as Gordon's ownership climbed. Then a Thursday injury tag on Wright and a reported coaching hint led the same research to call the backfield 'reopened' and cut Gordon's ownership by nearly half. Sunday's carries were 18 to 6 — the same lead, slightly wider than the week that first established it. What got conflated, and the distinction that would have caught it.
-
6-77-1 against 2-22-0: the ownership order a week of research flagged and lock never fixed
GameScript's own house pool spent the entire Week 5 2026 run projecting Ja'Marr Chase well above Tee Higgins while pricing Higgins' ownership higher anyway — a pool disagreeing with itself, not with the market. Two outside tools agreed with the projection. The box score did too, by 4.7x. What a model's own internal contradiction is worth, and why a different last-minute flag on the same slate's other backfield wasn't.
-
336 yards, 72 plays, zero touchdowns: what a dominated game script actually bought
Minnesota went into Week 4 as a double-digit home favorite over an 0-3 Dolphins team, in the slate's lowest-total game, with its own research flagging the backfield as the clean, lead-and-run read. Minnesota out-gained Miami by 138 yards and ran 35 more plays — and scored no touchdowns at all. The one touchdown in the game went to the side getting run off the field. What the final box score says about reading a blowout-shaped line.
-
25-for-34, 0 touchdowns: the QB question this site tracked for six days wasn't why Chicago won
This site's own Week 4 research spent six straight days and dozens of claims resolving one question — would Tyson Bagent or Case Keenum start for Chicago. It got the call right, hours before kickoff. The game was decided by a backfield split the same research flagged once, on day one, as having no clear read, then never revisited. A dated post-mortem on where research attention went and where the fantasy points actually came from.
-
Seattle won by 24. The line everyone suspected was wrong wasn't.
A piece published here on September 17 flagged GameScript's own SEA@ARI house line as a suspicious outlier, 6.5 points off DraftKings' board. Three days later, DraftKings had moved to meet it — and the game that followed blew past every number on the table anyway. What actually happened, and what the resolution gets right and wrong about trusting a disputed line.
-
1:00 lock, 2:55 answer: the injury call a Sunday slate forces blind
Without late swap, a DraftKings main slate locks at its earliest kickoff, but the NFL's own inactives rule — the list posts roughly 90 minutes before each game, not before the slate — means afternoon-window players' official status regularly arrives after the build is already locked. A Week 3 Baltimore receiver whose real Rio kickoff pushed that gap to nearly two hours is the clean, dated example. The mechanism behind it runs every single week.
-
-3.8 vs. -0.8: two backup quarterbacks lost their jobs the same week, and only one total believed it
New York and Washington both had their Week 3 starting quarterback go down from a Week 2 injury that turned out worse than first believed, discovered on the same Tuesday. One game's total dropped almost four points and grew its own wind-and-rain story to match. The other's total barely moved. What the size of a total's reaction to the same category of news tells you that the injury report alone can't.
-
McCaffrey's IR story was real — in September 2024: three weeks of catching contaminated DFS research
A daily research scan spent three weeks doing what any DFS player does before a slate — searching for injury and depth-chart news — and kept finding real articles about the wrong year, the wrong opponent, or the wrong team. Cross-referencing every player-team claim against the actual slate pool caught most of it, produced one false alarm of its own, and still missed things a human reading quickly would miss too. The failure modes, the catches, the one retraction, and the checklist that's left.
-
-3.5 to -10: the Week 2 spread that got less settled the more sources checked it
Seattle at Arizona has carried a real spread disagreement for four straight days — and the newest number, GameScript's own house line, is now the outlier on the opposite end from DraftKings. What the gap means for reading this specific game, and why a total agrees across sources when a spread doesn't.
-
OUT Friday, unpriced Sunday: the DFS ownership lag, measured in hours
A daily research scan tracked the same DFS ownership pages, from the same publishers, every day for two weeks before Week 1 2026 locked. The gap between an injury becoming official and that injury reaching a published ownership number wasn't occasional — it was structural, and in one case it never closed before kickoff. What the lag actually looked like, and how to read a projection page's byline instead of trusting its number.
-
50%, 50%, 29%, 58%: four days inside a forecast that never settled
One Week 1 game's second-half storm chance was checked daily for four days and never stabilized — it dropped, then jumped past its own starting point, with no event explaining either move. What that means for how long a weather-driven read is actually good for.
-
Showdown captain leverage: five slots are forced, one is a decision
A Showdown entry spends 600% of ownership across six slots — 500% on five flex spots the field fills the same way, and 100% on the captain. Here is the conserved-quantity arithmetic, the proof that points-per-dollar is mathematically useless in the captain slot, what GameScript's model actually assumes about captains, and why nobody here can tell you what real captain ownership looks like.
-
Six days out, the board settles — except for one number that won't move
Atlanta named its Week 1 starter and six outdoor stadiums got their first real forecasts on the same day. What that resolves, what still doesn't, and the one line dispute that has now outlasted everything else on the board.
-
Ten days out, four Week 1 numbers still don't agree with themselves
A full environment ranking for the Week 1 2026 Sunday main slate, plus an honest account of which totals are settled, which are genuinely still in dispute, and what two teams' offensive-line and quarterback situations mean for reading them.
-
Stacking is a variance decision, not a points decision
A QB-WR stack does not raise your expected score by a single point. It widens the distribution around the same mean, which is the entire reason to do it in a tournament and the entire reason not to in cash. With the arithmetic, what a bring-back really buys, and when a naked running back is fine.
-
The Week 1 line freeze: why a total from May isn't a total from September
A Week 1 number looks exactly like a Week 8 number, but it hasn't been priced the same way. It opens at schedule release, sits largely frozen for months, and moves only at a handful of calendar events. Here's how to tell a real move from three books mirroring the same stale line.
-
Game environment first: how implied totals should shape a DFS lineup
Implied team totals tell you which games are worth playing before you argue about which players are worth rostering. Here is how to read them, where they mislead, and how to turn one into a lineup rule.
-
Cash and GPP are different constraint problems, not different moods
The same player pool and the same projections produce two different rosters because the objective function and the feasible set are different. Here is what each optimizer knob actually changes, with the arithmetic, and why one lineup cannot serve both contests.
-
Leverage is a difference, not a low number
Low ownership and leverage are not synonyms, and the industry's usual leverage ratio ranks the wrong players first. Here is the arithmetic of a field-relative edge, why a good leverage measure has to be bounded, and when the chalk is simply correct.