Weekly Brief · Week 3

To: Game Ops and Trading
From: BI (a job candidate's work sample)
Date: Wednesday, September 23, 2026
Re: Lines that stay beatable, same-game stacks, and the 4-pick Power Play

Bottom line: some PrizePicks lines can be beaten by a skilled player, most of them stay that way until kickoff, and same-game stacks leak hold in a way that repeats, where the sharp players' swing was mostly luck. The three most urgent fixes are in the Action box, in order; each finding below ends with its own recommendation.

Real lines · simulated playersHow this is computed →

13 of 47beatable lines were fixed by kickoff in Weeks 1–2. Most stay open.

A day before kickoff, 47 lines were beatable: one side's fair chance (our estimate from sportsbook prices) was above 54.2%, the lowest breakeven on the payout table. That side went on to win 58.5% of the time, against 50.7% for every other line. Only 13 of the 47 moved toward the sportsbooks by kickoff, and the count still grew to 70 as new gaps opened in the last day.

Recommendation: Auto-flag any line that stays beatable for 12+ hours into a trader queue, starting with Receptions, where only 4 of 26 beatable lines moved.

See the Line Recap →

12%kept from same-game stackers, against 26% expected. That gap is built into how stacks pay.

Simulated players who stacked a QB with his receivers won fewer picks (45.6%) but left PrizePicks only 12% of their entry fees (its hold), against 26% expected. Their picks win together, and Power Plays pay on exactly that. Sharps missed by more: expected to leave 4.9%, they were paid 32% more than they put in. But that is mostly luck: 50 accounts picking the same few dozen lines win and lose together.

Recommendation: Review same-game stacking limits on Power Plays (QB + receiver Mores first) before tightening limits on sharp accounts. The stacking gap will repeat; the sharp swing mostly won't.

See Player Groups →

56.2%per pick to break even on a 4-pick Power Play, more than the 3-pick or 5-pick.

At 10x, a 4-pick Power Play needs 56.2% per pick to break even, more than the 3-pick (55.0%) or the 5-pick (54.9%). Skilled players will skip it, so its entry fees lean casual. The cheapest entry to beat is the 6-pick Flex Play at 54.2%, and that sets the beatable-line cutoff across this console.

Recommendation: Decide whether that's on purpose. At 11x it would need 54.9%, level with the 5-pick, and keep about 6 points less from a 50% player. If it's a deliberate high-margin entry, keep it and track who plays it.

Try it in the Payout Lab →

21%of whole-number Receptions lines ended in a tie, quietly voiding the pick.

49% of Receptions lines were whole numbers, and about 1 in 5 of those tied. A tie voids the pick and drops the entry to a smaller payout. Goblin and demon lines, by contrast, landed within 0.8 points of their fair chance across 4,975 lines, the best evidence that the model behind these numbers is right.

Recommendation: Size what ties cost in hold and in player experience on Receptions and Pass TDs, and test .5 lines on low counts.

See ties on the Line Recap →

6lines are beatable on this week's board of 239, and only 64% have a sportsbook price.

PrizePicks posts most lines before sportsbooks post player props, so early in the week those lines have nothing to check them against. The board refreshes daily and highlights any line the books disagree with.MarShawn Lloyd Rec Yds · ATL @ GBPrizePicks6.5Books5.5Less 57.8%

Recommendation: Look at MarShawn Lloyd Rec Yds first. Keep tighter limits on lines with no sportsbook price until the books post.

Open the Live Board →

25 pickswas enough for CLV to flag 100% of simulated sharps, a best case. Win rate needed hundreds.

CLV (closing line value) asks whether a player's picks were still the better side at kickoff. With a real edge near 5 points, wins and losses need hundreds of picks to separate skill from luck; CLV does it in tens. In the simulation, a 53% CLV rule caught every sharp and flagged no casual players. That is a best case, because the simulation defines sharps with the same market prices CLV measures. Real accounts will take more picks, but the order holds: CLV first, results second.

Recommendation: Queue accounts for review at an average CLV of 53% or more after 25 graded picks. Treat line-shopper flags as a pricing fix (match the other apps' lines), not an account action.

See the flag rule tested →

Sources: 1,001 PrizePicks lines from Weeks 1–2 graded on nflverse box scores, the live Week 3 board, and a 1,000-player simulation. Every figure has a CSV export or its SQL on Method & SQL.