This week's packs
DeadMoneyVault

Methodology

How the DeadMoneyVault Models Work

7 min read

DeadMoneyVault's picks come from one model per market — receptions, rushing yards and anytime touchdowns. Each prices every prop against the sportsbook market, the best are ranked into a top 10 per market, and the strongest legs become the week's 2-leg parlays. This page walks through the receptions model step by step, then covers how the rushing and touchdown models differ — what goes in, what comes out, and the guardrails in between.

Step 1 — a full catch distribution for every player

For each pass-catcher with a posted line, the model estimates the probability of every catch count, from zero upward. It combines two views:

  • A learned model trained on several seasons of play-by-play, using target share, routes, role changes, recent usage trends, matchup, game script and injury reports.
  • A team-level simulator that starts from the team's expected pass volume and splits targets among the receivers who are actually active — so when a starter is out, his targets move to teammates in a realistic way, and teammates' outcomes stay correlated.

The two are blended and calibrated on held-out seasons so the stated probabilities match how often things actually happen.

Step 2 — anchored to the market

The sportsbook price is the best single forecast available, so the model doesn't ignore it. Each projected probability is blended with the market's de-vigged fair probability, with weights learned from past seasons. The result leans away from the market only where the historical evidence supports it — which keeps the model honest and stops it from producing wild, confident disagreements that history says would be wrong.

Step 3 — only legs that clear the fair price

A leg is eligible for a parlay only if its probability is at least the market's fair probability for that side, and its price is no shorter than −250. That second rule exists on purpose: near-certain legs at −1000 or worse don't add value to a parlay, they just disguise a single bet.

Step 4 — pairs, priced for correlation

Eligible legs are paired into 2-leg parlays whose combined price lands between +100 and +300. Legs from different games are treated as independent; legs from the same game use the simulator's joint outcomes, because two receivers on one team compete for the same targets. The best pairs by expected value — with no player repeated — go on the week's board.

The other markets: rushing yards and anytime TD

The same principles — a full distribution, anchored to the market — drive the other two models:

  • Rushing yards simulates carries (driven by role and expected game script) times yards per carry drawn from each runner's real per-carry history, then blends the result with the market and calibrates it.
  • Anytime touchdown starts from how many touchdowns the team is expected to score and splits them by each player's goal-line, red-zone and overall usage share.

Each market's props are ranked by how far the model's fair price sits from the sportsbook's, and the top 10 go on that market's props board.

Step 5 — dealt as packs

The board's ranking sets each parlay's rarity, and free members open three packs a week drawn from it. See how packs and rarity work for the exact odds.

What the model can't do

It can't see news that breaks after the lines are pulled, it can't guarantee anything, and in a sharp market any real edge is small and noisy. Parlays also combine the sportsbook's margin on both legs. Treat every pick as the model's read of the market — information, not advice.


New to props? Start with what reception props are — the same ideas apply to every player prop — or see this week's packs.

Frequently asked

How does DeadMoneyVault make its picks?
Each market has its own model — receptions, rushing yards and anytime touchdowns. Each projects a full distribution for every player with a posted line and anchors that probability to the de-vigged sportsbook price. Each market's props are ranked into a top 10; for parlays, only legs priced at or above the market's fair probability are kept, paired into 2-leg parlays paying +100 to +300, and the best by expected value go on the week's board.
Does the model beat the sportsbooks?
No model can promise that. Sportsbook closing lines on props are sharp, so the model is deliberately anchored to the market and only leans away from it where years of history support it. Picks are estimates, not guarantees.
Why are so many picks unders?
Across recent seasons, player-prop overs have hit slightly less often than their prices implied — books tend to shade the popular over. Because the model is calibrated to that history, the legs that clear its fair-price bar are more often unders.

Keep reading