How it works

We could tell you
the model is brilliant.
Here's the receipt instead.

This page explains exactly what our model does, where it's good, and where it isn't, including the test our own projections failed.

We tested our projections. They weren't good enough to sell.

Most tools launch with "our proprietary model gives you an edge." We built ours, then ran an honest backtest on a completed NFL season. It walks forward week by week, never peeks at the answer, and is graded against the dumbest baselines we could find.

PREDICTOR (2024 · 3,694 player weeks)MEAN ABSOLUTE ERROR
our projection model5.11
a simple trailing average5.26
just using last week's score6.49

Our model won, by 3% over a trailing average. We tried 48 parameter combinations to push it further. None of them mattered. Then we checked it against Vegas player props, and the market was sharper than we were.

A 3% edge over a trailing average is not a product. So we didn't sell it as one.

That's the moment most DFS companies would have shipped anyway and written the marketing copy. We went looking for where an edge actually exists.

Points are a commodity. Ownership isn't.

Here's what that backtest really taught us. Player points are priced by professionals with millions of dollars on the line. You are not going to systematically beat that market at projecting them, and anyone who tells you they do is selling something.

But nobody prices what the field will do. In a tournament, you don't win by projecting Ja'Marr Chase 1.2 points better than everyone else. You win by knowing 40% of the field is already on him, and deciding what to do about it.

That decision is not automatically "fade." Chalk is usually chalk because it's the best play on the board, and blindly fading it is how people lose money while feeling clever. Sometimes you eat the chalk and take your leverage at another position entirely. What you can't do is make that call without knowing where the field is, and that's the number we build.

You don't beat the projection. You outposition the field.

So that's what we built. And ownership has a structural property that points never will:

Ownership is a closed system

Every DraftKings Classic lineup has exactly one QB. So across the entire field, QB ownership must sum to 100%. Running back sums to about 242%, which is two slots plus a share of the FLEX. That isn't an assumption we made. It's arithmetic the roster rules hand us for free.

Points prediction is unbounded, noisy, and priced by pros. Ownership is bounded, sums to a known total, and is driven by public heuristics the field applies out in the open: chase value, chase last week's big game, chase the shootout. That's a problem we can actually solve.

Open the app and you'll see the math checking itself.

At the top of every leverage report there's a line that reads "QB sums to 100% / 100% ✓ · RB 242% / 242% ✓". That's our ownership model proving, live, that it obeys roster physics.

You don't have to take our number on faith. We show you the constraint it has to satisfy, and whether it held. If our math ever breaks, you'll see the ✗ before we do.

Every number, shown

Projection, ceiling, ownership, leverage, and how each one was derived. No black box.

Your projections, our ownership

Trust someone else's numbers? Upload them and we'll tell you what the field does with them. Our projections aren't the product. The field model is.

The receipts

A winner screenshot proves nothing. For every one posted, thousands of losing lineups go unmentioned. That's survivorship bias dressed up as evidence, and you should discount it wherever you see it.

So we publish a public track record instead: every play we flag, graded, wins and misses together. Not the highlights. The whole thing. Free to everyone, including people who never pay us.

If we're wrong, you'll be able to prove it. That's the point.

The track record goes live with Week 1 of the NFL season.

Who's behind this

Slate Signal is built and run by one person in Ohio, a project cost analyst who spends his days in financial models and his nights building this one. Not a media company, not a content farm, no room full of analysts.

That's a disadvantage in some ways. It's an advantage in the one that counts: there's nobody here whose job depends on telling you the model is better than it is. When the backtest said our projections were mediocre, there was no marketing department to overrule it, so we shipped the finding and rebuilt the product around what actually worked.

We show you the work. Then you decide what it's worth.

What we won't tell you

Ready?

See what the field is missing.

Start free 7 day trial

Or use the free plan forever. No card.