Credence Is Live. Here’s What We Built for Draft Day.

Fantasy football does not have a rankings problem.

It has a context problem.

By the time you sit down for a draft, you probably already have more rankings than you know what to do with. Your league platform has them. Analysts have them. ADP has them. Consensus rankings combine dozens more.

But none of that changes the fact that your league may start three wide receivers instead of two. Your scoring may reward receptions differently. Your roster may already have two running backs when an unexpected run on receivers starts in Round 6.

At that point, the question isn't really, "Who is ranked higher?"

It's: What makes sense for my team, in my league, with the players who are still available?

That's the problem we built Credence to help solve.

A Draft Board Built Around Your League

Credence starts with full-season projections for every roster-relevant player.

But we don't just hand everyone the same top 200.

Add your league and Credence re-scores those projections using its actual scoring format, roster construction, and league size. You can import a league from Sleeper or ESPN, or enter the settings manually.

That matters because player value is relative.

The replacement level in a 10-team league isn't the same as in a 14-team league. Starting requirements change positional scarcity. Scoring rules change the value of different types of production.

We also don't think a projection of 247.3 fantasy points means a player will score 247.3 fantasy points.

Every season projection therefore comes with an 80% uncertainty range. The point estimate is still useful. It just isn't presented as if an NFL season has only one possible outcome.

Every account gets one personalized league for free.

You can see what Credence thinks before paying us anything.

Where the Paid Product Becomes Different

A personalized draft board is useful before the draft.

Then the draft actually starts.

The player you expected to take is gone. Your roster develops differently than you planned. Positional runs happen. The available player pool changes with every pick.

A static ranking doesn't know any of that.

The paid Credence draft optimizer follows the draft pick by pick and updates as the board changes. It considers the players still available, along with your league settings and the roster you're actually building.

That's the distinction we care about.

If all you want is another rankings list, you probably don't need to pay for Credence.

The subscription is for the point where the decision becomes specific:

Given my league, my roster, and what has happened in this draft, what should I be considering right now?

Paid accounts also support several leagues and include a post-draft roster report that evaluates your draft against your league's own replacement level.

It does not tell you that you "won the draft." We don't hold every opposing roster in every league, and draft grades aren't season results.

It tells you what we can actually measure.

Sometimes Credence Disagrees

A useful model shouldn't exist just to reproduce consensus with different formatting.

Credence currently sees the 2026 player pool differently in several places.

Our model is notably higher than market consensus on Christian McCaffrey, Zach Ertz, Ja'Marr Chase, and Alvin Kamara.

It's notably lower on Travis Hunter, Joe Burrow, Luther Burden III, and Bhayshul Tuten.

We don't know whether those differences will prove useful.

No 2026 games have been played. The season gets to answer that question.

There is also an important nuance to how our board works. The model producing projected fantasy points never looks at the draft market. A separate model determines player ordering, and at most positions that ranking can blend our view with market consensus.

So we're not going to pretend every difference on the final board is a completely independent signal.

What Credence gives you is another quantified view of the decision — one that can then be applied to the specific league you're actually drafting in.

Why We Think You Should Trust It Enough to Try It

The day before launching paid subscriptions, we found a problem in our own projections.

We audit the model against 13 years of actual outcomes from 2013 through 2025. During that process, we found that our projected point totals were systematically running low.

Running backs made the issue particularly clear. Their projected per-game production sat at roughly 86% of what those players actually produced. Wide receivers were around 85%, and tight ends around 84%.

The first suspicion was availability.

Maybe we were expecting players to miss too much time.

That wasn't the problem.

We were underpricing what players did when they were actually on the field.

We tested a correction against two alternative fixes and a do-nothing baseline using a pre-registered statistical test and out-of-sample validation. The test returned p = 0.0002, and the correction cleared our formal overfitting check at greater than 99.9% confidence.

Across the fantasy-relevant player pool, pooled projection bias moved from -12.85 points per player to +1.41.

The fix didn't change the player rankings within each position. It corrected the projection magnitude.

Running back point totals rose about 25% on average. Wide receivers and tight ends increased roughly 10–11%. Quarterbacks decreased about 7%.

We shipped the correction on August 15, the day before launch.

Not because finding a problem is something to celebrate.

Because finding one, figuring out why it happened, testing the correction, and publishing the result is what we think running an analytics product should look like.

You Can Check Our Work

That same idea is why our historical track record is public.

Every season since 2019, we grade the Credence board against what actually happened and compare it with the draft-day market and other fantasy ranking sources.

The results are not uniformly flattering.

Against the real-money draft crowd we use as our primary comparison, our player ordering and the market's have been statistically indistinguishable across those seven seasons. The difference is small enough that it could be noise.

That's a wash, not a win.

We also trail two staff-ranking products and one expert-consensus product by a modest but statistically real amount over that period.

Position by position, the comparison is mixed. We're modestly ahead at quarterback, wide receiver, and tight end, and modestly behind at running back.

Another large-platform ADP comparison shows a statistically clean gap in our favor. But that platform also ranks substantially deeper into the player pool. Restrict the comparison to similar player-pool depth, and that difference becomes small enough that it could also be noise.

We could show only the flattering comparison.

We don't think that would tell you very much.

The full track record is free and requires no account.

You don't have to trust our description of the model. You can inspect how it has actually performed.

Try It on Your League

Credence paid subscriptions are live today.

But you don't need to subscribe to find out whether the product is useful to you.

Create an account, import a league, and see how Credence values players under your actual settings. Look at the uncertainty ranges. Compare the board with your own rankings. Read the public track record.

If that's all you need, keep using the free version.

If you want Credence helping you make decisions as your actual draft unfolds, that's what the paid subscription is for.

The first 100 paid members lock in a $10 monthly price permanently. After that, the standard price is $20 per month.

Your draft probably won't go exactly the way your rankings expect.

That's the part we're building Credence for.