Career Tiers: Star, Starter, Contributor, Bust
Data Dictionary · how NFL career outcomes are classified
Career tiers are how I sort finished NFL careers across the whole platform, four ways: Star, Starter, Contributor, and Bust. Every drafted prospect with a completed career window gets a tier based on where his value over replacement (VBD) lands within his position. These are the outcome labels the DMX model gets trained on and tested against.
The four tiers — and the base rates
Here's how careers actually shook out across every offensive prospect from the 2001 through 2020 draft classes with a finished career window (n = 2,026):
| Tier | Share | N | What it means |
|---|---|---|---|
| Star | 16.5% | 335 | Top-of-position difference-maker; league-winning stretches |
| Starter | 16.3% | 331 | Multi-year lineup piece with real positional value |
| Contributor | 20.8% | 422 | Flex/streamer seasons, short peaks, depth value |
| Bust | 46.3% | 938 | Negligible dynasty return on the pick |
Offensive prospects, draft classes 2001–2020, career tier labeled. Live counts from the public database.
Look at that last row, because it's the whole story: nearly half of all drafted skill guys bust, and only about one in three ever becomes a Star or long-term Starter. That's the backdrop for every rookie pick you'll ever make. The question isn't “will this guy hit?” It's “does anything about this profile beat a coin flip that's rigged 2-to-1 against me?” That's the whole reason to score prospects at all — to move you off that base rate. The top DMX decile hits at 80.2%; the bottom sits at 7.6% (full table in the DMX guide).
How tiers are assigned
There's no opinion in these tiers. I take a player's career VBD over a fixed window after the draft, rank it against his position, and let the percentile pick the tier. Nobody's voting. A guy two percent under the Starter line gets tagged Contributor even if his career “feels” like a Starter's. Yeah, that's a little arbitrary right at the boundaries, and I say so — which is exactly why I publish the raw VBD next to the label, so you can see how close he sat to the line.
Why the model predicts tiers, not points
Here's something the validation makes clear: predicting the tier works about twice as well as predicting exact fantasy points (tier R² of 0.289–0.415 across positions, versus 0.102–0.207 for exact VBD — all twelve regressions are in the whitepaper). Careers clump into types a lot more cleanly than they line up on exact point totals. Guessing which kind of career a guy gets is just a more learnable problem than nailing his point total to the decimal, and for dynasty, it's the more useful answer anyway.
Where you'll see tiers
The Draft Board puts each historical prospect's actual outcome tier right next to his DMX score — that's the model's track record, out in the open, class by class. The Analytics page uses the tiers for residual analysis: the Stars I scored too low and the Busts I scored too high sit on ranked leaderboards, not buried where you can't find them.
Frequently asked questions
What percentage of drafted NFL skill players bust?
In the 2001-2020 draft classes with finished career windows, 46.3% of offensive skill prospects (938 of 2,026) came out as Busts — careers with next to no value over replacement. Only about 33% turned into Stars or long-term Starters.
What counts as a Star career?
A career in the top band of value over replacement (VBD) for his position — roughly the top 13-14% of outcomes. The tiers come straight from VBD percentiles, not from anybody's opinion.
Are career tiers subjective?
No. They come from career VBD percentiles within each position. The boundaries are a little arbitrary — a guy just under a line gets the lower tier — which is why I publish the raw values right next to the labels.
Can a player's tier change?
A tier gets locked once the labeling window is finished, so it's stable. Active players don't have a final tier yet — for them, DPX tracks where their dynasty value is right now.