Your draft is done. Week 1 is here. Before one box score convinces you the draft was a mistake, it is worth knowing what winning drafts actually looked like.
Every draft-strategy take you read in August has the same problem: nobody can check it. To test a strategy you need thousands of completed leagues with the full draft board and the eventual champion. We have both.
This post is what we found in 24,482 of them. It is an observation study, not a set of rules, and the caveats at the end are part of the finding.
How the study works
We sampled 12,500 Sleeper redraft leagues per season from the 2024 and 2025 seasons. To count, a league had to run a completed snake draft of 12 to 20 rounds, have 8 to 14 teams, seat at least four playoff teams, and have no best-ball or keeper settings.
That left 24,482 leagues and 264,406 teams. One champion per league, taken from the final match of the Sleeper winners bracket.
Each champion is measured against their own leaguemates. In a 12-team league every team starts with a 1-in-12 baseline chance of winning. A strategy’s “lift” is how much more often teams using it won versus that baseline. Because the comparison is inside the league, size and scoring cancel out.
Every number below clears a high statistical bar, and every table was recomputed on fully human-drafted teams to make sure autopick was not driving the result. It was not.
What champions did in both seasons
These are the findings that held in 2024 and again in 2025. That stability is why they carry the most weight.
| Draft pattern | 2024 title-odds lift | 2025 title-odds lift |
|---|---|---|
| Robust RB (3+ RBs in rounds 1–5) | +33.8% | +33.3% |
| Zero-RB (no RB in rounds 1–4) | −37.6% | −47.0% |
| Hero-RB (one RB in rounds 1–2, none in 3–5) | −10.3% | −4.4% |
| Late QB in 1QB leagues (first QB in round 9 or later) | +27.0% | +43.6% |
| Early kicker (round 11 or earlier) | −19.8% | −17.6% |
| Early defense (round 10 or earlier) | −25.9% | −23.6% |
| Heavily autodrafted (half or more of picks) | −21.4% | −25.8% |
Read the first two rows together. Teams that left the first five rounds with three or more running backs won about a third more often than their leaguemates, both years. Teams that took no running back in the first four rounds won far less often, and the gap widened in 2025.
In raw terms: in 2025, zero-RB teams won 4.9% of their leagues against a 9.3% baseline. Robust-RB teams won 12.3%.
The continuous numbers say the same thing more quietly. Champions drafted about 0.14 more running backs in rounds 1 through 5 than the field, drafted fewer total tight ends, and had roughly a quarter less autopick on their draft record.
What flipped between 2024 and 2025
The stable findings above are the load-bearing ones. These are the ones that moved, and they are the reason we publish per-season numbers instead of timeless rules.
| Draft pattern | 2024 | 2025 |
|---|---|---|
| Round-1 running back | −4.6% | +23.5% |
| Round-1 wide receiver | +6.1% | −15.9% |
| RB-RB start | +29.1% | +51.7% |
| WR-RB start | +25.5% | −0.9% |
| WR-WR start | −13.2% | −26.3% |
| Tight end in rounds 1–3 | −22.2% | +18.2% |
| Round-1 quarterback | −15.3% | −39.6% |
In 2024, champions won by opening wide receiver and scooping running back volume in rounds 2 through 5, the year mid-round backs badly outran their draft price. In 2025 the market had repriced elite backs upward, and paying up still won. Twelve-team champions opened with a running back 57.6% of the time against the field’s 44.9%.
The through-line is the same both years: champions concentrated early draft capital on running backs. Only the entry point moved.
The RB-RB start at +51.7% is the single largest result in the study, and it led every scoring format in 2025: standard, half-PPR and full PPR. It is not a standard-scoring artifact.
The tight end flip is a reminder about year effects. An early tight end was a leak in 2024 and an edge in 2025, because that year’s early tight end class delivered. Whether it repeats depends on the players, not the strategy.
Quarterbacks: the wait got longer
In 1QB leagues, taking your first quarterback in round 9 or later went from +27.0% in 2024 to +43.6% in 2025. Taking one in the first four rounds went from −13.3% to −25.5%.
Champions’ first quarterback came about 0.4 rounds later than the field’s in 2024 and about 0.9 rounds later in 2025.
Superflex did not rescue the early quarterback either. In 2025 superflex leagues, a first quarterback in rounds 1 through 4 ran −15%, while rounds 5 through 8 ran +45%. The round-9-and-later group looked even better but is a small sample, so treat it as a direction rather than a number.
Two openers stood out as the worst in the study: two quarterbacks in the first two rounds (−25% in 2024, −56% in 2025) and a quarterback plus tight end start (−69% and −59%).
What did not matter: your draft slot
Draft slot produced a clean-looking pattern each year and a different one each time. In 2024 the champion rate rose steadily from slot 1 to slot 12. In 2025 it was U-shaped, with slots 1 and 2 on top and slot 5 at the bottom.
That is the signature of which specific players sat at each slot in a given year, not of slot value. The only durable read is that strategy spreads of plus or minus 50% dwarf anything the slot did. You can win from anywhere.
The caveat that matters
This is correlation, not causation. Managers who draft this way also tend to manage rosters better in-season: they work the waiver wire, they set lineups, they show up. The autopick row is the clearest hint of that. Some of the lift belongs to the manager, not the strategy.
So the honest framing is “what champions did,” not “do this and you will win.” We are not claiming these patterns make anyone’s advice sharper. They are a description of winning rosters, measured carefully.
What to check on your roster before Week 1
You cannot redraft. You can look at what you have with the same lens.
- Count your running backs from rounds 1 through 5. Three or more is the shape champions had both years. One is the hero-RB shape, which ran mildly negative. Zero is the shape that won least.
- Find where your first quarterback went. If it was round 4 or earlier in a 1QB league, you paid a price that champions mostly did not pay.
- Check whether autopick drafted for you. Heavily autodrafted teams won about a quarter less often, and that is the one row that says more about the manager than the roster.
- Look at your kicker and defense rounds. Champions took both later than the field did.
None of that decides Week 1. It tells you which parts of your roster are already in the champion pattern and which parts you will be managing around all season.
Grade your draft against the pattern
If you want the count done for you, the free draft grade reads your latest Sleeper draft and scores every pick, and the free team grade does the same for the roster you have now. Neither needs a signup.
When you connect your league, the co-manager already knows your roster, your scoring, and every team in your league. Ask it whether your running back room is the champion shape or the hero shape, and it answers about your team, not a generic one.
Start free and ask about your roster
Data note: every number in this post comes from the SharperSunday Championship Draft Study, run 2026-07-11 on the 2024 and 2025 seasons (repo artifacts docs/CHAMPIONSHIP_DRAFT_STUDY_2026-07.md and docs/CHAMPIONSHIP_DRAFT_STUDY_2026-07.data.json). Sample: 24,482 Sleeper redraft snake leagues, 8 to 14 teams, no best-ball, no keepers, 264,406 teams. Lifts are within-league champion rates versus a 1-in-N baseline. Every headline figure clears |z| ≥ 4.5; four cells shown for continuity do not (2024 round-1 running back at −4.6%, 2024 two-quarterback start at −25%, 2025 hero-RB at −4.4%, and the 2025 superflex round-9-plus quarterback group), and none of them is treated as a finding on its own. The study is redraft-only and does not extrapolate to dynasty. Separately, SharperSunday’s corpus registry had identified 104,612 league champions as of 2026-09-07; that is a count of champions found, not the study sample. Player positions come from the Sleeper player database at analysis time. Findings are correlational and were controlled for autopick.