Does team-building strategy actually matter, and can we predict which teams make the playoffs? We trained a playoff-prediction model on every historical draft pick, and clustered 154,636 fantasy rosters across four scoring formats to find the archetypes that consistently make the playoffs.
Each scoring format trains its own XGBoost classifier on every historical draft pick, predicting whether the drafting team went on to make the playoffs. Stats below are per-model, not per-cluster.
XGBoost classifier · trained 2026-07-27 19:45 · 54 features
| Leagues | 10,230 | Teams | 115,319 |
| Training picks | 812,412 | Test picks | 202,857 |
| Total picks | 1,015,269 | CV search iterations | 300 |
XGBoost classifier · trained 2026-07-27 18:55 · 54 features
| Leagues | 1,342 | Teams | 15,284 |
| Training picks | 114,799 | Test picks | 28,922 |
| Total picks | 143,721 | CV search iterations | 300 |
XGBoost classifier · trained 2026-07-27 19:03 · 54 features
| Leagues | 1,330 | Teams | 14,813 |
| Training picks | 119,923 | Test picks | 29,570 |
| Total picks | 149,493 | CV search iterations | 300 |
XGBoost classifier · trained 2026-07-27 18:32 · 54 features
| Leagues | 363 | Teams | 4,038 |
| Training picks | 34,745 | Test picks | 8,697 |
| Total picks | 43,442 | CV search iterations | 300 |
Which roster-building strategy correlates with making the playoffs, per format.
118,886 teams analyzed in this format
| # | Archetype | Playoff % | Champ % | Teams | QB% | RB% | WR% | TE% |
|---|---|---|---|---|---|---|---|---|
| 1 | Balanced | 60.7% | 10.7% | 25.9% | 14.9% | 32.2% | 30.5% | 8.6% |
| 2 | RB Dominant | 60.6% | 11.1% | 14.4% | 15.4% | 38.7% | 22.8% | 9.0% |
| 3 | WR Dominant | 56.5% | 9.2% | 28.2% | 14.8% | 25.6% | 37.3% | 8.7% |
| 4 | WR Dominant (Extreme) | 45.1% | 6.7% | 15.2% | 14.7% | 18.9% | 43.8% | 8.9% |
| 5 | WR Leaning / TE Heavy | 42.9% | 5.2% | 16.3% | 17.8% | 24.5% | 29.7% | 12.2% |
15,830 teams analyzed in this format
| # | Archetype | Playoff % | Champ % | Teams | QB% | RB% | WR% | TE% |
|---|---|---|---|---|---|---|---|---|
| 1 | RB Leaning | 61.2% | 10.8% | 18.4% | 14.2% | 36.1% | 28.4% | 8.1% |
| 2 | WR Dominant | 61.1% | 11.6% | 27.0% | 13.7% | 28.4% | 37.3% | 7.9% |
| 3 | Balanced | 50.5% | 7.0% | 20.7% | 16.2% | 28.5% | 30.1% | 10.2% |
| 4 | WR Dominant (Extreme) | 49.6% | 7.7% | 16.7% | 13.9% | 20.0% | 45.0% | 8.2% |
| 5 | WR Dominant / TE Heavy | 43.3% | 5.2% | 17.2% | 16.3% | 21.4% | 36.1% | 11.5% |
15,452 teams analyzed in this format
| # | Archetype | Playoff % | Champ % | Teams | QB% | RB% | WR% | TE% |
|---|---|---|---|---|---|---|---|---|
| 1 | RB Dominant | 61.6% | 11.8% | 13.4% | 20.7% | 36.6% | 22.0% | 8.3% |
| 2 | Balanced / High QB | 58.0% | 8.8% | 24.4% | 26.3% | 26.8% | 24.7% | 9.8% |
| 3 | WR Leaning / Low QB | 53.7% | 9.2% | 21.1% | 18.2% | 27.8% | 31.8% | 8.7% |
| 4 | WR Dominant / High QB | 53.7% | 8.1% | 27.1% | 25.4% | 20.5% | 32.8% | 9.1% |
| 5 | WR Dominant (Extreme) / Low QB | 45.5% | 7.5% | 14.1% | 18.5% | 18.8% | 41.9% | 8.4% |
4,468 teams analyzed in this format
| # | Archetype | Playoff % | Champ % | Teams | QB% | RB% | WR% | TE% |
|---|---|---|---|---|---|---|---|---|
| 1 | RB Dominant | 61.0% | 11.9% | 13.9% | 20.4% | 34.5% | 25.1% | 7.8% |
| 2 | Balanced / High QB | 57.6% | 10.0% | 28.6% | 25.5% | 25.9% | 27.7% | 8.9% |
| 3 | WR Leaning / Low QB | 53.3% | 8.4% | 19.8% | 16.6% | 27.5% | 33.8% | 8.7% |
| 4 | WR Dominant / High QB | 51.4% | 7.4% | 26.4% | 24.0% | 20.2% | 35.2% | 9.0% |
| 5 | WR Dominant (Extreme) / Low QB | 42.9% | 6.6% | 11.3% | 17.4% | 17.8% | 43.6% | 8.6% |
Patterns consistent across all four scoring formats
In every format, heavy investment in QB and TE scoring share ranks last or near-last — often 20+ points below the top archetype's playoff rate.
RB-heavy rosters lead the standings in standard and 3WR formats. The Extreme WR strategy underperforms in every format tested.
Superflex leagues reward QB investment, but not to the extent that you would think. The QB + RB balanced build becomes the optimal path to the playoffs.
Team data was collected from the Sleeper API covering the 2025 fantasy season using a snowball-sampling approach starting from seed users. Only completed redraft leagues with 10–16 teams were included, and only those whose active roster slots matched one of the four supported formats exactly. Teams with missing scoring data were removed before analysis, leaving 154,636 teams across the four formats.
For each team, we tracked which players started each week and computed that position’s share of total season fantasy points (QB%, RB%, WR%, TE%, DEF%, K%). Flex and SuperFlex slots were resolved to the actual position the player filled that week — an RB starting in a FLEX slot contributes to RB scoring, not a separate flex bucket.
We then applied k-means clustering with k=5 — selected via elbow method and silhouette analysis — to group teams into five distinct archetypes per format based solely on their positional scoring mix. Each format is clustered independently because optimal roster construction differs meaningfully across standard, 3WR, and superflex league types.