Every season, fantasy managers build their rosters around projections from sources like CBS, ESPN, NumberFire, FantasyPros, and others. But which projections should you trust? Are some sources consistently better than others, or does accuracy swing so much that last year’s best source becomes this year’s worst? To find out, we analyzed seasonal projection accuracy across eleven sources from 2014 to 2025, covering the top projected players at all four offensive positions over 12 seasons.
We measured three things: accuracy using Mean Absolute Error (MAE), how much of the variance in player performance projections explain using R², and directional bias using Mean Error (ME). The full analysis, which includes detailed source-by-source rankings, R² tables, calibration plots, and bias charts, is available in our complete seasonal accuracy analysis. What follows are the key findings every fantasy manager should know.
Accuracy by Position
Quarterback Projections
Quarterback has become the hardest position to project, with error rates climbing sharply across every source as rushing production and passing-game volatility have reshaped the position. FantasyPros led QB accuracy over the full 12 seasons (61.0 MAE), with FFA Average right behind (61.7 MAE), and both aggregation approaches beat every individual source. In the last three seasons, NFL emerged as the most accurate source (71.2 MAE), while ESPN, strong early in the sample, fell to last (86.7 MAE).
Running Back Projections
CBS led RB accuracy over the full historical period (52.2 MAE), narrowly ahead of FantasyPros (52.4 MAE), and held its lead in recent seasons (54.2 MAE). FFA Average and FFA Weighted finished in a tie over the last three seasons (56.9 MAE each), staying in the top five without the sharp year-to-year swings other sources show. RTSports lagged the others, finishing last among qualifying sources (64.3 MAE recent).
Wide Receiver Projections
Wide receiver is the most tightly clustered position, with the top seven sources separated by just 2.1 MAE historically. FantasyPros was the historical leader (40.2 MAE), while FFToday took over in recent seasons (40.9 MAE), with FFA Average and FFA Weighted close behind in fourth and fifth (43.8 and 43.9 MAE). CBS illustrated how volatile a single source can be, dropping from fourth historically to ninth in the last three years (47.5 MAE).
Tight End Projections
FantasyPros led TE accuracy historically (31.4 MAE), edging FFToday (31.5 MAE), with FFA Average sixth (32.3 MAE). The position improved across the board recently, with FFToday holding a narrow lead (29.4 MAE) and FantasyPros, ESPN, and NFL all within a single point. FFA Weighted and FFA Average tracked just behind (30.4 and 30.5 MAE), while CBS, FantasySharks, and RTSports rounded out the bottom.
Variance Explained (R²)
Even the best seasonal projections explain only a limited share of the variation among top players, roughly 14% to 26% of within-position variance depending on position, with running backs the most predictable and quarterbacks the least. This is lower than the figure from our original 2015 analysis, but that is largely because restricting the sample to the top projected players compresses the range and naturally lowers R². Over the last three seasons, FFA Average posted a higher R² than FantasyPros at three of four positions. FFA projections rank among the top sources for R² at most positions, reinforcing why minimizing error through aggregation matters so much when a substantial share of fantasy outcomes remains unpredictable.
Calibration: How Projections Exaggerate the Spread
The most useful structural finding in our analysis is how projections miss. At every position, projections exhibit an exaggerated spread. This means the highest-projected players tend to underperform their projections, while the lowest-projected players meet or slightly exceed theirs. This shows up as a calibration slope below 1.0 at every position, steepest at QB (0.67) and TE (0.72), with WR the best-calibrated (0.85) and RB in between (0.79). The practical implication for your draft is simple. The gap between a source’s top-projected player and its fifth-projected player is almost always smaller in reality than on paper, which means the perceived drop-off from RB3 to RB8 is overstated, and late-round starters carry more upside than their projections suggest.
Bias by Position
Unlike weekly DFS projections, which are nearly balanced directionally, seasonal projections are systematically too optimistic. The average ME across all sources and positions was +21.6 points, meaning the top projected players underperformed their preseason projections, an effect driven by the downside risk of elite players falling short. The bias is most dramatic at quarterback, where it has nearly doubled from +22.5 points historically to +46.5 in recent seasons as the position has grown more volatile. FFToday was consistently the most conservative source across positions, while RTSports over projected the most. For how these seasonal biases compare with the small, balanced biases in weekly projections, see our companion DFS analysis.
Aggregating Sources Works
The single most robust finding across twelve seasons is that aggregation works. FFA Average outperformed individual sources in 69% of head-to-head comparisons, finishing with an average rank of 4.2 out of 11 sources, and the two aggregation-based approaches, FFA and FantasyPros, consistently beat every individual model. The reason is the wisdom of the crowd, combining sources smooths out the idiosyncratic errors that cause individual sources like CBS and ESPN to swing from first to last from one season to the next.
The weighted and simple averages are essentially interchangeable. FFA Average (47.4 MAE) edged FFA Weighted (47.8 MAE) over the full sample, winning 64% of head-to-head comparisons. Source accuracy simply does not persist reliably enough from year to year for historical weighting to provide a consistent edge, so equal weighting is the simpler and safer bet, the same conclusion we reached in our weekly DFS analysis.
The Bottom Line
In any given season, you can expect FFA projections to perform at or near the top for every position, without the catastrophic year-to-year drops that plague individual sources. FFA Average ranked in the top six at every position historically, and no individual source matches that cross-positional reliability: FFToday is strong at WR and TE but weaker at QB, CBS leads at RB but has fallen off elsewhere, and NumberFire is solid at TE but struggles at QB and RB. Layered on top of that reliability is the analytical depth to act on it, a weighted average, manual source weighting, and league-specific scoring in the web app. For the full positional breakdowns, season-by-season rankings, R² tables, calibration plots, and bias charts, check out the complete analysis, and to evaluate accuracy for yourself, head to the web app.
Stay Connected with Fantasy Football Analytics
For more data-driven fantasy football analysis and projections, follow us on social media:
- X (Twitter): @FFAnalyticsNet
- Facebook: FFAnalytics
- Instagram: @fantasyfootballanalytics
Visit FantasyFootballAnalytics.net for our latest tools and articles.
The post We Analyzed 12 Seasons of Fantasy Football Projections. Here’s What We Found. appeared first on Fantasy Football Analytics.