Can a fantasy football novice automate his way to championship glory?
This week
highest-value moveTeam & system
consensus (R) · priority #1Optimal lineup
proj 110| Slot | Player | Tm | Proj |
|---|---|---|---|
| QB | Jalen Hurts | PHI | 17.1 |
| RB1 | Derrick Henry | BAL | 18.9 |
| RB2 | Aaron Jones Sr. | MIN | 16.8 |
| WR1 | DeVonta Smith | PHI | 11.2 |
| WR2 | Carnell Tate | TEN | 8.6 |
| TE | Juwan Johnson | NO | 8.6 |
| W/R/T | Jeremiyah Love | ARI | 13.7 |
| K | Trey Smack | GB | stream |
| DEF | Saints | NO | stream |
System health
consensus (R)League
championship odds · all 12| Team | Title odds | Title | Playoff | |
|---|---|---|---|---|
| 1 | Let’s Get Weird! | 16.4% | 56% | |
| 2 | Jay Money | 14.7% | 50% | |
| 3 | When Life Gives U Mikes… | 13.8% | 48% | |
| 4 | Say Hello To My Kittle Friend | 11.8% | 45% | |
| 5 | When Life Gives U Makai Lemon | 9.2% | 38% | |
| 6 | Lone Male Juror | 9.0% | 38% | |
| 7 | Bijan’s Bitches | 7.9% | 35% | |
| 8 | Bench Warmers Anonymous | 7.4% | 33% | |
| 9 | HAIL | 4.6% | 24% | |
| 10 | Real Rout Daniel | 4.2% | 21% | |
| 11 | Saquon Deez Nuts | 0.8% | 8% | |
| 12 | Them Scousers | 0.3% | 3% |
How it works
data → brain → youData sources
The brain
What you get
Opportunity metrics
xFP · WOPR · snap share → buy-low / sell-high.
Correlated sim
League + stack correlation with real right-skew.
Weighted ensemble
Blends sources by graded accuracy, weekly.
Trade finder
Scans every roster for fair, +ΔP swaps.
The evidence
research this is built on · gradedEvery strategy below was researched and graded before it was coded. SUPPORTED = backed by data; JUDGMENT = directionally sound but effect size unproven, so it’s implemented cautiously. Click any card for how it works, why it matters, and the source.
SUPPORTEDBlended projections beat any single sourceThe coach averages multiple sources instead of trusting one, then reweights by accuracy.
How: a weighted ensemble (consensus + ESPN), re-scored to your exact rules and reweighted each week toward whichever source has been most accurate for THIS league.
Value: in a 12-season backtest, the consensus average beat individual sources in 69% of head-to-head comparisons; single sources are volatile year to year.
Source: Fantasy Football Analytics
SUPPORTEDUsage predicts scoring, not recent pointsOpportunity (targets, snaps, air yards) is weighed far above last week’s fantasy total.
How: target share correlates ~0.70 year-over-year and explains ~60–70% of WR per-game variance; it stabilizes after ~4 games, and snap share stabilizes fastest. Efficiency stats (yards-per-carry) have almost no predictive value.
Value: stops the coach chasing a one-week spike (mostly noise) and buys players whose ROLE, not their luck, is rising.
Source: SumerSports research
SUPPORTEDStream kickers & defenses; never roster backupsA fresh K and DEF each week by matchup, not a season-long hold.
How: of top-10 defenses through Week 12 (2022–24), only 40% stayed top-10 over the final six weeks. Kickers are nearly flat, 20 of 28 bunched between 7.6–9.6 points per game.
Value: the bench spot a backup K/DEF would waste becomes a weekly matchup edge instead.
Source: ESPN streaming data (Cockcroft / Clay).
SUPPORTEDVegas implied totals predict real scoringBetting markets set each team’s expected points, the coach reads them for streaming.
How: implied total = (game total ÷ 2) − (spread ÷ 2). Historically teams implied under 14 scored ~13.7; 24–27.75 scored ~25.9; 28+ scored ~29.7.
Value: targets defenses facing offenses implied ≤ ~18–19 (the scoring-cliff zone) and kickers on high-total offenses.
Source: The Odds API
SUPPORTEDYour league’s scoring quirks are exploitableProjections are re-scored to your exact rules, and the coach targets the edges they create.
4-pt passing TDs flatten QB value (a 6-pt format spreads QB1–QB9 ~36% vs ~29%), making streaming QB viable and rushing QBs more valuable.
Kicker distance tiers (40–49 = 4, 50+ = 5) reward leg strength, Brandon Aubrey has averaged 10.44 pts, +2.27/game (+27.8%) above average.
The DEF cliff (21–27 allowed = ZERO, 28+ = negative) makes matchup selection matter more than unit quality.
JUDGMENTFavorite plays the floor; underdog plays the ceilingFavored → lean safe. Underdog → lean high-variance. Used only as a tie-breaker.
How: grounded in Skinner’s variance-optimization framework: maximize Z = (μ_you − μ_opp) / √(σ²_you + σ²_opp). The crossover is exactly 50% win probability when adding variance is free.
Value: a real but modest edge, so it’s implemented conservatively (never sacrificing >10–15% of projected points); the head-to-head effect size is unpublished.
Source: Skinner, “Scoring strategies for the underdog,” Journal of Quantitative Analysis in Sports (2011).
JUDGMENTHoard #1 waiver priority; spend it only on a needle-moverYour rolling #1 priority is a one-time asset, the coach guards it.
How: winning a claim drops you to last and never replenishes, so the coach only spends #1 above ~1% championship gain; below that it says wait and grab the player off free agency for zero cost.
Value: avoids the classic mistake of burning your best claim on a two-week injury fill-in.
Data sources
everything it reads, all freeEight-plus live feeds are fused into one player model every week.
Yahoo FantasyYour live roster, the waiver wire, matchups, and injury tags.
Cadence: read continuously
Source: via your authenticated session
Sleeper APIInjury status, depth charts, and trending adds/drops (crowd demand).
Cadence: daily
Source: docs.sleeper.com
nflversePlay-by-play, snap counts, target share, air yards, red-zone data.
Cadence: after every game
Source: nflverse.nflverse.com
ffopportunityExpected fantasy points (xFP) from an xgboost model on nflverse.
Cadence: weekly, in-season
Source: ffopportunity.ffverse.com
Consensus projectionsMulti-source projection aggregation (ffanalytics).
Cadence: Tue & Sat
Source: fantasyfootballanalytics.net
The Odds APISpreads & totals → implied team totals for streaming.
Cadence: Sunday AM
Source: the-odds-api.com
NFL inactivesOfficial actives, posted ~90 min before kickoff.
Cadence: polled Sunday (most time-critical)
Source: www.nfl.com/inactives
FantasyCalcCrowd-sourced trade values from real trades.
Cadence: on demand
Source: fantasycalc.com
The mathematics
under the hoodThe engine is deterministic statistics, not vibes. Here’s the machinery.
SIM10,000 simulated seasons per decisionA Monte-Carlo season run to the reseeded top-4 bracket, ten thousand times.
Each week every team draws a score; the real Yahoo schedule decides wins; the top 4 seed; Weeks 16–17 are a reseeded bracket where the higher seed wins ties. ≥10,000 seasons per decision for stable odds.
STATSGamma distributions & per-position volatilityFantasy scores aren’t a bell curve, they’re right-skewed and floored at zero.
So scores are modeled as gamma, not normal. Per-position weekly coefficient of variation: QB 0.36 · RB 0.54 · WR 0.58 · TE 0.63. TEs and deep WRs are the most volatile; QBs the steadiest, which is exactly why streaming a QB is safer here than streaming a TE.
MODELCorrelation: the stack ceilingTeammates boom together, and whole weeks run hot or cold. Both are modeled.
A shared latent weekly factor makes all teams co-move slightly (~8% league-wide), and same-NFL-team starters correlate more (+0.2 to +0.6), the “stack” ceiling that wins two-week playoff brackets.
MODELAvailability: why depth mattersEach starter can miss a week; if he does, your best backup fills in.
A deep roster loses less when a starter sits, because the sim swaps in your best backup and prices in the uncertainty. That is what finally makes handcuffs and depth pay off.
MODELWeighted ensemble + a learning loopSources are blended by accuracy, and the coach grades itself every week.
Every projection is logged and later scored against the real result; the mean-absolute-error per source becomes its weight, so the blend keeps shifting toward whatever is most accurate for your scoring.
METRICWOPR: weighted opportunityOne number for how much a receiver is really being used.
Combines target share and air-yards share into a single opportunity score. In one study it explained about 75% of receiving fantasy points (R² ≈ 0.75).
METRICΔP(championship)Every possible move is scored by how much it raises your title odds, nothing else.
This targets title probability directly, so a move that helps you win the whole thing outranks one that just scores more points on paper. Computed with common random numbers so even tiny edges are measured precisely, then every waiver, start/sit, and trade is ranked by this single delta.