Draft board methodology
LeagueFrenzy ranks players for your custom scoring and roster shape: fantasy football projections from event rates, value over replacement at a league-specific baseline, rookie priors that keep hype in the market term, and Monte Carlo-style outcome ranges for boom-bust context. This page is the deeper methodology — written for readers who want the pipeline, not a pitch deck.
Prefer the shorter VBD primer first? Value-based drafting.
The value engine starts with expected counting stats — how often a player is involved, and how efficiently that involvement becomes fantasy events — then scores those event rates under your league rules so two identical skill lines can produce different points.
Each player carries expected counts for scoring events — receptions, rush yards, solo tackles, and the rest. Veterans with history get player-specific rates; rookies and thin samples use structured priors instead of copying elite veterans.
Those rates are scored against your league rules, not a default PPR sheet. Linear rules multiply rate by points per unit; tiered bonuses use a deterministic average-game approximation so a 100-yard bonus only pays when your league actually has one.
Projected points become value over replacement by subtracting a position-specific replacement level set from your roster slots, team count, and who is actually free. Rankings follow that surplus, then blend with market consensus so the board stays grounded.
Value over replacement is projected points minus the replacement level at that position. The replacement baseline is not a universal constant — it moves with roster construction, scoring rules, and free-agent quality. That is why custom fantasy rankings diverge from a one-size-fits-all sheet.
Replacement level at a position is roughly the Nth-best projected body, where N scales with team count times how many starters that position must fill. Dedicated slots count fully; flex and superflex slots spread across eligible positions by how much surplus those positions actually produce.
Add a second QB-eligible slot and the replacement QB gets much better, so elite QBs gain value over replacement. Pay real tackle scoring and startable linebackers exist that generic boards never rank. The baseline moves with your rules — so the whole board moves with it.
A starter is only worth the alternative you could actually pick up. Replacement is also informed by startable free agents, so a thin waivers position does not invent huge surplus from thin air. Synthetic public boards without real rosters stay on pure starter-count math.
Rookies are the easiest place for a model to launder hype into fake production. The value engine keeps draft-capital and college-production priors on the projection side, and market consensus on the blend side — then, on dynasty boards, applies development curves across future seasons.
Rookies with no NFL history do not inherit an elite-veteran projection just because market hype is loud. Offense rookies project from draft-capital bands; IDP rookies combine draft capital with college production so the value-over-replacement term stays independent of buzz.
Consensus ADP and market value still influence the final ranking blend — they just do not rewrite the projected event rates. A hyped first-rounder can still rank high; the engine is explicit about which half of the score is prior and which half is market.
On dynasty horizons, multi-year value over replacement applies an age-based development shape: young players ramp, peak by position, then decline. Year zero matches the current projection; missing ages degrade safely rather than inventing a curve.
A single projected total hides boom-bust risk. Monte Carlo methodology produces outcome ranges — floor, median, and ceiling bands in a p10 / p50 / p90 spirit — so you can see whether a ranking is a narrow volume bet or a wide distribution with real upside and real downside.
Season outcomes are noisy. Monte Carlo methodology treats a projection as a range — simulating usage, efficiency, and availability variance to produce floor / median / ceiling style bands (p10 through p90-style ladders under the hood).
A high ceiling with a soft floor is a boom-bust profile; a tight band around the median is a safer volume play. Those ranges are decision context, not a promise that every public board surface shows the full ladder to every user yet.
Reference distributions can be rescaled onto a league projection so the band brackets the number you see. Heavy per-game bonus tiers or unusual IDP weighting reshape tails, not just the level — per-league simulation is the honest long-term fix.
Trustworthy draft board methodology says where it stops. These limits are intentional product choices or known calibration gaps — not fine print after a hard sell.
The model deliberately keeps market consensus in its own term rather than letting it inflate event rates. That is a feature for custom-scoring leagues — and it means the engine will not chase every social narrative the way a pure ADP board would.
Breaking news, depth-chart churn, and injury reports that commercial sources fence behind licenses are not fully wired into every projection path. Availability filters remove clearly non-draftable statuses when we have them; they do not replace a human reading the injury report.
Tiered scoring still approximates every game as the average game outside of simulation work. IDP market signals are thinner than offense. Rookie IDP priors remain an earlier editorial pass than the fitted offense bands. We measure accuracy with backtests and publish methodology rather than hiding the seams.
It is not a claim of perfect future knowledge, not strength-of-schedule theater, and not a secret sauce dump of blend weights. It is a deterministic pipeline from your rules to ranked surplus — reproducible for the same inputs.
Want the shorter primer on value over replacement? Read value-based drafting. Browse all guides, or the blog.