DEVY WAR was built out of a personal obsession. There are other strong tools in the space — but this one lets you adjust your assumptions about draft capital and immediately see how those assumptions change a player's value. In your league. On your terms.
The goal is to measure value across completely different asset types — active NFL players, devy prospects, and draft picks — on the same scale, within the context of your league.
This is not "our" rankings. The model lets you pick the league, set the projected draft capital for college prospects, and choose which data to use for NFL players — then produces WAR values based on your assumptions. Two users with different views on where a player gets drafted will see different numbers. That's the point.
WAR measures how much production a player provides relative to a replacement-level option — defined by your league's roster requirements. In a shallow league, replacement is strong. In a deep or TE-premium format, replacement drops off quickly. The model adjusts automatically.
Each player's WAR is calculated from recent production, positional replacement levels, age decay curves, tier-stratified cumulative survival, and a discounted position-specific horizon. The output is a single number, but it is built from a full yearly breakdown under the hood — one you can inspect for any player.
Early-career players (first 1–3 NFL seasons) receive additional handling via Bayesian blending — their actual stats are blended with a historical trajectory prior derived from players at the same position and draft round. This prevents a talented rookie who missed games from being permanently undervalued. The blend fades entirely to actuals by year 4.
Devy players are evaluated on probabilistic outcomes tied to expected draft capital. The model uses empirical data from historical NFL draft classes to derive position-specific hit rates by pick slot — all four positions use pick number, since empirical analysis showed it has stronger correlation with career production than class rank at every position.
Those probabilities feed into an expected-value formula that accounts for both the upside (hitting and producing above replacement) and the downside (busting and contributing nothing). The result is not a prediction — it is an expected value estimate directly comparable to NFL players.
Hit rates vary significantly by position. QBs are the hardest to predict — even a top-5 pick has a meaningful bust rate, and QBs drafted outside the first round rarely produce. RBs and WRs have more stable hit patterns across the first two rounds. The model reflects these asymmetries rather than treating a top TE prospect and a top WR prospect as equivalent just because they're both consensus #1 at their position.
The model lets you override any prospect's projected draft pick — and trusts your projection if you do. But "trust the projection" assumes a meaningful projection exists. That holds for a college senior with three years of film and combine data. It holds less well for a high-school recruit who hasn't played a college snap. So a data confidence taper scales devy WAR by how observable the prospect's profile is — full strength for 2026 seniors, a small step down for 2027 juniors, more for 2028 sophomores, and roughly half-strength for 2029 freshmen and HS recruits. The magnitudes are anchored to public research on HS recruit → NFL draft hit rates (top-100 HS recruits reach the NFL at ~35-45% over 4-5 years vs ~70% for declared seniors with top-50 projections) and to year-over-year mock-to-actual-pick correlation (r≈0.80 in the draft year, r≈0.55 one year out, r≈0.35 two years out). The taper acknowledges that the further from real data you are, the less confidence the model should claim — without throwing the prospect's projection out entirely.
For WR and TE prospects with college stats available, a breakout age signal further adjusts their ceiling WAR. Players who produced early relative to their draft year historically outperform their pick slot — independently of draft capital. The adjustment is derived from regression analysis across historical college classes and applied automatically where data exists.
For TE prospects specifically, the model also looks at NFL Combine 40-yard time and peak college receiving production to discriminate receiving-archetype prospects from blocking archetypes. The two signals are independently statistically significant — historically, the slow-and-low-college-production cell hits at 5.6% while fast-and-high-college-production hits at 58.3%. Pick capital alone cannot tell those archetypes apart, so the model adjusts where the empirical signal exists. The TE multiplier only fires for drafted prospects with full combine + final-college-season data — pre-NFL devy underclassmen stay neutral so a freshman with a thin college stat line isn't penalized against a high-schooler with no reps yet.
WRs use a narrower version of the same idea. Combine metrics didn't discriminate WR hits from misses — the NFL has already pre-filtered for speed at the position. But R2–R3 WRs with strong college receiving production historically hit at materially higher rates than peers, so the model applies a small lift to that specific empirical cell. The adjustment is conservative and lift-only — never penalizes a prospect for unknowns or under-utilization in college.
RBs have their own version: it turns out neither combine speed nor college rushing volume predicts NFL fantasy success — but college receiving production does. Modern fantasy RBs need pass-catching ability, and the "3-down back" archetype hits at materially higher rates than pure rushers. R2–R3 RBs who demonstrated meaningful college receiving (peak ≥400 yards in a season) get a small lift in the model. Same lift-only philosophy: no penalty for "pure rusher" archetype, just acknowledgment that demonstrated pass-catching is empirical evidence for the 3-down role.
QBs are the most counterintuitive: college passing yards do not predict NFL fantasy success — but career college rushing yards strongly do. This is the Konami-code archetype that dynasty players have been talking about for years, with empirical backing. Hits averaged 2.4× more college rushing yards than misses. Mobile QBs (Lamar, Allen, Hurts, Daniels) dominate fantasy because rushing stacks 50–100+ extra points on top of passing each season. The model gives non–R1 QBs with strong college rushing a lift — late-round mobile QBs (Russell Wilson, Dak Prescott historically) are the archetype this captures. Pocket passers are not penalized through this lift — but they are checked against the empirical record (see next).
On top of the lift multipliers above, the model also runs an empirical cohort check. For every drafted devy prospect, it looks up the historical hit rate of their specific archetype — the combination of pick capital plus archetype signal (college production for WR/RB, college rushing for QB, athletic profile + college production for TE). If that archetype's empirical hit rate is below 10% historically, the model caps WAR at the empirical floor regardless of what pick capital alone would imply. The strongest example: R6–R7 pocket-passer QBs hit at 0 of 16 historically. Pick capital alone says "low but non-zero"; the empirical record says "this archetype virtually doesn't hit." The cap matches the empirical record. We don't fabricate evidence about cohorts the data hasn't seen — empty cells don't fire the cap. Together, the lift multipliers and the cohort check encode the full empirical structure: scale where signals discriminate, cap where archetypes confirm low.
The trade calculator lets you build any trade scenario — players, devy prospects, and draft picks on both sides — and see the WAR differential instantly. It also shows the KTC market value side-by-side so you can see where the model and the market agree, and where they diverge.
The calculator supports both L-WAR (Lifetime) and C-WAR (Contender) modes. Switch to C-WAR when evaluating trades from a win-now perspective — it strips out age decay and focuses purely on the next two seasons of production, making it easier to compare a veteran producing at a high level against a younger player with more long-term upside.
Draft picks are valued against your league's actual available board. A future first-round pick is worth the WAR of the Nth best available player at that slot — not a generic estimate. Future picks show you which players in that class are already rostered, so you can see exactly how much the board has been picked over.
For uneven trades, the calculator surfaces the best available free agents to help you understand what the extra roster spot is worth — making the comparison fair even when the player counts don't match.
For NFL players, you control which data feeds the model. The projection set picker offers several lenses: multi-year weighted actuals (the most stable signal), single-season actuals (reactive to what actually happened), and forward-looking projection sets as they become available. Switching sets recalculates all rankings and trade values instantly.
For devy prospects, the model starts with baseline pick projections — but you don't have to agree with these. In fact, this is the main area where most users disagree, and the tool can be useful for everyone. The model just gives you a starting point. You can (are expected to) edit any prospect's projected NFL draft slot, and WAR recalculates in real time across rankings, the draft board, and the trade calculator. Your overrides apply globally across all your leagues.
The model is built to honor your capital projection as the conviction. It does not second-guess your view by discounting for "the projection might be wrong." If you set a 2029 WR to top-5 capital, the model tells you what that player's WAR is if you're right — modulo only the small non-capital risks (injury, development) that exist regardless of where the prospect actually gets drafted. This is what makes the rankings genuinely yours.
This is especially important for future draft picks. The model uses current WAR estimates for upcoming draft classes, which will improve as those classes are scouted and mock drafts mature. Keeping your overrides updated is how you get the most accurate pick valuations today — before the consensus catches up.
Dynasty decisions are usually comparisons between unlike things — age vs upside, production vs projection, certainty vs probability. And they depend heavily on where your team is in its window.
This model does not remove uncertainty. It makes those trade-offs explicit and comparable. Is a 27-year-old RB with two great seasons left worth more than a devy WR who might go top-15 in two years? Most tools give you separate rankings for each. DEVY WAR puts them on the same axis.
For contending teams, L-WAR isn't always the right lens — an aging star who will produce at an elite level for two more years and then retire is worth far more to you than his lifetime WAR suggests. That's exactly what C-WAR is for. Switch modes and the rankings and trade calculator reflect your window, not a generic long-term horizon.
Combined mode surfaces the entire asset universe — active NFL players, college prospects, and draft picks — on one unified list. A high-WAR devy player appearing above a mid-tier veteran isn't an anomaly. It's the model telling you something about long-term value that a traditional ranking system can't.
This is a model, not an answer key. Outputs depend on scoring, roster construction, and assumptions about the future. The more you maintain your personal overrides, the more useful it becomes.
Values update as inputs and assumptions evolve.