This piece first ran with a $107M surplus lead and a flat null that killed the owner’s thesis. Both were artefacts. A data seam in 2025 credits contracts with player-weeks nobody rostered, and on the clean window West Waco lead and we are second. The null was mis-specified, and once the baseline is held the flywheel does turn — but on retention and extension only. The half of the thesis about going out and landing new stars is the half with nothing behind it.
Correction — what the first version of this article got wrong
Claim one, withdrawn. The first version led with “first by $107M, and none of it made at the top.” Six-season surplus does read Scranton $370.5M and West Waco $263.2M, a gap of $107.3M, and those figures still reproduce. But they are contaminated. 2025 is the only season in which the store carries unrostered player-weeks, and surplus is summed with no week filter, so 2025 credits contracts with points scored in weeks the franchise did not own the player. On the seam-free 2020–2024 window the order reverses: West Waco Wildcatters $262.5M, Scranton Stranglers second at $233.8M. The within-season z-score control that piece ran does not rescue the claim.
Claim two, reversed. The first version reported −0.005 for “last year’s surplus against the change in elite payroll” and concluded the flywheel does not turn. That test was mis-specified — it regressed a change on a predictor correlated with the baseline, without holding the baseline, which is biased downward by construction. Specified properly it reads +0.331. The leg this article declared dead is alive, and the owner’s thesis is supported rather than refuted on that leg.
Two further claims are weakened rather than withdrawn and are corrected in place below: the split-half table was reported without a null bar, and the head-to-head-wins check validates far less than it was made to carry. Nothing has been quietly edited — every figure that moved is named here or in the next section.
Two live threats to anything measured here, and two standing properties of the metric. The threats are stated first because they are what broke the first version of this piece, and a table this clean is exactly the kind that gets read past its own definition.
Threat one — the 2025 data seam. build-realized-value.js sums fantasy.player.points grouped by player and season, with no week filter and no roster-week filter. For 2020–2024 the store holds rostered player-weeks only, so that is harmless. For 2025 it holds 8,665 additional player-weeks that nobody rostered, and every one of them gets credited to whoever held the contract at snapshot time. The tell is mechanical: mean weeks per contract-season runs 8.02, 8.74, 8.90, 9.12 and 8.65 across the first five seasons, then 10.21 in 2025.
| Season | Contract- seasons | Mean weeks per row | Mean points per row |
|---|---|---|---|
| 2020 | 689 | 8.02 | 72.8 |
| 2021 | 685 | 8.73 | 77.0 |
| 2022 | 680 | 8.90 | 76.3 |
| 2023 | 667 | 9.12 | 77.6 |
| 2024 | 690 | 8.65 | 74.6 |
| 2025 | 675 | 10.21 | 81.4 |
It is not spread evenly, which is what makes it fatal rather than merely noisy. Across the sixteen franchises, the excess weeks a franchise picks up in 2025 correlate with its 2025 surplus at 0.773 (Spearman 0.782). Scranton is one of the two biggest beneficiaries — 10.89 weeks a row against an 8.44 baseline, +2.44 — on a 2025 surplus of $136.7M, which is 36.9% of the entire six-season total. South Park picks up +2.34 and books $170.3M. Las Vegas moves the other way, −0.67, and books −$111.2M. Season-demeaned, weeks-per-row against surplus runs r = 0.667 inside 2025 against r = 0.103 across 2020–2024. The 2025 column is measuring roster churn as skill.
Threat two — the change test was mis-specified. Correlating prior surplus against the change in elite payroll without holding the baseline level constant induces a mechanical negative bias, and both ingredients are present and large. Baseline elite payroll mean-reverts: level against its own change correlates at −0.465. Prior surplus correlates with that baseline at 0.539. Their product, −0.251, is a downward pull sitting inside the published statistic before any real effect gets a chance to appear. The full treatment is two sections down.
Standing property one — surplus is zero-sum by construction. value.realized converts points over replacement into dollars using an exchange rate recalibrated every season as that season’s total committed salary divided by that season’s total weighted production. League-wide surplus sums to $0.000M in every year measured — not approximately, exactly. Any “first by” claim is a claim about the other fifteen franchises, never a claim that value was created out of nothing.
Standing property two — the downside is floored. Production over replacement is clipped at zero, so a player who finishes below replacement level scores a surplus of exactly minus his salary and never worse. That makes the bottom of the table a payroll-weighted count of contracts that returned nothing rather than a measure of harm done, and it is the single biggest reason the floor teams sit at three hundred million negative.
A seam in one season of six, worth 36.9% of our own total, and a specification error worth −0.251 of correlation. Neither was visible in the published tables; both were visible in the store’s own catalogue and in the shape of the regression. Not one of the piece’s five robustness checks tested either.
The elite band still does not pay — 0.889 on the dollar, and 0.871 with the seam season dropped. The payroll finding is built from the salary ledger, which no phantom player-week can touch. And on the one seam-proof valuation column in this piece, the price leg, Scranton is first of sixteen at $152.5M.
The leg the first version called dead, re-run against the objection it never faced. Three numbers below: what was published, what was wrong with it, and what the question actually answers to.
The published statistic was the correlation between a franchise’s surplus in season N and the change in its elite payroll from N to N+1, across 80 franchise-season pairs. It came back −0.005 with a bootstrap interval straddling zero, and the piece called it a flat null and built its conclusion on it.
It is a sum, not a measurement. A franchise carrying a large elite payroll this year is a franchise likely to shed elite payroll next year — that is regression, and it measures −0.465. That same franchise is also more likely to have generated surplus, at 0.539. Multiply those and you get a bias of −0.251 baked into the statistic, pointing the wrong way, before the effect under test has said anything. Hold the baseline constant and the effect is there.
negative bars run right to left
the published null was the true effect and the mechanical bias, cancelling
| Test | Statistic | Reads as |
|---|---|---|
| surplus in N → elite payroll in N+1, level | +0.310 | coupled |
| surplus in N → elite payroll in N, same season | +0.369 | coupled |
| elite payroll in N → surplus in N+1, level | −0.046 | nothing |
| surplus in N → change in elite payroll, as published | −0.005 | mis-specified |
| baseline elite payroll vs its own change | −0.465 | mean reversion |
| prior surplus vs that baseline | 0.539 | the other half |
| product — the mechanical bias | −0.251 | the whole null |
| surplus in N → change, baseline held | +0.331 | the flywheel |
| same, within franchise | 0.338 | holds |
| in elite seats, not dollars: raw / partial | −0.012 / 0.354 | same shape |
| surplus in N → newly acquired elite dollars | −0.015 | nothing |
The partial correlation is +0.331, with a 95% interval of 0.173 to 0.482 and 0 of 10,000 franchise-block bootstrap draws at or below zero. Within franchise it is 0.338. The OLS slope with the baseline held is 0.215 — about 21 cents of elite payroll for every dollar of surplus generated the year before. And it is not an artefact of the dollar denominator: ask the same question in elite seats and the raw statistic is −0.012 while the partial is 0.354, the identical pattern.
Now the part that is genuinely new, and it is sharper than either the old conclusion or a simple “thesis confirmed”. The elite payroll a franchise adds can come from two places: keeping and paying up for a player it already had, or going out and signing one it did not. Those separate cleanly, because the second needs no baseline control at all — a newly acquired elite contract-season is one at outside-market rank 1–50 for a player the franchise did not hold the season before. It is not a rare event: the league averages $27.9M of newly acquired elite salary per franchise-season. Against prior surplus, within franchise, it correlates at −0.015.
So the flywheel turns on one bearing and freewheels on the other. Surplus becomes elite payroll through retention and extension, and not through acquisition. The owner’s thesis was “selecting good players that you can get at good contract prices, freeing up salary to spend to go out and land other studs.” The freeing-up is real, and the spending is real at 21 cents on the dollar. The going out is the step that is not in six seasons of data. Substitute the seam-proof price leg for surplus and the acquisition side improves to 0.162 — better than nothing, still the weaker of the two channels by a distance, and still not the mechanism the sentence describes.
Realised surplus per franchise, 2020–2025: what each contract returned in league-rate dollars, minus what it cost. This is the table the first version led on. It reproduces to the decimal and it is the one most exposed to the seam, so it is printed with its correction attached rather than alone.
| Franchise | Surplus | Salary paid | Return multiple |
|---|---|---|---|
| Scranton Stranglers | $370.5M | $1,203.2M | 1.308 |
| West Waco Wildcatters | $263.2M | $1,171.3M | 1.225 |
| South Park Cows | $221.1M | $986.3M | 1.224 |
| Willamette Valley Coopers | $115.8M | $1,132.9M | 1.102 |
| San Antonio Stetsons | $115.1M | $1,094.4M | 1.105 |
| Hollywood Wookies | $109.2M | $917.9M | 1.119 |
| Enniscorthy Electric Eels | $64.8M | $1,065.7M | 1.061 |
| Washington Redskin Potatoes | $59.1M | $969.3M | 1.061 |
| Carbon County Coal Crushers | −$3.9M | $999.0M | 0.996 |
| Oeiras Silver Swans | −$17.5M | $1,075.8M | 0.984 |
| Houston Longhorns | −$53.4M | $1,039.5M | 0.949 |
| Chicago Maesters of the Midway | −$56.5M | $1,007.1M | 0.944 |
| South Philly Pigeon Boys | −$101.8M | $857.9M | 0.881 |
| Kawasaki Samurai | −$342.4M | $1,007.7M | 0.660 |
| Las Vegas Gamblers | −$368.8M | $944.8M | 0.610 |
| San Diego Hopheads | −$374.7M | $1,071.6M | 0.650 |
Cut the seam season out and the top of it changes hands. If a lead is a property of a franchise it should not depend on whether the contaminated season is in the sample. This one does, and it does not survive either shortening.
| Window | Leader | Leader’s surplus | Scranton | Gap |
|---|---|---|---|---|
| 2020–2025 — with the seam | Scranton Stranglers | $370.5M | 1st | — |
| 2020–2024 — seam-free | West Waco Wildcatters | $262.5M | 2nd, $233.8M | −$28.7M |
| 2020–2023 | West Waco Wildcatters | $225.6M | 3rd, $144.3M | −$81.3M |
The dispersion control the first version ran does not fix this, and it is worth saying why it looked like it would. Franchise surplus has a standard deviation of $51.3M in 2020 and $76.3M in 2025, and 71.8% of Scranton’s six-season total was booked in the last three years ($265.9M against $104.6M). Re-scoring each franchise-season as a z-score inside its own season removes the widening spread — but z is taken within season, so a season whose rows all carry extra weeks keeps its full vote and the franchises that gained most weeks keep their advantage inside it. Drop 2025 instead and the z leader is West Waco at 4.56, with Scranton second at 4.10.
The shape of the original finding does survive the cut, even though the order does not. Split the seam-free book at the elite line and all three leaders lose money above it and make it below — the difference between them is entirely how much they make below.
| Franchise | Total | Elite surplus | Elite return | Everything else surplus | Return |
|---|---|---|---|---|---|
| West Waco Wildcatters | $263.2M | −$29.0M | 0.949 | $292.2M | 1.482 |
| Scranton Stranglers | $370.5M | −$7.1M | 0.987 | $377.6M | 1.589 |
| Willamette Valley Coopers | $115.8M | −$136.4M | 0.784 | $252.2M | 1.504 |
We are the best of the three above the elite line at 0.901, and it buys nothing, because the spread up there is 0.873 to 0.899 and the spread below it is 1.505 to 1.744. West Waco beat us on the clean window by being better at the cheap end — 1.744 against our 1.505 — which is precisely the mechanism the first version claimed for us. It was the right mechanism attributed to the wrong franchise.
If the objection to the surplus table is that 2025 credits points nobody earned for us, the answer is to ask the question in a currency that contains no points at all.
Inside each season and position group, fit the league’s salary line against the log of the outside market’s average auction value, and take the residual: what the league would have paid for this player, minus what we actually paid. Call it the price leg. It is a function of our salary and an outside market’s price and nothing else — not one point scored in this league enters it — so no quantity of phantom player-weeks can move it. It is also the leg denominated in cap room, which is the thing that could actually fund a signing.
3,138 rows carry the outside price needed to compute it, and the league column sums to −0.000M, zero-sum as any such construction must be. Scranton is first of sixteen on it: $152.5M over six seasons, $25.4M a season. That is 41% of the published surplus figure, and 44% of the $344.3M of it that is matched at all.
And the price leg tells the same story as the corrected change test, independently. Regressed on the change in elite payroll it reads 0.135 raw and 0.256 with the baseline held, 95% interval 0.094 to 0.407, with 7 of 10,000 draws at or below zero. Two different constructions of “what a good contract year is worth”, one of them immune to the thing that broke the headline, and both say the room gets spent.
Split the six seasons into two disjoint halves of three, rank the sixteen franchises in each half, and correlate. There are exactly ten ways to do that. The first version printed all ten and drew the wrong conclusion from them.
| First half | Second half | Spearman | |
|---|---|---|---|
| 2020 / 2021 / 2022 | 2023 / 2024 / 2025 | 0.194 | the chronological one |
| 2020 / 2024 / 2025 | 2021 / 2022 / 2023 | 0.324 | |
| 2020 / 2021 / 2024 | 2022 / 2023 / 2025 | 0.391 | |
| 2020 / 2023 / 2025 | 2021 / 2022 / 2024 | 0.426 | |
| 2020 / 2021 / 2023 | 2022 / 2024 / 2025 | 0.441 | |
| 2020 / 2021 / 2025 | 2022 / 2023 / 2024 | 0.444 | |
| 2020 / 2022 / 2023 | 2021 / 2024 / 2025 | 0.415 | the null bar sits at 0.503 |
| 2020 / 2022 / 2025 | 2021 / 2023 / 2024 | 0.624 | clears it |
| 2020 / 2023 / 2024 | 2021 / 2022 / 2025 | 0.650 | clears it |
| 2020 / 2022 / 2024 | 2021 / 2023 / 2025 | 0.697 | the interleaved one |
“All ten are positive” was the wrong bar, and this piece used it. Ten splits of the same 96 franchise-seasons are not ten independent tests, and positive is not the same as meaningful. The honest reference is the 95th percentile of the absolute Spearman between two random rankings of sixteen items, which is 0.503. Only 3 of the 10 splits clear it — 0.624, 0.660 and 0.697 — and the seven that do not include the chronological split at 0.194, which was already flagged. The mean of 0.461 and the median of 0.434 both sit below the bar.
What that leaves standing is narrower and still real. The interleaved split — even seasons against odd — reaches 0.697 with a permutation p of 0.0013, and the 80 consecutive franchise-season pairs correlate year-over-year at r = 0.507. Within an era, contract surplus is a stable property of a franchise. Across eras it drifts, and six of sixteen franchises changed name or owner inside this window. Three of ten is a real signal reported at its real strength, not the wall-to-wall repeatability the first version claimed.
The one leg the seam cannot reach at all, because it is built from the salary ledger and phantom player-weeks add points, not dollars. Every team faces the same $220.0M soft cap: how much does each use?
| Franchise | Payroll per season | Unused cap | Cost per point over replacement |
|---|---|---|---|
| Scranton Stranglers | $204.6M | $15.4M | $116k |
| West Waco Wildcatters | $202.5M | $17.5M | $122k |
| Willamette Valley Coopers | $194.4M | $25.6M | $137k |
| San Antonio Stetsons | $187.6M | $32.4M | $140k |
| San Diego Hopheads | $185.9M | $34.1M | $227k |
| South Philly Pigeon Boys | $149.9M | $70.1M | $173k |
| league | $178.3M | $41.7M | $151k |
Scranton runs the highest six-season payroll in the league. $204.6M a season against a $220.0M cap is $15.4M of headroom, the thinnest of the sixteen, and top-three in every individual season (ranks 2, 1, 2, 3, 3, 2). West Waco — who lead the clean window — is immediately behind at $202.5M and $17.5M. The two franchises that come out on top of the corrected generation table are also the two that spend closest to the ceiling, which is the one part of the original thesis that never depended on the valuation model at all.
San Diego is the counter-example that makes it a finding rather than a tautology. Fifth in payroll at $185.9M a season, less unused cap than eleven other franchises — and $228k per point over replacement, the second-worst rate printed. Spending to the ceiling is necessary and nowhere near sufficient; San Diego proves you can run out of room and still finish last in surplus at −$374.7M.
Every contract-season is banded by the outside market’s opinion of that player — MFL’s global average auction value across roughly two thousand other leagues, a price nobody in this league set. Then: what does a dollar spent in each band return?
Every tier below the elite one returns more than it costs — and the further down, the better
| Outside rank | Contract- seasons | Salary | Surplus | Return per dollar |
|---|---|---|---|---|
| 1–50 — elite | 300 | $6,090.7M | −$676.5M | 0.889 |
| 51–150 | 597 | $5,663.1M | $131.5M | 1.023 |
| 151–300 | 799 | $2,481.4M | $261.5M | 1.105 |
| 301+ | 1442 | $1,713.3M | $287.2M | 1.168 |
| unranked outside | 948 | $596.1M | −$3.7M | 0.994 |
The league has put $6,090.7M into the fifty players the outside market rates highest each year, and got 0.889 back on the dollar. That is −$676.5M. It is the one headline of the first version that came through the review intact, and it survives both controls: drop the seam season and the elite band reads 0.872, slightly worse.
But the monotone ladder underneath it is partly a position ladder, not a price ladder, and that has to be said. There are zero defensive players in the elite band, while IDP is 73% of the salary in the unranked band and 33% at 301+. Comparing 1–50 against 301+ is therefore partly comparing offence against defence, which is not a comparison this shop makes in dollars. Restricted to offence only, the ladder flattens sharply at the bottom.
| Outside rank | All six seasons | 2020–24 seam-free | Offence only | Rows (offence) | IDP share of salary |
|---|---|---|---|---|---|
| 1–50 — elite | 0.889 | 0.871 | 0.889 | 300 | 0% |
| 51–150 | 1.023 | 1.047 | 1.054 | 477 | 15% |
| 151–300 | 1.105 | 1.114 | 1.129 | 379 | 41% |
| 301+ | 1.168 | 1.174 | 1.073 | 819 | 33% |
| unranked outside | 0.994 | 0.919 | 0.265 | 215 | 73% |
The elite band is 100% offence by composition — RB 45%, WR 37%, QB 14%, TE 4% — so the like-for-like test is elite offence against the rest of offence. That reads 0.889 on $6,090.7M against 1.055 on $7,611.3M. The finding survives, at about two-thirds of its advertised size: it is a 0.889-against-1.055 gap, not a 0.889-against-1.168 gap. Separately, IDP overall returns 1.091 on $2,842.5M, which is the actual explanation for most of the bottom of the original ladder and is a positional finding, not a price one.
| Franchise | Elite salary | Elite surplus | Elite return | Everything else salary | Everything else surplus | Return |
|---|---|---|---|---|---|---|
| Scranton Stranglers | $562.2M | −$7.1M | 0.987 | $641.0M | +$377.6M | 1.589 |
| West Waco Wildcatters | $564.8M | −$29.0M | 0.949 | $606.6M | +$292.2M | 1.482 |
| South Park Cows | $387.0M | $125.5M | 1.324 | $599.3M | +$95.6M | 1.159 |
| Willamette Valley Coopers | $632.3M | −$136.4M | 0.784 | $500.7M | +$252.2M | 1.504 |
| San Antonio Stetsons | $595.2M | −$182.2M | 0.694 | $499.2M | +$297.3M | 1.596 |
| Hollywood Wookies | $411.7M | $31.1M | 1.076 | $506.2M | +$78.1M | 1.154 |
| Washington Redskin Potatoes | $260.9M | $20.0M | 1.077 | $708.4M | +$39.1M | 1.055 |
| Houston Longhorns | $498.9M | −$109.1M | 0.781 | $540.6M | +$55.7M | 1.103 |
| Kawasaki Samurai | $191.6M | −$65.1M | 0.660 | $816.0M | −$277.2M | 0.660 |
| Las Vegas Gamblers | $166.9M | −$38.1M | 0.772 | $777.9M | −$330.7M | 0.575 |
| San Diego Hopheads | $265.3M | −$76.7M | 0.711 | $806.4M | −$298.0M | 0.630 |
South Park is the only franchise in the league that makes money at the elite tier — +$125.5M on $387.0M, a return of 1.326, against a league band average of 0.889. Nobody else is meaningfully above water up there; Hollywood and Washington clear 1.00 by a nose on small books. At the other end, San Antonio has given back $182.2M inside the elite line despite being the best sub-elite operator in the league at 1.595, and Willamette has given back $136.4M on the largest elite payroll of the sixteen. Those two are the clearest “good below, careless above” cases the data contains, and between them that is over three hundred million.
The floor is a different failure and it is not about the elite tier at all. Kawasaki, Las Vegas and San Diego carry the three smallest elite books in the league, and lose $277.2M, $330.7M and $298.0M respectively below the line. They are not the victims of star pricing. They are the reason the plus column exists.
Abstractions about bands are easy to misread. These are the actual Scranton contract-seasons at both ends of the distribution, with the outside market’s rank for the same player in the same year.
| Season | Player | Outside rank | Paid | Points | Surplus |
|---|---|---|---|---|---|
| 2025 † | Puka Nacua · WR | 24 | $0.5M | 241 | +$29.2M |
| 2024 | Baker Mayfield · QB | 124 | $7.5M | 277 | +$23.0M |
| 2024 | Sam Darnold · QB | 179 | $4.2M | 254 | +$23.1M |
| 2021 | Jonathan Taylor · RB | 19 | $27.9M | 336 | +$17.6M |
| 2023 | Jonathan Taylor · RB | 37 | $27.9M | 108 | −$18.6M |
| 2023 | Saquon Barkley · RB | 11 | $52.0M | 173 | −$32.8M |
| 2024 | DK Metcalf · WR | 40 | $44.6M | 134 | −$35.9M |
Every entry in the top half is a player the outside market did not rate at the top — two quarterbacks the wider world ranked 124th and 179th who returned 277 and 254 points. Every entry in the bottom half is a top-forty name on an eight-figure deal. Same franchise, same six years, and the sign flips with the price tag. Only 105 of Scranton’s 272 contract-seasons carry positive surplus at all; the ten biggest account for 29% of every positive dollar in that book.
† The Nacua row is the one 2025 line here, so its 241 points may include weeks the roster did not hold him. It is left in because it is the cleanest illustration of the mechanism in the file, and flagged because the rest of this article is about not letting that pass silently.
Payroll share by position against the league, next to a knob-free efficiency measure: this franchise’s raw fantasy points per dollar at that position, divided by the league’s. Above 1.00 means our dollars there bought more points than the league’s dollars did.
| Pos | Our share of payroll | League share | Our points per dollar ÷ the league’s |
|---|---|---|---|
| WR | 22.9% | 33.9% | 1.582 |
| TE | 5.2% | 7.6% | 1.500 |
| CB | 1.4% | 1.9% | 1.434 |
| QB | 12.6% | 12.6% | 1.161 |
| DL | 8.2% | 6.9% | 1.111 |
| RB | 41.6% | 28.7% | 0.858 |
| LB | 5.4% | 5.7% | 0.809 |
| S | 2.8% | 2.7% | 0.692 |
spending-impact.md already settled where money buys wins in this league, and it is cited rather than re-derived: RB spend is the one clean staircase into Victory Points, QB spend is flat and non-monotonic, and CB and S spend correlates negatively with winning. strategy.md turns that into an order — fill corner and safety at or near the minimum. The allocation above obeys both files: 1.4% at corner and 2.8% at safety, at or below league share.
The right-hand column is the uncomfortable part, and the seam does not explain it away. Running back is where we put 41.6% of the payroll and it is the position where our dollars buy fewer points than the league’s — 0.858. Receiver is where we are best in the league by a distance at 1.579, and we are eleven points of payroll share underweight there. One honest confound: our running back money is concentrated at the very top of the position, where the elite band’s 0.889 already tells you the returns are worst, so part of that 0.858 is a statement about the price band rather than about judgement.
The wins check does not validate what it was made to validate, and this is the third correction. The first version reported Spearman 0.801 between six-season surplus and head-to-head wins as external confirmation. Two things wrong with that. Surplus and realised production correlate at 0.954, so surplus nearly restates production, and production is what wins games — the check is close to circular. And raw production tracks wins better than surplus does: raw VORP against wins is 0.885 and raw points 0.823, against surplus’s 0.801. Every piece of valuation machinery layered on top makes the wins correlation worse, not better. The partial correlation of surplus with wins holding VORP is −0.330 with p = 0.234 and an interval of −0.713 to 0.239, so the sign is unreliable at n = 16 — but nothing there says surplus adds information about winning beyond the points it is built from. The rank order does hold at the extremes: we are first in surplus and first in wins at 67 of 83, and Las Vegas is fifteenth in surplus and last in wins at 19 of 83, with Kawasaki at 24 and San Diego at 34.
“Surplus” is substantially a bust count. 2,274 of the 4,086 contract-seasons — 56% — returned nothing over replacement, and $4,130.2M of the league’s $16,544.5M in salary sat inside them. The share of a franchise’s payroll parked in those contracts correlates with its six-season surplus at −0.762. That is most of the table. Scranton’s bust rate is 48% of rows, which is league-typical, but its busts average $1.2M against a $4.4M roster mean — a ratio of 0.277, second-lowest of sixteen behind Hollywood. “When we are wrong, we are wrong cheaply” is price discipline, and is also partly injury luck no method here can separate out. The seam pushes this the flattering way, too: extra weeks make a 2025 contract less likely to be counted a bust.
The dollar figure is amplified, and by a lot. In raw fantasy points per million dollars of salary — no replacement level, no position weighting, no exchange rate — Scranton scores 20.3 against a league 18.9. That is +7.1%, and fourth of sixteen, not first; drop the seam season and it is +5.0%, still fourth. Adjusting instead inside 205 season × position × market-band cells, so a franchise is only compared against others buying the same kind of player in the same year, we come out at 1.175 and third, behind San Antonio at 1.269 and West Waco at 1.235. Four defensible measures, four different rankings, and the first version published the most flattering one as the headline.
The position weight is a knob, not a measurement. build-value-vector.js maps a measured win correlation to a weight with a chosen floor of 0.5 and running back anchored at 1.0. Every dollar figure in this piece inherits it. That is why the positional table above uses raw points per dollar instead, and why no per-position surplus number is printed anywhere.
What the ledger cannot see is what a cut costs. Salaries come from a per-season roster snapshot, so a player released mid-season and the cut penalty that follows him — 100% of current salary plus 50% of remaining years, per rules.md §12 — simply are not in the data. A franchise that escapes a bad contract has that salary removed from its ledger rather than charged against it. Dead money is invisible in both directions, which means the floor teams’ losses are understated, and it means this measures contracts held, not contracts entered.
The bands are an outside, offence-first market. Only 3,138 of 4,086 rows (77%) carry an outside AAV at all, and the coverage is not random: most of the two thousand leagues that price is drawn from do not start IDP, which is why there are zero defensive players in the elite band. Every band statement is a statement about offence, and the like-for-like table above is the version to quote.
On the unit, because a standing directive covers it. D12 says contract asset value is trade points and not dollars, and that is deliberately not what this is. D12 governs contract.points, a forward-looking ranking whose dollar form could not express what a cheap young player is worth. This is value.realized — a backward-looking measurement of scoring against salary actually paid — and D18 exempts it by name: “It is not a model and it is honestly denominated.”
retain · 0.215
Last year’s surplus does buy this year’s elite payroll, at about 21 cents on the dollar, once the baseline is held: partial +0.331, interval 0.173 to 0.482, 0 of 10,000 draws at or below zero. That channel is extension and retention. It is already working and it is the one worth protecting.
acquire · −0.015
The league moves $27.9M of newly acquired elite salary per franchise-season, so the opportunity exists and is taken. It just has no relationship to who generated surplus the year before. Plan the cap around keeping and extending; treat an outside elite signing as an opportunity, not as the mechanism.
Do not budget for the elite tier as an investment. It returns 0.889 on the dollar league-wide, 0.871 without the seam, and 0.889 against 1.055 in the like-for-like offence comparison. Only South Park has beaten it over six seasons, at 1.324. Spending there is a purchase of certainty, not of value, and it should be budgeted as such — which is exactly what the retention finding says we already do, and the acquisition null says we correctly do not do more of.
And the concrete counterparty read. Contracts are worth selling to the franchises whose sub-elite books return least — Las Vegas at 0.573, San Diego at 0.630, Kawasaki at 0.660 — because those three are the reason the league’s plus column exists at all. That is the same trio the six-season table puts at the floor, arrived at from a different direction, and it is unchanged by every correction in this piece.
The first thing to fix is not a strategy, it is the store. Until build-realized-value.js filters to roster-weeks, every season after 2024 will be inflated for whoever churned hardest, and no ranking that includes it can be quoted. Until then, quote the 2020–2024 window, or quote the price leg, which is immune by construction.
Surplus. Insight value.realized, built by build-realized-value.js under spec 09. For each player-season: production over replacement is that season’s realised BWAFF points minus the score at the measured starter boundary for his position group, floored at zero; that is multiplied by a position weight; the season’s exchange rate is that season’s total committed salary divided by its total weighted production; and surplus is production valued at that rate, minus salary. Rebuilt per season so 2021 production is never priced in 2026 dollars.
The seam. That build sums fantasy.player.points with GROUP BY entity_id, season and no week or roster-week predicate. metrics-catalogue.md already recorded the consequence and nothing had read it: the fantasy family carries 17,282 facts for 2025 against roughly 8,600 for every other season, because player-scores contributes 8,665 unrostered player-weeks in that season alone. Everything computed on 2020–2024 in this piece is unaffected; everything computed on all six seasons is labelled as such.
The universe. 4,086 scored contract-seasons over 16 franchises and 6 seasons, from insight.contract.ledger, itself parsed from raw/rosters/. A further 435 ledger rows exist for these seasons and are unscored because their position does not map to a scoring group. Payroll figures use the full ledger; every surplus figure uses the scored subset. Head-to-head wins are fantasy.franchise.won, 83 games per franchise.
Bands. market.player.aav_rank — MFL’s global average auction value across roughly two thousand leagues outside this one, in the same season. Used only to sort players into tiers, never to value them, because it is the one price in this analysis that nobody in this league set.
The price leg. Inside each season × position-group stratum, the league’s salary is regressed on the log of that outside AAV; the price leg for a contract is the fitted salary minus the salary actually paid. 3,138 rows carry the join. It contains no in-league scoring of any kind, so it is unaffected by the seam.
The tests. Split-half repeatability is Spearman on franchise totals across all ten disjoint 3-and-3 season partitions, with permutation p-values from 20,000 shuffles at a fixed seed, read against a null bar of 0.503 — the 95th percentile of |rho| for two random rankings of sixteen, from 20,000 draws. The redeployment tests are partial correlations holding the baseline elite payroll, with 10,000-draw bootstraps that resample whole franchises, because five observations from one franchise are not five independent observations. Panel tests remove each franchise’s own mean from both sides.
Four things it cannot see. Cuts, dead money and cut penalties, as above. Trades — a contract that moves mid-window is credited to whichever franchise held it at snapshot time, so nothing here says who signed a good deal, only who was holding it. Draft capital, entirely: a franchise that traded picks to acquire a cheap contract paid a price this measurement never charges it. And injury, which is indistinguishable from misjudgement in a zero-floored production number.
Confidence, down from the 0.52 this piece first published, and deliberately left unnumbered. The honest sample is five seasons, not six. Inside it the top of the ranking is West Waco’s and ours is second by $28.7M, which is not a separation this n can defend as skill; only 3 of 10 split-half partitions clear the null bar. The claims this piece will stand behind are the elite band’s 0.890, the payroll finding, the price-leg rank, and the retention-versus-acquisition split at +0.331 against −0.015. The six-season ranking is not one of them.