Six seasons of in-season pickups, reconstructed from weekly roster membership and scored against what a real starter at that position actually produced — not by how often they got used, but by whether the starts were any good. The answer is not a gradient. It is a cliff, and it runs exactly along the offense/defense line.
A start count is usage, not performance. A quality start is one at or above the median starter’s score that season; a dud is below the 25th percentile — a week he was in your lineup and actively lost it. This is quality minus duds, per pickup made.
At quarterback, running back and receiver, a street pickup delivers more actively-bad starts than good ones. A street quarterback’s start is a dud 52% of the time — a coin flip that you lost the week by making it. CB, S and DL are net positive in all six seasons individually; LB in four of six; TE sits on zero.
| Pos | Starts per pickup | Quality of starts | Duds of starts | Elite of starts | QS per pickup | Duds per pickup | Net |
|---|---|---|---|---|---|---|---|
| CB | 3.1 | 50% | 24% | 23% | 1.56 | 0.74 | +0.82 |
| S | 2.3 | 54% | 25% | 24% | 1.24 | 0.58 | +0.66 |
| DL | 2.3 | 45% | 26% | 21% | 1.03 | 0.59 | +0.43 |
| LB | 2.2 | 43% | 30% | 20% | 0.95 | 0.65 | +0.30 |
| TE | 1.4 | 39% | 35% | 20% | 0.53 | 0.49 | +0.05 |
| WR | 1.5 | 33% | 41% | 15% | 0.50 | 0.63 | −0.13 |
| RB | 1.6 | 31% | 45% | 14% | 0.51 | 0.73 | −0.22 |
| QB | 1.0 | 30% | 52% | 8% | 0.31 | 0.53 | −0.22 |
54% of a street safety’s starts are quality, ahead of corner’s 50%. CB only leads on volume — 3.1 starts per pickup against 2.3 — which measures how often you are forced to start one, not how good he is.
Roughly a quarter of street CB and S starts land in the top quartile of all starters at the position. At quarterback it is 8%. The defensive wire is genuinely boom-capable; the offensive wire is not.
The usage view, kept because the gap between it and the quality view is the point. P(a street pickup becomes a real starter) × (his points per game ÷ a median starter’s) — it puts the cliff in the same place, one position further down.
The street covers it — don’t spend an August roster spot
The street cannot — a bench body is the only insurance there is
Street pickups only. Roster-to-roster moves are a different act at a different price and are split out further down.
| Pos | n | Started same wk | Ever started | Became a real starter | ppg vs median | Median starter | Index |
|---|---|---|---|---|---|---|---|
| CB | 248 | 71% | 89% | 35% | +0.13 | 5.5 | 34.0 |
| DL | 184 | 49% | 72% | 23% | −0.19 | 4.5 | 23.4 |
| S | 139 | 48% | 73% | 25% | −0.44 | 7.0 | 21.1 |
| LB | 231 | 44% | 67% | 25% | −2.54 | 8.0 | 16.6 |
| RB | 213 | 38% | 55% | 17% | −6.24 | 10.9 | 7.0 |
| TE | 111 | 39% | 55% | 14% | −3.33 | 8.3 | 6.5 |
| WR | 240 | 36% | 60% | 14% | −4.85 | 10.4 | 6.5 |
| QB | 98 | 35% | 52% | 9% | −11.49 | 21.2 | 3.0 |
The bar is the median starter at that position — and at CB that is 5.5 points a week against 21.2 at QB. So the finding is that CB replacement is cheap because CB is cheap, not that corners off the street are good.
This answers one question: can I skip carrying a backup? It says nothing about replacing an above-median starter of our own. Danielle Hunter at DL9 projects roughly 8.5 a week against a 4.5 DL median; nothing on the wire replaces that. The index is about depth, never about starters.
The same numbers without the positional bar underneath them, so the column above can’t be read out of context.
| Pos | Median starter | Street add | Add as % of a starter |
|---|---|---|---|
| DL | 4.5 | 4.51 | 100% |
| CB | 5.5 | 5.32 | 97% |
| S | 7.0 | 5.87 | 84% |
| LB | 8.0 | 5.37 | 67% |
| TE | 8.3 | 3.99 | 48% |
| WR | 10.4 | 4.76 | 46% |
| RB | 10.9 | 4.49 | 41% |
| QB | 21.2 | 6.82 | 32% |
Points per game over the positional median, by when in the season the pickup was made. A late add is made with information about who is actually playing.
| Pos | Weeks 2–5 | Weeks 6–10 | Weeks 11–14 | Late premium |
|---|---|---|---|---|
| RB | −7.01 | −6.53 | −4.77 | +2.24 |
| WR | −5.55 | −4.93 | −3.60 | +1.95 |
| CB | −0.38 | −0.15 | +1.57 | +1.95 |
| QB | −11.96 | −11.88 | −10.48 | +1.49 |
| DL | −0.80 | −0.16 | +0.66 | +1.47 |
| TE | −3.82 | −3.31 | −2.80 | +1.02 |
| S | −0.46 | −0.19 | −0.69 | −0.24 |
| LB | −2.79 | −1.97 | −3.17 | −0.38 |
A trade or a claim brings a player somebody else already valued, and it shows almost everywhere — except at defensive line, where it reverses.
| Pos | Street n | Ever started | ppg vs median | Roster-to-roster n | Ever started | ppg vs median |
|---|---|---|---|---|---|---|
| CB | 248 | 89% | +0.13 | 34 | 94% | +1.97 |
| S | 139 | 73% | −0.44 | 23 | 91% | +1.80 |
| DL | 184 | 72% | −0.19 | 27 | 63% | −0.70 |
| LB | 231 | 67% | −2.54 | 29 | 86% | −1.80 |
| RB | 213 | 55% | −6.24 | 66 | 88% | −2.43 |
| WR | 240 | 60% | −4.85 | 82 | 83% | −3.20 |
| TE | 111 | 55% | −3.33 | 29 | 69% | −1.66 |
| QB | 98 | 52% | −11.49 | 51 | 82% | −6.35 |
Share of street pickups that ever started, by season. CB never drops below 85; QB and TE swing more than thirty points.
| Pos | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | Spread |
|---|---|---|---|---|---|---|---|
| CB | 86% | 88% | 85% | 95% | 93% | 86% | 10 pts |
| S | 74% | 64% | 81% | 83% | 72% | 65% | 19 pts |
| DL | 71% | 69% | 86% | 74% | 70% | 65% | 20 pts |
| LB | 62% | 60% | 71% | 79% | 62% | 79% | 19 pts |
| WR | 59% | 62% | 59% | 59% | 64% | 61% | 5 pts |
| RB | 64% | 58% | 43% | 52% | 56% | 58% | 21 pts |
| TE | 54% | 54% | 67% | 37% | 58% | 67% | 30 pts |
| QB | 72% | 50% | 44% | 41% | 47% | 56% | 31 pts |
Street pickups by week, all positions, all six seasons. The wire is busiest through the bye stretch and thins out after week 10 — which is also when the pickups get better.
Street pickups per franchise per season, and how many of the 96 franchise-seasons never made one at that position at all.
| Pos | Mean / season | Busiest franchise-season | Never touched it | Best single add on record |
|---|---|---|---|---|
| CB | 2.6 | 11 | 13 of 96 | 2023 wk10 — +9.5/wk over 5 |
| LB | 2.4 | 11 | 22 of 96 | 2021 wk9 — +7.8/wk over 6 |
| DL | 1.9 | 11 | 24 of 96 | 2024 wk5 — +7.7/wk over 10 |
| S | 1.4 | 7 | 31 of 96 | 2022 wk2 — +7.6/wk over 4 |
| WR | 2.5 | 10 | 15 of 96 | 2024 wk8 — +10.6/wk over 7 |
| RB | 2.2 | 10 | 18 of 96 | 2020 wk10 — +10.5/wk over 4 |
| TE | 1.2 | 8 | 32 of 96 | 2021 wk9 — +12.2/wk over 6 |
| QB | 1.0 | 6 | 41 of 96 | 2024 wk8 — +4.8/wk over 7 |
CB · DL · S · LB
Don’t buy depth here in August. Every one of these returns more good starts than bad, in every season measured. Even a bye-week hole is cheaper to patch in week 5, with information, than to insure against in August.
QB · RB · WR · TE
Here the wire is worse than nothing: more duds than quality starts, every year. If a starter goes down and no one is behind him, the replacement you find will lose you the week roughly half the time. There is no wire fix to buy later.
An add is a player on a franchise’s week-W roster who was not on it in week W−1, reconstructed from roster membership rather than a transaction feed — so it silently includes IR activations and taxi promotions alongside real pickups. It is split by whether the player was on any roster the previous week: not anywhere means street, somewhere means a trade or a claim. Each add is scored against the median weekly score of an actual starter at his position, computed per season so scoring drift can’t leak across years. “Became a real starter” means held four or more weeks and started in four or more of them, so a one-week bye plug doesn’t read as roster-changing. Start quality is graded against the same distribution: at or above the median is a quality start, below the 25th percentile is a dud, at or above the 75th is elite. Quality is only measured over weeks he was actually started, because that is the only period he could have helped or hurt — a pickup who never started scores no quality starts and no duds.
Three things it cannot see. Acquisition price — blind-bid dollars are cap dollars, so a position can be street-replaceable and still expensive to replace. Who was dropped to make room, so “the add helped” is never measured net of its cost. And weeks 15 onward, because the archive stops at week 14 (2020 at 13), which truncates the tail of every add’s production.
Confidence ceiling 0.49 — realised, not forecast. This measures what happened to 1,464 historical pickups. It does not forecast that the 2026 wire will behave the same way.