← Research · Above Replacement
Value over replacement is the spine of every serious draft model: a player is worth what he scores above the guy you could get for free. Set that bar wrong and every number downstream is wrong. So what should "free" mean — and does anyone check?
The streaming bar sounds rigorous but hides a trap: "start the best-projected free agent each week" credits the winner's full projection, and selecting the maximum of twenty noisy forecasts systematically selects the optimistic ones. So we reconstructed five years of a real 12-team league's waiver wire from box scores and measured what streaming actually returned.
| Position | Projected streamer bar | Realized off the actual wire | Verdict |
|---|---|---|---|
| DST | 105 | 97 | Holds — streaming defense is real |
| QB | ~300 | 144–199 | Fails — off by 100+ points |
Why the QB bar collapsed: as more of the league rostered a second quarterback (rostered QBs rose from 15 to 25 over the five years), the wire emptied. A replacement bar built on a wire that no longer exists inflates every QB's "waitability" — and under-prices the elite ones.
The bar must be the best validated, executable alternative — never a blend, never a projection nobody can realize. In practice: a streaming bar only where streaming demonstrably works (defense), a worst-starter bar everywhere else, and both derived from your league's actual size, lineup, and scoring. When we corrected the QB bar this way, elite quarterbacks gained roughly twenty points of draft value overnight — the difference between "wait forever" and "elite or wait," which eight seasons of rankings data independently supports.
Next in this line of work: rerunning the wire reconstruction across hundreds of public leagues to measure how replacement level varies by league environment. The corrected bars are what the Above Replacement draft tool computes for every connected league.