← Research · Above Replacement
"Stream your kicker and defense" is standard advice. Stream them by what is the question nobody tests. We backtested the candidate signals across eight seasons, hindsight-free — each week's pick uses only information available before kickoff, drawn from a realistic pool of waiver-wire leftovers.
| Strategy | Season points |
|---|---|
| DST — lowest opponent implied total (Vegas) | 137 |
| DST — chase last week's points | 103 |
| K — highest own-team implied total (Vegas) | 148 |
| K — chase last week's points | 128 |
| K — the season-long "set and forget" K12 | ~135 |
Two rules, both from the betting market:
Everything else we tried added nothing once the market number was in: trailing fantasy points (worse by 20–34 points a season), kicker "leg strength" history, combinations of signals. Most surprising: the same matchup logic fails at quarterback — picking weekly QB matchups scored 216 against 253 for simply keeping the best QB you own. Streaming is a K/DST edge, not a universal one.
A Vegas implied total is the distilled opinion of the most heavily incentivized forecasters in sports, updated to the minute. A season-old preseason ranking — which is what most draft-day defense picks are — carries no measurable signal. The practical conclusion: never spend a meaningful draft pick on K or DST, take one of each in the last two rounds, and stream by the lines from week one.
Backtests can flatter their author. So this season, every weekly pick these rules produce is logged before games and scored after — misses included — on the accuracy page. Judge the rules there, not here.
Methodology: eight seasons of weekly results; pick pools restricted to players realistically on waivers (preseason leftovers, and league-reconstructed wires where box scores allowed); implied totals from closing lines. The rules power the streaming logic in the Above Replacement toolkit.