The AlgorithmRecord · The Spicy Boys

Track record

Every recommendation this system makes is logged before the week is played and scored against what actually happened. This page shows only graded, finished weeks — never live advice.
Weeks graded
1
Weeks logged
2
Metrics version
2026-08-22.1
Graded at
2026-08-28T16:24

Headlines

Metrics

Which source ranks a week best
Which source ranks a week best — FantasyPros ECR, Sleeper's projection, or our ensemble?
No week has been graded against actuals yet.
n=0 graded weeks
Waiver list order
Is the waiver list in the right order, and did the adds beat the drops?
No waiver target has been graded yet.
n=0 graded weeks
Start / sit calls
Were the start/sit calls right, and how much would the recommended lineup have left on the bench?
No close call has been graded yet.
n=0 graded weeks
K / DEF streaming order
Is the K / DEF streaming order right?
No streaming pick has been graded yet.
n=0 graded weeks
Draft P(available at my next pick)
When the draft assistant said a player had a p% chance of lasting to my next pick, did p% of them last?
421 shown players across 5 valid draft(s): predicted 49% available on average, 47% actually were; worst 0.1-0.9 bin off by 0.1445. Only 5 independent observation(s) so far (n_independent < 30) — a sighting, not a result.
n=421 valid drafts
Draft regret vs the alternatives shown
Did the players the engine was showing beat the player I actually took?
no logged draft's season has been played yet
n=0 graded picks
Late-round upside chips
Do the late-round darts the panel put an UPSIDE chip on actually hit?
64 upside claims logged; the drafted season has not been played yet
n=0 players shown with an upside chip
Breakout / regression flags
Do the breakout and regression flags land?
no flag list for this season (analysis.json is built for the upcoming season; a season's flags can only be graded once that season has been played)
n=0 flagged players
Jordan vs model
Do Jordan's own calls beat the model's flags, and each other by confidence?
5 call(s) logged; season 2026 has not been played, so none has reached its horizon
n=0 distinct player x horizon

Per week

WeekPlayedLoggedWith actualsVerdict
2024 wk 5yes79411Ranking a week (Spearman vs actual points, n=350): ecr +0.631 · sleeper +0.598 · ensemble +0.626 — best: ecr.
2026 wk 1not yet500Source ranking: not graded (no player has an actual and all three predictors this week).

What would change the model

Nothing on this page is a live recommendation: no FAAB figure, no current-week waiver target, no roster. It is the scoreboard, kept so the system can be judged on it.