Strategy Scoreboard
Three strategies, same settled-pick universe, bootstrap 95% confidence intervals on ROI and hit rate. The narrow point estimates everyone quotes are noise at small samples. What you want to watch is whether the gold bar (Lakeshore's CI) sits clearly to the right of the others — that's when the model is beating Pinnacle in a way the sample size can actually support. Snapshots accumulate daily so you can watch the CIs collapse over time.
Three-way comparison
Each strategy is evaluated on every settled pick where we stamped both a Pinnacle closing line
and a FanDuel decimal — the only universe where all three are directly comparable. Stakes are
flat $$10 per pick so payout asymmetries don't distort ROI. Pinnacle-favorite-at-FanDuel
derives the opposing-side decimal from the vig-free inverse when our pick disagreed with Pinnacle,
which is approximate but the only path without storing both sides per game.
Computing bootstrap CIs...
How CIs are collapsing — daily history
One snapshot per UTC day. Watch the CI width column shrink as the sample grows; that's the proper
signal that the comparison is becoming decision-relevant. Below ~120 picks the bands will overlap
heavily and the comparison can't distinguish strategies.
Loading snapshot history...
Pinnacle closing lines stamp via the snapshot loop introduced on 2026-05-23. Picks settled before that date can't be backfilled because Pinnacle data wasn't being captured. The universe grows daily as new games settle with stamps attached.