The Daily Grind

The Current Experiment: Do Handicappers Actually Beat the Number?

The Grind is our working notebook — whatever hypothesis we're actively testing lives here. Right now that's handicappers. We scrape the picks people post publicly, re-grade every one of them against the box score ourselves, and price a paper bet at the real market ask. No money is on any of it.

How the data gets here

Two feeds, both public and both timestamped before first pitch. From X we pull tracked handicappers' posts directly. From YouTube we pull the daily "free picks" videos and read the transcript — that tier of channel makes money from ads and subscriptions, so the entire card is given away out loud. We currently track 122 handicappers across both — the roster grows on its own, so this count comes from the ledger, not from memory.

Every pick is re-graded from the final score. We never use the capper's own claimed result: when we checked, our grade and their claim disagreed on a quarter of the overlapping picks, and on every single one we could verify independently against the box score, ours was the correct one.

Yesterday's card

On Tuesday, August 11, 2026 we independently graded 67 picks from 21 handicappers. Pooled, they went 32-35 for -8.06 units (-12.0%) at the prices they posted.

HandicapperSourcePicksW-LUnits
Free Picks PickDawgzYouTube43-1+2.53u
Tony's PicksYouTube74-3+1.89u
LVTruckX22-0+1.64u
FeedintipsX22-0+1.62u
NickicashinX74-3+1.03u
CarlowguybetsX11-0+0.84u
Guy Boston SportsYouTube21-1+0.44u
AnderspicksX21-1+0.31u
WagerTalk TV: Sports Picks and BetYouTube21-1+0.05u
Greedo PlaysYouTube21-1-0.20u
WagsX42-2-0.23u
SharpfootyX21-1-0.24u
StumpTheSpread - Free Sports PicksYouTube52-3-0.50u
MrbigbetsX42-2-0.76u
BettingProsYouTube10-1-1.00u
AI Sports Predictor YouTube10-1-1.00u
BetspidX10-1-1.00u
Doc's Sports Picks, Best Bets, PreYouTube73-4-1.38u
ThesharpestsyndicateX10-1-1.50u
ParlaysafariX30-3-5.00u
Picks4DayzzzX72-5-5.59u

The forward test

Historical records are the easy part and they lie. A handicapper picked out of a crowd for having the best record is mostly the winner of a coin-flip tournament, so the only number we'll trust is one that accrues after we commit. Every qualifying pick gets logged pre-game at the price we could actually have paid, and settled from the box score.

The forward test has logged 107 paper legs at real Polymarket asks, of which 107 have settled: 52-55, -28.47 (-5.3%). Solo-capper legs toward the review gate: 55/100.

What we already know won't work

Certifying an individual handicapper is not possible at these sample sizes. To distinguish a genuine 4% edge from luck takes roughly 2,400 graded picks — six years at a pick a day. We tested the obvious gates against twenty thousand simulated zero-skill bettors: requiring a positive record over fifteen picks lets 49% of pure noise through. So we don't rank anybody as "proven". We rank for attention, and we let the pooled result decide, because it accumulates about twenty times faster than any one person's.

Paper only. Nothing on this page is a bet we placed, and none of it is advice. The retired A/B/C model's full daily history is preserved in the archive.