Methodology

How it works

I scrape sharp Twitter, Reddit POTD threads, and public-money splits, post the picks before tipoff with a written read, and grade every result in public. Wins and losses both stay on the record.

The Pipeline

Scrape

Sources scanned

Analyze

AI extraction

Score

Confidence 1-10

Posted before tip-off

Grade

W / L / P

Track

CLV + ROI

Watch the sources

I pull from 159 active sources — sharp Twitter accounts, Reddit POTD threads, pro handicappers, and public-money splits. Every post with a specific, actionable bet gets captured.

Tracking what actually matters: line movement, ATS trends with 100+ game samples, sharp money indicators, and Reddit consensus. Vague hot takes get filtered out before they ever hit the slate.

Pull out the actual bet

Each post gets parsed into the canonical fields: team, line, odds, sport, bet type, units. Spreads, moneylines, totals, props, parlays — all normalized so they can be graded the same way later.

A 4-agent debate (matchup, market, public-vs-sharp, situational) + a judge runs on every pick to flag the weak ones. Picks backed by stats + matchup data score higher; gut-feel ones score lower. That score becomes the confidence rating.

Score the conviction

Every pick gets a 1-10 confidence score based on the strength of the underlying read. Source track record matters too — handicappers with 10+ graded wins get a small bump.

Sources that produce winners earn reputation that boosts their next picks. Sources that whiff get a tag. Self-correcting over time toward the sharper signals.

Grade in public, track CLV

After each game, picks are graded Win, Loss, or Push from the actual final. CLV — Closing Line Value — is also tracked: the difference between the line when the pick posted vs. the line at tipoff.

CLV is what serious bettors look at — consistently beating the close means you're getting better numbers than the market. Win-rate is what the public reads; CLV is what tells you whether the picks are real.