SV Research · Paper 01 · Ongoing

Auditing a decade of a public betting record.

An open research project. We're reconstructing and independently grading a public handicapper's entire over/under record — every pick, from every source we can find — and showing our work as we go. Here's where we are, exactly how we're doing it, and what we still need.

Where we are

The subject advertises a career over/under record of roughly 775-643-74 (~55%) across the seasons he lists. Break-even against standard -110 pricing is 52.4%. We're rebuilding that record from scratch — every season, every pick — to see what it looks like when all of it is counted and independently graded.

Known gaps we're still chasing: 2014 (not archived anywhere we've found), the 2019 summer, and scattered early-season dates. Have a source, a screenshot, or an archive we're missing? Send it — see the bottom of the page.

What we're doing with the data right now: collection is largely done. We're in the verification pass — cross-checking every pick's line against independent market data to confirm the number was actually available to bet, and re-grading each result against the official box score. The ledger below is a live working dataset: it shows what we've collected and verified so far, and it updates as the pass completes. Nothing here is presented as final.

What counts, and what doesn't

The bar is strict, and it's the whole point. A play only counts if it was posted publicly on a reputable, timestamped forum — never self-reported — and if its line was independently verified to have actually been available. If we can't confirm the number cited was real, it's out.

How we grade and calculate

Every pick is graded independently — we never take anyone's word for the result.

• The result — settled against official box-score data. We pull the game's final score, total the runs, and grade the pick against his line — win, loss, or push. No interpretation, no self-reporting.
• Profit, in units — profitability is measured in units at every play's actual odds, summed across the whole record — not a simplified "hit 52.4% at -110" assumption. Net-positive means he beat the vig; net-negative means he didn't. It's an independent question from raw win rate.
• Line & odds — his stated line and price are cross-checked against historical odds from reputable sources, confirming the number was real and actually available to bet — not a favorable invention.
• Units & ROI — flat one unit per play, at his actual odds where we have them, to show real profit or loss — not just a win rate.
• Self-report concordance — where he logged his own result, we compare ours to his to measure how accurately he reports.

The question we're answering

It comes down to one hypothesis: is Matt Zylbert better than an average bettor? Several independent tests get us to an answer.

• Win rate, in context — win rate on its own tells you nothing. We examine it against the odds he actually played and the units deployed on those odds — across every season, not only the ones he advertises.
• Profit in units — combined units at every play's real odds: does the record actually make money, or just look like it?
• Statistical significance — is any edge real, or within the noise of a coin flip?
• Integrity — concordance with his own logs, and whether picks were posted before the games.

Season by season

The record spans nine independent publications — from handicapper archives like VegasInsider and Action Network to his own public posts on Twitter/X. The Source column shows exactly where each season's picks came from; his single highest-volume year (2013, 552 plays) is also one of his better seasons.

The data, so far

This is what we've collected — every play, open. Search it, sort it, and open any pick at its source. Take any row — the date and teams, his side and line, our independent grade, and the actual game total — and check it against a box score yourself. The ledger grows as we recover more.