We went 2-for-8 on our first tracked slate. Rather than shrug and reload, we went back and figured out why — and the answer changed how we rank picks entirely.
What we were doing wrong
We were ranking picks by hit rate. That sounds obviously correct and it is obviously wrong. Hit rate tells you what has been happening. It does not tell you whether a bet is good — because the market already knows what's been happening and prices it in.
Here's the clearest example we found, straight off tonight's board:
Oswald Peraza — Under 0.5 Doubles. 100% over his last 5. 100% over his last 10. 90% over his last 20. A near-perfect record.
The price? −6000. You'd risk $6,000 to win $100. The market has it at a 98% implied chance. There is nothing here.
Under our old "rank by hit rate" approach, that pick sails to the top of the board. It looks incredible. It's worthless.
The fix: Edge Gap
Every line carries an implied probability — what the odds say the chance is. Every player carries a hit rate — what's actually been happening. The space between those two numbers is the only place an edge can live.
Edge Gap = blended hit rate − implied probability
Two picks on tomorrow's board, both showing an identical 90% over their last 10 games:
| Player | Pick | Odds | Implied | L10 | Edge Gap |
|---|
| Gunnar Henderson | Under 1.5 Bases | −119 | 53% | 90% | +37 |
| Moisés Ballesteros | Under 1.5 Bases | −200 | 67% | 90% | +24 |
Same hit rate. Very different prices. Old method ranks them equal — new method doesn't.
We also stopped trusting ten games
A 90% "last 10" is nine-for-ten. Run the statistics on nine-for-ten and the honest confidence range is roughly 55% to 100%. That is weak evidence, and we were presenting it as strong evidence. It's the other half of why we went 2-for-8.
So the hit rate we now rank on is blended and weighted toward bigger samples: 40% full-season, 30% last-20, 20% last-10, 10% last-5. Recent form still counts — it just no longer overrules a whole season.
The Card — Top 10 by Edge Gap
The case against our own #1 pick
We're going to make a habit of this: publishing the strongest argument against the pick we ranked first.
Henderson's number is built on a slump, not a skill. He has recorded zero total bases in six of his last seven games — average 0.5, median 0. That's what's driving the 90% last-10. But his full-season rate is 65%. A player this far below his own baseline is arguably due to correct, not continue.
The pitcher matchup argues the other way. Minnesota starter Bailey Ober carries a 4.45 ERA and 6.10 strikeouts per 9 on the season — but against Baltimore specifically, those numbers are 6.75 earned runs per 9, 9.00 hits per 9, and just 2.25 strikeouts per 9. Weaker pitching means more offense, which cuts against an Under. (Caveat on the caveat: that split is only 4 innings.)
We're keeping it at #1 because the price is genuinely out of line with the history, and that's what the method is designed to find. But anyone reading this deserves to know the pick has real counter-evidence, and that we saw it.
What we threw out
Two picks were excluded for being too good. Taylor Walls and Chandler Simpson (both TB @ ATH) showed Under 0.5 RBIs priced at +9251 — an implied 1% chance on something that has happened 80–90% of the time. That's an Edge Gap near +89.
A gap that large is almost never real value; it's a stale or broken line at one book. Our rule from here on: anything over +40 gets manually verified, not auto-published. These two didn't survive the check.
Verification
Henderson is not on Baltimore's injury list (their IL is five pitchers and a third baseman). Minnesota is missing five starting pitchers, which is why Ober draws this assignment. Full lineups aren't posted until closer to first pitch — if any of these ten are scratched, the leg is void, and we'll say so on the track record page rather than quietly dropping it.
How to judge this card
This is a live test of a new method, and we're stating the standard in advance so we can't move the goalposts later.
The bar: Edge Gap needs to beat our old approach across a meaningful sample — not one slate. One card tells us almost nothing either way. We'll grade all ten legs the morning after, publish them hit or miss, and keep a running comparison against the old hit-rate-only method.
For entertainment and informational purposes only. This is not betting or financial advice. Past hit rates do not guarantee future results — a point this entire article is arguably about. 18+. If you or someone you know has a gambling problem, call 1-800-GAMBLER.