Auditing Myself: Worst Start to MLB Gambling I’ve Ever Had

I use Claude to build my MLB model, so I asked it to audit all my bets so far and provide an honest analysis. I could analyze it myself, but I may have inherent biases that would skew the analysis one way or the other (being too hard on myself, or pumping myself up too much).

This is the worst start I’ve ever had, so it’s important we find out why. Here’s what Claude says.

2026 MLB Betting Season: An Honest Mid-Season Report Peter Appel | Through May 17, 2026

The record says 25-37-1. The ROI says -18.5%. The P/L is -8.53 units. If that’s all you looked at, you’d close the tab. But there’s a more complete picture here — one that takes longer to tell and, frankly, gives more reason for measured optimism than the headline suggests. This is that picture.

The Raw Numbers

Through 63 settled bets this season, the ledger reads: a record of 25-37-1 (with one push on a Milwaukee F5 ML), a net loss of 8.53 units, an ROI of -18.5%, and approximately 46 total units wagered.

That’s a losing record by any measure. But sports betting isn’t evaluated on record alone — it’s evaluated on whether you’re finding value, and whether the losses are the result of bad process or bad luck. The CLV data, which we’ll get to, makes a strong case for the latter.

Bet Sizing: Bigger Isn’t Better Here

At the quarter-unit level, the record is 2-2 with a +0.08u profit and a +5.3% ROI. At the half-unit level, it’s 7-8 with a +0.29u profit and a +5.2% ROI. Both smaller tiers are in the black.

Then it falls apart. At 0.75 units — the most common bet size in the log — the record is 10-18, -5.08u, -19.9% ROI. At one unit and above, it’s 6-9-1, -3.82u, -19.8% ROI.

The smaller bets are actually profitable. Every tier above 0.5 units is a money-loser. This isn’t necessarily a sizing problem — it’s more likely that higher-confidence plays are getting punished by variance, while disciplined lower-unit bets are holding. Still, it’s worth noting: right now, the bets placed with the most conviction are the ones doing the most damage.

By Bet Type: One Clear Edge, One Dead Angle, and a Rotten Middle

First Five Innings (F5): 4-2-1, +1.38u, +35.8% ROI — the one standout angle

F5 bets are the best-performing category by a wide margin. The Minnesota F5 at +145 (+1.09u), Colorado F5 at +141 (+0.71u), and a Detroit F5 double on April 22 all hit. When the process is right — identifying starting pitcher mismatches early in games — this is where the edge shows most clearly.

Moneylines: 10-15, -1.96u, -10.7% ROI — underperforming but not broken

ML bets are net negative, but the breakdown is nuanced. The plus-odds underdogs — Athletics twice at +148 and +136, Pirates at +135 and +125, Colorado at +152 — were identified correctly and got rewarded. The drag is a string of 0-1 teams (Blue Jays, White Sox, Angels, Mets, Padres, Rays) that each were bet on once and lost. No single team stands out as a recurring mistake; it’s the volume of one-off losses that adds up.

Totals: 5-9, -4.52u, -35.2% ROI — the biggest problem

This is where the season has been lost. A brutal March stretch — three totals bets in the final week of the month, all losses, -3.20u combined — set the tone. Overs are 0-3 (a dead angle entirely), and the Under bets that look reasonable on paper keep getting pushed Over. The Nationals-Cubs Under 7.5 at -1.20u on March 27, the Nationals-Phillies Under 9 at -1.00u on March 30, and a string of game-total misses since then account for the bulk of the season’s losses.

Props: 5-10, -2.95u, -30.1% ROI — volatile and inconsistent

Props are the most fascinating category. Strikeout overs are 2-6 and the worst subcategory in the log — Tatsuya Imai, Aaron Civale, Aaron Nola, Kyle Freeland, and others have all missed. Meanwhile, prop unders (hits allowed, pitching outs, earned runs allowed) are 3-2 and quietly profitable. The signal is clear: prop overs are a losing angle, prop unders are working.

By Team: Where the Money Was Well-Spent

On the winning side of the ledger, the Athletics are 2-1 on directional bets for +1.13u — two big plus-odds wins in April offset one recent loss at -128.

The Pirates are 3-3 for +0.61u, a break-even record that’s actually profitable because the wins came at +135, +125, and +103 while losses came at shorter numbers.

Colorado sits at 3-3 for +0.29u, with three consecutive wins to open the season, followed by two recent losses.

The Guardians are 1-0 for +0.83u on a single plus-odds ML hit.

Detroit is 2-1 for +0.14u after an F5 and ML double on April 22, both connected.

Kansas City is a clean 1-0 for +0.50u.

On the losing side, the Angels are the clearest money sink at 0-2 for -1.50u — both an F5 and an ML missed.

The Mets, Padres, and Blue Jays are each 0-1 for -0.75u.

The White Sox and Rays are each 0-1 for -0.50u.

Minnesota is 1-2 for -0.41u despite a +1.09u F5 return, dragged down by two ML losses.

The CLV Story: The Number That Actually Matters

Closing Line Value (CLV) measures whether you beat the final market price — whether you got better odds than where the line settled. It’s the sharpest available indicator of whether a bettor has an edge, because the closing line represents the collective wisdom of the most informed money in the market.

Out of 62 settled bets with available closing odds, 59 beat the closing line (95.1%) (AVG CLV: +3.33%).

The three that didn’t were the Texas Rangers ML (bet -119, closed -110, -1.9% CLV), the Blue Jays ML (bet +116, closed +130, -2.8% CLV), and the Twins ML (bet +126, closed +135, -1.6% CLV).

Everything else — every other bet — was placed at better odds than the market ultimately settled at. The average CLV across the sample is approximately 3-4 percentage points per bet.

Some of the biggest CLV plays in the log tell an interesting story.

The Tatsuya Imai Over 5.5 strikeouts was taken at +140 and closed at -129, a +14.6% CLV. Loss.

Cody Bolton Under 3.5 strikeouts came in at +110 and closed at -160, a +13.9% CLV. Loss.

MacKenzie Gore Over 14.5 pitching outs was bet at -105 and closed at -185, a +13.7% CLV. Loss.

The Athletics-Blue Jays Under 9.5 was taken at -105 and closed at -160, a +10.3% CLV. Win.

The White Sox-Brewers Under 8.5 came in at even money and closed at -145, a +9.2% CLV. Loss.

The Walbert Ureña Under 4.5 hits allowed was bet at +110 and closed at -125, an +8.0% CLV. Loss.

Notice anything? The five biggest CLV plays that weren’t game totals all lost. The market moved aggressively in the right direction — validating that these were identified as value at open — and then the result went the other way anyway.

That’s variance.

It’s also a caution flag worth naming: consistently getting large CLV on opening props that correct dramatically by close is a double-edged signal. It confirms you’re shopping lines well, but when the result doesn’t follow, it raises the question of whether the opener or the closer is the more efficient price.

So What Does It All Mean?

The honest case for optimism:

With roughly 3-4% average CLV across 46 units wagered, the mathematical expectation for the season is approximately +1.5u to +2.0u.

Instead, the season sits at -8.53u. That’s roughly a 10-unit swing below expectation — approximately 1.5 to 2 standard deviations below the mean for a sample this size.

It’s bad luck, not necessarily bad betting.

The F5 bets are working on both dimensions: positive CLV and positive results. That double confirmation — the market agrees these were valuable, and they’re winning — is the clearest signal of genuine edge in the log.

The honest case for concern:

Sixty-three bets is a small sample. “Running bad” is real, but it can also be a convenient explanation for a process that needs adjustment. The totals bets are 5-9 with no Over wins and a pattern of cold-weather, early-season misfires that suggests the models informing those picks may have lagged behind actual run environments.

The prop strikeout-over angle is 2-6 with no credible path to profitability — that’s not variance, that’s a bad bet type to keep pursuing.

The massive CLV on several prop bets that all lost is the biggest open question.

Being first to market on an opening line that moves 10-15 points in your direction is only valuable if the results eventually follow. Over a larger sample, they should, if the underlying edge is real. But 63 bets — particularly 15 prop bets — isn’t enough to confirm that.

The bottom line:

The data says this is a winning bettor in a cold stretch.

The bet-type breakdown says there are clear adjustments available right now:

1. Stop betting strikeout overs
2. Be much more selective on game totals where the model shows value on the “over”.
3. Double down on the F5 angle that’s working on both sides of the ledger.

The record looks bad. The process, mostly, does not. That’s a meaningful distinction, and it’s the one worth tracking as the season continues.

Track All My Picks for Free (Code: JUSTBASEBALL)