MLB Backtest May 2026 — The Strategy LOST MONEY

vs blind favorite betting

Strategy	Record	Staked	Net	ROI
My analysis	7-6	$89	-$10.40	-11.7%
Blind favorite	10-10	$100	-$11.50	-11.5%

Strategy slightly outperformed (less lost) but both negative. Without sport-specific data (starting pitchers especially), MLB betting is approximately random.

2. The 6 losses analyzed

Loss 1 · LAD vs SF Game 1

Pick: LAD ML at -180 (STRONG, $8) · Logic: LAD massively better record (26-18 vs 18-26)

Actual: SF won 6-2 (upset)

What happened: Daily pitching matchup probably favored SF starter.

Lesson: Record gap doesn't capture daily pitcher quality.

Loss 2 · BAL vs NYY

Pick: NYY ML at -150 (STRONG, $8) · Logic: NYY better record (27-17 vs 20-24)

Actual: BAL won 7-0 (big upset, shutout)

What happened: BAL pitcher dominated, NYY had a bad offensive day.

Lesson: 1-game shutouts happen in MLB regardless of records.

Loss 3 · PIT vs COL Game 1

Pick: PIT ML at -170 (STRONG, $8) · Logic: PIT better record (24-20 vs 17-27) + home

Actual: COL won 10-4 (upset blowout)

What happened: PIT pitching collapsed, COL hit.

Lesson: Even home favorites with 7-game record edge can lose badly.

Loss 4 · TOR vs TB

Pick: TB ML at -160 (STRONG, $8) · Logic: TB elite record (28-14 vs 19-24)

Actual: TOR won 5-3

What happened: Home underdog with bullpen advantage flipped it.

Lesson: TB on road, possibly traveling, lost what should have been winnable.

Loss 5 · HOU vs SEA Game 2

Pick: SEA ML at -125 (MEDIUM, $5) · Logic: SEA still better team after winning G1

Actual: HOU won 4-3 (bounce-back)

What happened: HOU starter pitched better, bounce-back factor real.

Lesson: Bounce-back after blowout might be a real thing.

Loss 6 · ATH vs STL Game 2

Pick: STL ML at -115 (MEDIUM, $5) · Logic: Same as G1 (STL better team)

Actual: ATH won 6-2 (big home win)

What happened: ATH starter dealt, STL bats quiet.

Lesson: Same matchup, different result — daily variance dominates.

3. What worked (the 7 wins)

Pattern in wins: talent gap + situational alignment. Pattern in losses: talent gap alone wasn't enough.

4. The 7 skipped games — was skipping correct?

Matchup	Result	Skip verdict
TEX vs AZ G1 (both ~.500)	TEX won 7-4	Would have won — 1 missed
CIN vs WSH (close)	WSH won 8-7 (1-run upset)	Skip correct
BOS vs PHI	BOS won 3-1 (upset)	Skip correct
NYM vs DET (close)	NYM won 3-2 (1-run)	Skip correct
MIN vs MIA (identical records)	MIA won 9-5 (upset)	Skip correct
MIL vs SD (close)	SD won 3-1	Skip safe
TEX vs AZ G2 (close)	TEX won 6-5 (1-run)	Skip safe

Verdict: 5/7 skip decisions avoided actual losses or coin flips. Only 1 case (TEX) would have been a winning bet. 35% skip rate was correct — the strategy worked here.

5. Five new insights — the actual learning

1 · Confidence calibration is broken

STRONG bets: 4-4 (-19.4% ROI). MEDIUM bets: 3-2 (+8% ROI). When I think I'm most certain, I'm probably wrong.

FIX: Lower stake on "strong" picks until calibration improves. Possibly invert: $3 on STRONG, $5 on MEDIUM.

2 · MLB requires starting pitcher data

Season records do NOT capture daily pitcher quality. Need: ERA, recent form, vs-team history, home/away splits.

FIX: Without starting pitcher data, MLB betting is essentially random. Build that pipeline before next MLB bets.

3 · Bounce-back factor might be real

Both LAD and PIT bounced back from upset losses by winning big. Sample is 2/2 — way too small but worth tracking.

FIX: Track every "bounce-back" scenario going forward. Validate or invalidate after 20+ samples.

4 · Heavy favorites in MLB are dangerous

All -150 to -200 picks lost in this sample. A -180 favorite needs to win 64% to break even — too high a bar for daily MLB.

FIX: Skip MLB favorites with juice over -150 unless starting pitcher advantage is clear.

5 · Sample size > strategy quality

20 games is statistical noise. Need 100+ to validate or invalidate any approach.

FIX: Track all bets continuously. Don't conclude from short samples.

6. Honest answer to "is my analysis worth anything?"

NBA playoffs (20 games): +39% ROI vs blind favorites +44%. Net edge: zero.

MLB regular season (20 games): -11.7% ROI vs blind favorites -11.5%. Net edge: zero (slightly worse).

Conclusion: After 40 games tested, my analysis has NOT demonstrated edge over "bet the favorite." This isn't surprising — beating the market is HARD.

What this DOES NOT mean: that we stop tracking and analyzing.
What this DOES mean: be honest about edge. Don't bet what you can't afford to lose. Treat this as entertainment + learning project, not income source.

The framework's REAL value:

We now know confidence is mis-calibrated
We know MLB needs pitching data
We know narrative factors hurt more than help
We know what to look for as we expand the data

Over 200-300 games tracked the system may identify real edges. Over 20 games we identified our weaknesses. That's still progress.

7. Action items going forward

Add starting pitcher data to all MLB analyses — need a data source
Lower stakes on STRONG picks until calibration improves
Skip more coin-flip games — push toward 50%+ skip rate in MLB
Track bounce-back scenarios specifically (potential real edge)
Don't expand to other sports until we beat the market in one
Update /bets/methodology/ with these findings
Hard weekly bet limit — $25-50/week for analysis testing, treated as learning expense

📉 MLB Backtest — May 12-14, 2026 (20 games)

Strategy v1 lost money in MLB

1. Top-line results