⚾ MLB Bad Bets — The Five-Unit Lesson
We put the biggest tag in this column's history on the Rockies. They lost by five. Here's the honest post-mortem — and why the model is walking right back into Coors today.
⚡ The 30-Second Read
Yesterday we told you the model wanted 5.00 units on Colorado — the heaviest tag this column has ever posted. We also told you, in the same breath, that it would make you feel insane.
It lost. CIN 7, COL 2. Not close, not unlucky — the Reds put a seven-spot on the board at Coors and the model’s loudest number of the season got its doors blown off. By the 5PM re-run the edge had ballooned to +21.1%, which only made the miss louder.
We’re not going to bury that. It’s the lead. When you post a five-unit lock and it loses by five, you owe the room a straight accounting, not a subject change. That’s the whole reason this column grades in public.
The rest of the day held it together: the model went 7-7 with PIT@CLE postponed, because Boston cashed twice (including a 10-0 game-two beatdown), Atlanta rolled 15-1, and MIN and STL did their jobs. A 5-unit loss on a 7-7 day is exactly why we size at fractional Kelly instead of betting the label — one bad lock doesn’t torch the week.
Ledger: 359-283. 55.9%. +3.5 points above the line. The cushion took a real bite — the COL unit did that — but it’s green fifty-two days in.
And here’s the part that’ll test your patience: the model is right back in Coors today, on Chicago at 5.00 units again. Same building, same tag, different team. More below, because if we’re going to take that swing we’re going to explain it.
YESTERDAY 7-7 (PIT@CLE ppd)
THE LOCK COL 5.00u ✗ lost 7-2 at Coors
AGREE 4-3
FLIP 3-4
BIG WINS BOS ✓✓ · ATL 15-1 · MIN ✓ · STL ✓
LEDGER 359-283 55.9% +3.5pp (thru Jul 11 base)The board doesn’t get to skip the days it looks stupid. First pitch 1:10.
🎬 Yesterday: The Post-Mortem
Let’s do the COL loss properly, because a five-unit tag earns a real autopsy.
What the model saw: Colorado at home, Coors Field, park factor 112, 91 degrees, against a Reds team sitting 43-52 and last in the NL Central. The model had the Rockies at 64.4% to a market price of 51.2% — a genuinely enormous gap. On paper it was the cleanest edge on the board.
What actually happened: Cincinnati scored seven. That’s it. That’s the entire failure. The model priced Colorado’s offense into the altitude and weather correctly and got the run prevention catastrophically wrong — which is the exact risk of backing a 39-59 team with the worst ERA in baseball no matter how friendly the park is. We flagged that risk in yesterday’s column, in writing, and took the swing anyway. The swing missed.
What we’re not going to say: that it was variance. A 64% pick losing is normal — it happens 36% of the time. But losing 7-2, getting outscored by five, is not a coin landing wrong. That’s a projection that was wrong about how many runs the Reds could hang on a bad Rockies staff. The size made the miss expensive, but the size wasn’t the mistake — the read was. Own the read.
The day survived because the rest of the board was excellent. Boston went 2-for-2 — the +10.4% morning play cashed and the nightcap won 10-0 behind Rivera. Atlanta buried Texas 15-1 as an AGREE +4.7%. Minnesota flipped onto the +121 Cubs and cashed 5-2. St. Louis flipped onto Arizona and won 5-4. Add WSH’s ugly 23-4 loss and SEA and NYY dropping their flips, and it nets to 7-7 with the Pirates postponed.
A 7-7 day with a five-unit loss inside it and the ledger only moves three-tenths of a point. That’s the entire argument for fractional Kelly.
📋 The Full Tape
Model: 7-7 (PIT@CLE postponed; TB@BOS doubleheader). *Re-run split.
🔬 Methodology Audit
The COL loss is the headline, but the quieter story is that the model’s biggest edges were a coin flip yesterday and its medium ones carried. The two loudest plays — COL +21.1% and STL +12.9% — split, one blown out and one cashed. The +8-to-+11% tier went 3-1 (MIN, both BOS, only NYY missed). When the model is very sure, respect the variance. When it’s moderately sure, that’s the sweet spot — same lesson the first half kept teaching.
WSH is worth a note: graded a +2.2% morning play, the Nats hung 23 runs on Oakland. That’s not a model win to be proud of — a coin-flip edge that happened to land in a demolition. We’ll take the checkmark and not pretend we saw a 23-spot coming.
📒 Running Ledger
DAY 52 359-283 55.9%
BREAKEVEN 52.4%
EDGE +3.5pp
GRADED PICKS 642The COL unit shaved the cushion from +3.7 to +3.5 — the biggest single-pick bite of the run, which is exactly what a 5-unit loss is supposed to feel like. Still green.
🎯 Today’s Slate — Saturday, July 18
15 games. Three locks — and the model is walking right back into Coors.
🔒 The LOCKs
WSH +15.0% @ OAK — FLIP Littell vs Ginn (PF 91, 60°F). Top edge on the board and it’s a big one — model has Washington at 61.7% against a market that’s got Oakland at 53.4%, a full 15-point gap. WSH is 4-6 L10 but scoring 6.9 RPG (model +8%), projection 5.3 – 4.1. The model flips off Oakland onto the +105 Nationals. Washington just dropped 23 on the A’s yesterday; the model thinks the bats travel. The bullpen that leads baseball in blown saves is the risk, as always.
CWS +12.0% @ TOR — AGREE · 5.00u Martin vs Bieber (PF 101, dome). Here’s the five-unit tag — and this time it’s an AGREE, not a Coors moonshot. Model has the White Sox at 63.6% to a market price of 51.6%. Projection CWS 5.1 – 3.8, total 8.95. Live ML CWS -117. This is the pick we’ve been making all month — we rode the White Sox three straight boards before the break and they’re now tied for the AL Central lead. The model isn’t chasing yesterday’s loss; it’s backing the team it’s been right about, in a dome, off an AGREE read. Different profile entirely from the COL swing.
PIT +11.3% @ CLE — FLIP Jones vs Williams (PF 98, cross 89°F). Same game as yesterday — it got postponed and reappears today. Model flips off the -126 Guardians onto the +105 Pirates again, PIT 58.0% vs market 53.4%. Pittsburgh’s 6-4 L10 at 5.9 RPG (model +8%), over 7.5 flags at +10.8%. Fourth straight board on the Pirates. The model is stubborn about Pittsburgh and we’re logging whether that’s conviction or a bug.
💪 The STRONGs
ATL +11.1% vs TEX — AGREE Gore vs Murphy (PF 100, 88°F). Atlanta buried Texas 15-1 yesterday and the model reloads — ATL 62.7% vs market 51.6%, projected 5.0 – 3.7. Under 9.0 flags at +10.0%. Live ML ATL -117. A near-lock at an AGREE profile.
MIN +9.1% @ CHC — FLIP Bradley vs Boyd at Wrigley (PF 102, in 10mph 79°F). Model flips onto the +121 Twins again after they cashed here yesterday — MIN 52.4% vs market’s implied CHC 56.7%. But it’s flagged both ways: projected gap only 0.2 runs (size small) and the card notes the flip carries a historically thinner edge. Take it light.
✅ SOLID (4–6%)
NYM +8.1% @ PHI (AGREE) — Manaea vs Luzardo. Kelly 3.31u. Model NYM 53.5%... wait — model has PHI at 53.5%, takes the +149 Mets dog on value against a puffed PHI -181. Book and model on the same side. Actually a STRONG-tier edge landing here on price.
STL +7.5% @ ARI (FLIP) — May vs Pfaadt. Model flips onto STL +100 after cashing this exact matchup yesterday. Under 9.0 flags at a huge +18.7%.
MIA +5.9% @ MIL (AGREE) — Meyer vs Drohan, dome. MIA +113, but tight: projected gap 0.1 runs, size small.
📎 LEAN (4–6% edge, thin structure)
SF +5.1% @ SEA (AGREE) — Webb vs Woo, dome. SF +113, Kelly 2.33u, book and model aligned. Projected dead even 3.7 – 3.7 — a true pick’em the model gives a slight SF lean.
⚪ PASS (<4%)
The bottom of this board is deep and thin — six games the model won’t touch:
NYY +3.9% vs LAD (FLIP) — Sheehan vs Weathers, the marquee game, but a 0.3-run gap and a thin flip. Pass.
KC +3.7% vs SD (FLIP) — 0.2-run gap. Pass.
COL +3.0% vs CIN (FLIP) — yes, Colorado again, the other side of yesterday. Model’s cold 3-7 L10, weather adds +3%. Pass. (Note: 24 hours ago the model wanted 5 units on COL. Today it’s a 3% flip it won’t play. That’s the re-run doing its job — the number moved and the model moved with it.)
BOS +0.9% vs TB (AGREE) — Boston’s on a 10-game win streak (model +8%) but the gap’s a rounding error. Pass.
BAL +0.6% @ HOU (FLIP) — 0.1-run gap. Pass.
DET +0.3% @ LAA (AGREE) — Skubal on the mound, model and book both on Detroit at -201, no value. Pass.
⚠️ Caveats
❌ Not betting advice. Model output.
💰 Kelly: 0.25×, 1u = 1% bankroll. CWS at 5.00u is today’s heaviest — and yes, we’re aware the last 5-unit tag lost by five yesterday. The difference: that was a moonshot on a bad team in a launching park; this is an AGREE on a team we’ve been right about for a month, in a dome. Respect it or skip it.
🌋 One record-tier-adjacent edge today — WSH +15.0% is exactly at the record-tier line. Size it like the grenade it is.
🔁 Flip watch: the model is deeply stubborn on PIT (fourth straight board) and reloading MIN and STL after they cashed. Reloading winners is fine; reloading a pick that keeps not resolving (PIT keeps getting postponed) is worth watching.
🥶 The pens are still lying. We’re one game removed from the break — bullpen-fatigue signal is barely back online. Trust the top of the card.
🆘 1-800-GAMBLER.
🔗 Full board + model JSON in Discord.
🏁 Where We Stand
Day 52. 359-283. 55.9%. +3.5 points above the 52.4% breakeven line.
Yesterday the model made its loudest call of the season and got embarrassed. We put five units on the Rockies, they lost by five, and the only honest thing to do is say so at the top and take the lesson: the read was wrong, not the sizing philosophy. The day still came in 7-7 because Boston and Atlanta carried it, and the ledger barely flinched — which is the entire point of never betting the label like it’s a promise.
Today the model does something that takes nerve: it walks back into the exact building that just beat it, and it asks for five units again — this time on the White Sox, in a dome, on an AGREE read it’s been right about for a month. That’s not stubbornness. That’s a model that doesn’t flinch because one number lost. We grade the process. The process put up 55.9% over fifty-two days, including yesterday. It’s earned the benefit of the doubt.
First pitch: 1:10 ET.
See you tonight in Last Night’s Tape.
— Bad Bets