You can't watch every game. You don't have to.

A pitcher loses two miles an hour in the sixth. Ninety-four seconds later the book has fixed the price. Your phone buzzes at second eleven.

9:04 PM
Diamond Labsnow
SPIN_VS_AVG fired — CLE @ TB, Over 6.5
Fired at 54¢ · SPIN_VS_AVG +7.8% over 241 settled
Diamond Labs10:43 PM
TRAILING_LATE fired — ATL @ LAD, Under 6.5
Fired at 46¢ · TRAILING_LATE +10.8% over 344 settled
Diamond Labs9:47 PM
VELO_DROP fired — SD @ MIL, Over 9.5
Fired at 52¢ · VELO_DROP +9.9% over 201 settled
Real alerts, Wednesday, October 14.

Ninety-four seconds. You spend eleven.

You're at dinner. The phone is face down. Here is what happens without you.

9:14:02

The starter loses 2.1 mph off his fastball.

Top of the sixth. Nobody in the building reacts. The broadcast doesn't mention it.

9:14:05

The rest of the game is simulated again from scratch.

Not a line nudge. A new distribution, built off the new game state.

9:14:07

The number moves. The book hasn't.

That gap is the only reason to interrupt you. Most gaps this size are our mistake, not theirs.

9:14:09

So is anything actually behind it?

VELO_DROP matches, and VELO_DROP has a published forward record. If nothing had matched, your phone stays dark and you never find out this happened.

9:14:13

Your phone buzzes on the table.

Eleven seconds after the pitch. One line: the game, the side, the price, the signal.

9:15:02

You turn it over, read it, decide.

Or you don't, and you put it back down. Nothing here tells you what to do with your money.

9:15:36

The book re-hangs. Window shut.

Ninety-four seconds, start to finish. This is why nobody catches these by watching.

Every alert tells you who to blame.

An alert without provenance is a tip. Here is the difference, in one screen.

DIAMOND LABS  ·  9:04 PM
SPIN_VS_AVG fired
CLE @ TB  —  Over 6.5
FIRED AT
54¢
INNING
3
RECORD
+7.8%
SPIN_VS_AVG · +7.8% across 241 settled bets — open the record

A market, not a lean

Exact game, exact side, exact number. Nothing to interpret.

The price it fired at

Timestamped. Afterwards you can check whether you got our number or a worse one.

One named signal

Not “the model likes it.” A specific rule with a specific history, and you can open that history from the alert.

And a way to fire us

If TRAILING_LATE stops paying, we pull it and you stop getting the alert. That rule is written down before the fact, not decided afterwards.

4 of our 19 signals clear the bar to send alerts. We publish the 15 that don't.

Every signal, every interval, the raw file

You were never going to be watching.

The edges don't arrive when it's convenient. That's most of why they're still there.

Saturday, 8:40 pm

Dinner with your wife. Phone face down. It buzzes once between courses, you read one line, it goes back down. Ten seconds, not an evening in front of a screen.

Tuesday, 1:04 am

A west-coast game is in the eighth and you're asleep. The best prices all night are the ones nobody is awake for. In the morning it's still there, with the price it fired at, so you can see exactly what you missed.

Thursday, all night

1 games. Nothing clears. Your phone doesn't buzz once — and that's the system working. Most nights are quiet nights.

Most nights, nothing.

An alert goes out when a graded signal fires and the gap survives a stress test. That's it. We'd rather send you nothing than send you noise.

1
alerts on Wednesday, October 14
1
games watched
4
signals cleared to fire

Where it fires, and where it doesn't.

A system that finds an edge in every market is a system with a bug.

Baseballfiring · 4 signals

a team down one to three in the seventh or later, facing the other side's best arms; the starter's velocity falls off his own season average mid-outing; spin rate dropping below the pitcher's own baseline; the lineup coming around on the starter a third time.

Basketballtested · nothing cleared

Our error is worse than the market's in both directions. No alerts until that changes.

Footballtested · nothing cleared

Hundreds of hypotheses, zero survivors once you correct for having run hundreds of them.

Event marketswired · not yet tested

Settled contracts are in the index as a price reference. When something clears, it starts firing here too.

Invite only, and here's the real reason.

The edges are thin. They don't survive everyone taking them at once. The research stays free either way.