METHOD

What Jodie measures — and how well

Jodie joins two measurements no one else keeps side by side: how thousands of names are trading together right now, and what their filings disclose about each other. It does not tell anyone what to buy, sell, or expect from a price. This page is the ledger: what it measures, and the numbers we hold ourselves to.

The work

01Build the live map. Jodie strips out the broad market move, then watches for names whose leftover movement locks together — separating established groups from ones still forming. This map answers: is this real, right now?
02Build the structural map. Continuously, before anything moves, Jodie reads the filings: disclosed customer, supplier, and competitor relationships, and annual reports that read alike. Every link is kept with the excerpt it came from. This map answers: who is connected to whom, and why?
03Join them. Every live event is checked against the structural map. When a moving group crosses a disclosed relationship with a name you follow, that is the alert — movement you can see, a connection you can verify. Neither map alone would have told you.
04Flag concentrating activity. When trading in a single name pulls well clear of its own norm, Jodie marks it — a heads-up that a move may be near.

What we have measured

Each map is tested on its own, and so is the join. Here are the numbers, with sample sizes and the base rates they beat.

52%vs 30% before

THE LIVE MAP

Groups that are real

Names in a confirmed group kept moving together in the sessions after we flagged them — well above the previous method and above chance for a group this tight.

n = 50 confirmed episodes, same test window
99.5%

THE STRUCTURAL MAP

Relationships we can prove

Disclosed customer, supplier, and competitor links, each checked back against the exact filing it was drawn from.

LLM-extracted, verified against source filings
42.9%vs 30% base rate

THE STRUCTURAL MAP

Filings that read alike, move alike

When two companies' annual reports use strikingly similar language, they went on to move together more often than the base rate for any two names.

p = 0.0015
+0.04695% CI [+0.017, +0.075]

THE JOIN

The read-through is real

Names linked by filings to a group that is lighting up co-moved with it beyond what their shared sector alone would explain — the measurement behind “this touches something you follow.”

sector-controlled, interval excludes zero
1.3×range 1.25–1.33×

ACTIVITY

Concentrated activity precedes movement

When Jodie flags concentrating activity in a name, realized movement over the next one to three days ran about a third higher than normal — in size, not direction.

n = 14,767 observations

What we do not claim

We do not call direction. Jodie measures that activity is concentrating, not which way a price will break. On direction our reads are no better than a coin flip, so we never make them.

We are not faster than the news. At the level of the single name that moves first, we are usually behind the headline. Jodie's job is to connect a move across a whole group and its filings — not to break the story.

We ease off under stress. For about a week after a volatility or credit shock, group behavior gets less reliable. In that window Jodie marks itself less sure and shows fewer calls, not more.

What Jodie refuses to do

No forecasts. No buy or sell calls. No price targets. No track records or leaderboards. Concentrated activity is context, never a recommendation.