Spread Scanner

Updated — · horizon — · —

Loading the latest scan…

The scanner finds names whose volatility has compressed — coiled springs that tend to release with an outsized move. The options layer then reads what the market charges for that move, and that is what decides which side of the trade you want to be on:

One trade, placed in the same week every year. Set the week you would buy, how long you would hold it and how far the name has to move — this counts the years that closed past that move against the years that did not, over the same weekly history behind the Charts tab. The verdict is the exit: eight weeks at +1% asks whether the name is 1% up at the close of week eight, not whether it got there and gave it back. Whether it was ever touched is reported beside it, never as it.
The target can be zero or negative. An in-the-money vertical is already past its breakeven at today's price, so its real question is did it hold up rather than did it travel — set −3% and this counts the years that closed no worse than 3% down.
A target is a distance, not a level: $250 meant something very different in 2016, so you set a percentage and each year's target price is worked out from that year's own entry. Every one of them is in the table.

Name

Direction

Buy week — 37

Hold, weeks Target move by exit, % Years to repeat

Loading the weekly history…

Price it as a debit spread — real cash in, cash out, net

Every definition below ships inside data/scan.json alongside the numbers it describes, so an explanation here can never drift from the field it explains.

What the numbers mean

The strategies

How the pipeline works

Backend — data only

Each weekday after the US close a GitHub Action fetches the ETF holdings that define the universe, re-screens them on industry plus debt and cash ratios, downloads daily prices, recomputes the indicators without look-ahead, reads the option chains for the top-ranked names — the front month and a ≈13-month expiry — runs both strategy engines, and writes JSON:

data/scan.json — signals, IV read, recommendations, and this copy
data/charts.json — downsampled price history, and the same closes grouped by calendar month
data/weekly.json — the same history as one row per ISO week, which the Repeat test walks
data/backtest.json · data/calibration.json
data/signals.csv — the flat table, for spreadsheets

Python renders no HTML at all. Deciding what to trade got complicated enough that it needed to be data, not markup.

Frontend — this page

Hand-written, checked into public/, no build step and no external assets. It fetches the JSON above and renders it. Nothing regenerates it, so you can restyle the dashboard without touching the scanner — and read the same data from anything else that speaks JSON.

Run it locally with python run.py then python -m http.server --directory public 8765. Opening the file over file:// will not work: the browser blocks fetch there.