Methodology: How Subsecond Detects Movers

What the scanner watches

Subsecond evaluates 5,000+ US-listed tickers once every second, from 4 AM to 8 PM ET, across all three sessions: pre-market, regular hours and after-hours. Every metric is computed per session, so a pre-market reading is compared against pre-market history rather than against the whole day.

Five rolling windows run at the same time: 10 seconds, 30 seconds, 1 minute, 5 minutes and 15 minutes. Around 50 metrics are filterable, in families covering price and change, volume and dollar volume, gap percent, relative volume, and share structure such as float, short interest and days to cover.

How relative volume is computed

Relative volume compares the volume a ticker has actually traded in a window against the volume it would normally have traded in the same window. The baseline is built from 50 prior sessions, stored in 10-second bins, and kept separately for each session. A rolling window's expected volume is assembled from the same clock-time bins on those prior days, with the bins at the edges of the window pro-rated so a window that starts mid-bin is not over-counted.

A noise floor suppresses the reading until a ticker has at least five sessions of history, so a new listing does not print a fake ratio on its first morning.

What "flagged at 6:14 AM" means in an article

It is the wall-clock time, in Eastern Time, at which the scanner's rule first matched for that ticker. It records when a condition became true. It is not a trade signal, not an entry, and not a claim that the move continued after that moment.

Data

Subsecond reads the full consolidated US equities tape. Quotes and prints can be delayed or revised by the exchange after the fact, and halts, auctions and outages all affect what a window sees. A reading is a description of that tape at the instant it was taken, not a settled record.

Limitations

A volume spike is not a catalyst. It says people are trading, not why, and not in which direction. Low-float names produce extreme ratios on genuinely small volume, so a large multiple on a few tens of thousands of shares is arithmetic rather than significance. New listings have no baseline for their first five sessions. Nothing the scanner reports predicts direction.