What is a dataset?

A table with one row per day and one column per thing you are tracking. Columns are either computed from daily bars or filled in by you.

There are three kinds, and the kind is fixed when the dataset is made.

• Daily logOne row per session. This is what nearly everything here is. A daily log can hold computed columns, which is what lets it backfill years of history in one go.
• Event logRows at any time, each stamped with a date and a time. For things that happen several times a session, or not at all: a failed breakout, a V-reversal. An event log cannot hold computed columns, because computing them means one row per date and an event log does not promise that.
• ImportedWhatever came out of your spreadsheet. The rows are yours; the app adds nothing to them.

Every dataset also carries an instrument — NQ, ES, ^VIX — unless it is an event log. The instrument is what the computed columns fetch bars for. Without one, a dataset can still hold everything you type into it, but nothing will be worked out for you.

The label under a dataset's name on the desk tells you which of these it is, in one word: computed, manual, imported, or journal.