Data & concepts
Z-scores, the economic floor, freshness, and what a tool touches.
A short tour of the ideas the tools assume. None of these are settings you configure: they tell you how to read what comes back.
Vocabulary
Eight terms carry the domain:
- Screen: a predicate/filter over the universe. Run it live with
screen_events. - View: a saved Screen that runs on a schedule. Create one with
create_view, list yours withlist_views. - Match: one event entering a View's screen. The feed of these is
list_matches. - Alert: one delivery that batches your Matches: an email, or a post to an alert destination (a Slack or Discord channel, or a URL you set). Delivery, not a tool.
- Digest: the other delivery of a View: its whole table sent at an hour the owner
picks. Set in the app, not by a tool. Not the same as the
batchedemail mode. - Issue: one sent Digest, with Movers, New and Gone counted against the previous Issue. See Read an Issue.
- Price: the seller's price before marketplace fees. Columns and fields without
aipin the name carry it:ea.lowest_price_currentis the lowest price. - All-in price: what the buyer pays, fees included. Columns and fields with
aipin the name carry it:ea.lowest_aip_price_currentis the lowest all-in price. Where a marketplace states no all-in price the field is null; it is never worked out from the price.
The tools create_alert, list_alerts and list_firings no longer exist: use
create_view, list_views and list_matches.
Analytics snapshots
Most tools return rows from three joined tables:
event_analytics(ea.): the heart of it, with self-relative z-scores (price, pace, volatility regime, inventory velocity), cross-sectionalpct_rankand robust MAD-z columns, quartile primitives, face-value rollups, and adata_quality_score.events(e.): identity and scheduling, with name, venue, local date, status.performer_analytics(pa.): per-performer signals, social and performer-average price z-scores.
Two flavors of z-score
- Self-relative (e.g.
ea.price_self_z_7d): how unusual this event is against its own recent history. Good for "is this moving for it?" - Cross-sectional (the
_xs_z_subcat_tte/_pct_rank_subcat_ttefamily): how this event ranks against its peer set (same sub-category, similar time-to-event). Good for "is this a standout right now?" These ranking primitives are the fast columns tosorton.
The economic floor
By default the server drops events with very few or very cheap listings so noisy minor-league inventory doesn't top the rankings.
See the predicate DSL. Disable
with apply_floor=false when you explicitly want the raw universe.
Freshness
Rows aren't all recomputed at the same time. screen_events keeps the events priced within 3
days of the newest price date in the market by default (exclude_stale=true). When
"is it moving now" needs a tighter window, gate on ea.last_price_snapshot_date with a
relative { "days_ago": N }.
data_quality_score
A per-row score from 0 to 1: higher means more history behind the signals. Filter on it directly.
What a tool touches
Every connection is signed in, so no tool is open to the world (see Connect). What differs is the subject: most tools read the shared universe, and the Views, Matches, destinations and watchlist tools read and write things you own. The tools reference is grouped that way, and each tool's own page in the app carries its access annotations.