Crypto only
Spot and perpetuals on the venues the platform reads. The equities connector is paused over a symbol-format problem and the screener does not offer it.
Buying guide
Every screenerlists indicators.None of themmatter most
Screeners are compared on indicator counts because that is the easy number. The five properties below decide whether the screen you build is trustworthy, and most product pages do not mention any of them.
The five
1. Does it evaluate on closed candles?
An indicator on a forming bar changes every tick and can cross your threshold and cross back before the bar closes. A screen reading it fires on a value that never existed at the close, and will not reproduce in a backtest. Ask, because it is usually not stated — and it was a real defect here before closed-only became the default.
2. Does it keep every firing?
Without a history you judge the screen by the hit you remember. With one you can ask how often it fires, on what, in which regime, and what happened next. Firings should also be idempotent: a condition true for six hours is one event, not six hours of alerts.
3. Does it distinguish "nothing found" from "could not look"?
If market data stalls, a screen that returns zero results looks exactly like a quiet market. A degraded run must say so. This is the difference between a screener you can leave running and one you have to babysit.
4. Is the universe per venue, from the venue?
A symbol on two exchanges is two series with different liquidity and funding. Worse, a hand-typed market that does not exist scans zero symbols and reports a clean run — we lost a real run that way, which is why venue, market type and quote now come from the instrument registry rather than from typing.
5. What can a hit become?
A screener that only draws on a chart makes you the whole pipeline. A hit here can start an assistant turn with the rows already in context, become a backtest, or feed a watch that waits for an entry condition — and none of those places an order without you.
What we would still ask about this one
The honest limitations, in the same place as the argument.
About two years of history
A real ceiling on any condition needing a long look-back, and on any backtest of what the screen finds.
Fewer venues than a bot platform
Venues are added by certification rather than by checkbox, and each page says which verb applies.
No social or on-chain conditions
Those feeds have no vendor behind them here, so screening on them is not offered rather than offered thinly.
Questions about choosing a screener
How many indicators do I actually need?
Fewer than any product page implies. A trend filter, a momentum measure, a volatility measure and a rank across the universe will express most screens people actually run.
Should the AI write my screens?
It is a good drafter and a poor decider. Have it produce the specification, read the specification, then save it — which is the flow here, because the failures in a screen spec are silent.
How do I tell if my screen is any good?
The firing history plus a backtest of what it found. A screen that fires forty times a day is noise regardless of the one hit you remember.