Setup Odds. How often did this setup actually work?
Free · Open source · MITThe session opened above the prior close by at least the threshold; did price trade back to the prior close by 12:00?
Setupgap >= %Outcometouches prior.close by 12:00
39.6%
of 53 sessions saw the outcome. 95% range: 27.6% to 53.1%.
- Every measured session
- 64.1% of 515 (59.8% to 68.1%)
- Lift
- -24.5 points, 0.62 times the baseline
Example: counted on the tool's demo bars, which are made up for testing, not market data. Load your own bars to get a real count.
Did it hold up over time?
Each year
When it happened
median10:09
half of them between9:50 and 11:03
90% by11:17
Warnings
- Sessions that could not be measured: 2 (1 needs bars from before the file starts or after it ends; 1 the session closed early, before regular hours).
demo seed 1: 530,497 bars, 2023-01-03 to 2024-12-31, 515 sessions measured.
No bare percentages. Every number shows what it’s based on.
- The rate
- Sessions where it happened, out of the sessions where it could have.
- The 95% range
- A Wilson interval, which stays honest with few sessions and close to 0% or 100%.
- First half, second half
- The matching sessions split by date, each with its own range.
- Each year
- A count for every year, so you can see if one good year carries the number.
- When it happened
- The time of day: the median, the middle half and the 90th percentile.
- The clock
- Every session follows the New York exchange calendar, holidays and early closes included.
How often isn't an edge.
It counts what happened. It doesn't place trades, pay costs or use stops.
Ask enough questions and luck answers one.
A range holds for one question asked once. Try twenty variations and one will look good by chance.
Sessions aren't independent.
Calm days and wild days come in streaks, so the real uncertainty is wider than the range shows.
The engine is open. Run the same count yourself.
The tool on this page is the open source engine, running in your browser. Install the same engine as a package to run it on your own files, and read its setup language and every preset in the repository.
A frequency is a question. A backtest answers it.
“Test this setup as a strategy” opens AlgoLift with your setup written out for Algo, our copilot, to build into a strategy. You read the request and send it yourself. Building and backtesting needs an AlgoLift account; this tool doesn’t.
Do my bars leave my computer?
No. The tool reads your file in this tab and does all the counting there.
What bars can I use?
One-minute bars in a CSV file with a header: a time (or a date and a time), open, high, low and close, and volume if you have it. A time can carry its own offset, be read as New York, Chicago or UTC time, or be epoch seconds. Sessions follow the CME equity-index futures calendar on the New York clock, 2018 through 2027.
Do I need an account?
No. You only need one to test the setup as a strategy in AlgoLift.