How often did it actually work?
Stop guessing whether your setup has an edge. Edge Stats answers from your own market data: how often it worked, under the exact conditions you trade it, with the sample size and confidence interval attached so you know when to trust the number.
edgestats query · DEMO_STK
stability 88.2% (n=51) vs 70.6% (n=51) · halves agree ✓
minutes to fill median 1 · p25 0 · p75 16 · p90 100
per-year 2023 88.0% (50) · 2024 71.2% (52)
Historical conditional frequencies with sample sizes. Not predictions, not advice.
+ one more condition
⚠ LOW SAMPLE. N < 30; treat as anecdote, not edge
stability 57.1% (n=7) vs 85.7% (n=7) · halves agree ✓
minutes to fill median 25.5 · p25 10 · p75 63.5 · p90 109.7
per-year 2023 66.7% (6) · 2024 75.0% (8)
Historical conditional frequencies with sample sizes. Not predictions, not advice.
Real output · engine 0.1.0 · store 79763594fe94 · calendars 427e1cdc5046 · deterministic demo store, synthetic bars
Know if your setup works, down to the weekday.
Does the gap fill more often on a Monday? Pick a setup, pick a ticker, and the report answers with real numbers and real sample sizes. Gap fills, opening range breaks, inside days, and everything between.
run on the bundled demo store · every bar carries its sample size
Every classic report, ready on day one.
Gap fills, opening range breaks, inside days, session behavior: the reports data-driven traders live by ship as 42 presets across 11 categories. Every preset cites the LuxAlgo Library indicator that draws the same setup on your chart, so the statistic and the visual come from one playbook.
.png&w=3840&q=75)
Let your AI run the analysis.
Nine read-only tools let Claude, Cursor, or any AI assistant run any report or custom question against your data mid-conversation, and show the sessions behind the answer. Analysis that used to be a spreadsheet weekend becomes one message. Local today; the same tool shapes are headed for the hosted LuxAlgo MCP.
edge_query
Any composed P(outcome given conditions), in the full envelope.
edge_fields
The registry: every outcome, predicate, and field, with definitions.
edge_sessions
The historical sessions behind a result.
edge_report
A preset from the catalog, with parameters.
edge_reports_list
The catalog, with parameter specs.
edge_freshness
Configured symbols, last bar per symbol, calendar versions.
edge_export
CSV or parquet written locally, path returned.
edge_trades
Which trade tags your imported trades produced, and how to query them.
edge_live
Live Board state: forming, active, and resolved setups.
Your bars, from wherever they live.
Seven adapters, two of them free and keyless. Daily canaries pull a sample from every vendor and fail CI on schema drift.
Broker sync and fill imports run through the open LuxAlgo Broker SDK, read-only by design
No naked percentages
There is no code path that prints a percentage without its sample size.
Intervals, always
Every estimate carries N and a Wilson 95% confidence interval.
Guarded small samples
A warning below 30 sessions. No estimate at all below 10.
Stability, split
First half versus second half, plus a recency view over the last 250 sessions.
Any condition you can think of.
By ticker, by weekday, by session, by gap size, on event days or off them. Any condition you actually trade can be asked directly, and combined freely with the others, so the question you ask is exactly the trade you take.
$gapFill WHERE dayOfWeek = Tue AND gapPct BETWEEN 0.05% AND 0.6%
$orbBreak(15m, up) WHERE gapUp AND NOT eventDay('CPI')
$eventOccurs('TRADED_WIN') WHERE eventDay('TRADED') AND prevNr7
$gapFill WHERE dayOfWeek = Tue AND NOT eventDay('FOMC')
Answers land in about 20 milliseconds over years of 1-minute bars (the architecture doc’s own measurement), because the heavy work happens once when your data syncs, not every time you ask.
Prove it on the demo store first.
One clone, one install, and the deterministic demo store answers in seconds: ~900k synthetic bars with the full envelope on every query. Point an adapter at real data when you believe the machinery.
Distributions, not promises.
Every answer carries its N.
No order execution, ever. No hosted service, no telemetry, no accounts. Just historical conditional frequencies with sample sizes, computed on your machine.
MIT code · CC BY 4.0 bundled data · © LuxAlgo Global, LLC · contribute
Every Edge Stats result carries the same fixed line: historical conditional frequencies with sample sizes, not predictions and not advice. Past frequencies do not guarantee future results.