FILTRIX / Lesson 02 of 4
How to Use Context Analysis
Build a FILTRIX context, set the close-position band, understand the historical sample and result buckets, then save the full sample for execution testing.
Watch the walkthrough
Context Analysis video walkthrough
Read the full video transcript
Welcome to Filtrix. In this walkthrough, you will build a Context Analysis, read its historical results, and save the sample for Execution Analysis. Start with one question: when a clearly defined setup appeared before, where did those days close? Context describes historical outcomes. It does not simulate your entry or exit.
Open Context in the navigation. Use Add Filter, then choose a column, an operator, and a value. For a first exercise, use one familiar symbol, or define a broader pattern with price, gap, and volume conditions. Keep the setup easy to explain. For an intraday test, use only information that would have been available before your intended entry.
In Analysis Parameters, choose a date range and the In-range Band. Close position is the close minus the day's low, divided by the day's high minus its low. Zero is the low; one is the high. With a band from zero point five to zero point eight, closes between fifty and eighty percent of the daily range are in range. This is not a fifty to eighty percent return. Select Run Analysis.
Read the sample counts before the percentages. Total found in history counts matching alert records. Included in calculation, n, counts usable symbol and date examples after duplicate symbol-days and unusable outcomes are removed. A difference between these numbers deserves attention. A large n alone does not prove an edge: check whether a few symbols, dates, or market conditions dominate the sample.
Above range means the close finished above the upper boundary. Below range means it finished below the lower boundary. In range includes the boundaries. Within each bucket, Up, Down, and Equal compare the regular close with the regular open. Those percentages describe that bucket, not the whole sample. Above range is a location in the daily range, not a trade win rate.
Next, inspect the Preview rows. Check the dates, symbols, opening and closing prices, and percentage changes. Look for repeated names, unusual observations, and conditions that do not match your original question. Use another date period to check whether the pattern holds up. Changing filters until one historical result looks attractive can overfit the sample.
Give the context a descriptive name and select Save Analysis. For an execution test, keep the save mode on All. Selecting only Above range or Below range uses the day's eventual result to select the sample, which can bias an intraday backtest. The export selector is separate from the save selector. Keep a record of the filters and dates used for each comparison.
You now have a defined sample and a clear description of its historical closes. Use Execution Analysis to ask the next question: how would a specific entry and exit rule have behaved on these symbol-days? Open the next tutorial to follow that workflow. Historical patterns are a starting point for research, not a promise about the next trade.
What You Will Learn
- Build a repeatable setup from filters and a date range.
- Understand matching alert records versus usable symbol-days (n).
- Read the close-position band and Above / In / Below range buckets.
- Inspect the rows and save an unbiased sample for Execution Analysis.
Before You Start
- A FILTRIX account with available analysis quota.
- One setup question and the conditions known before your intended entry.
- A date period for research and another period for validation.
Define the setup with filters
Open Context from the navigation. Each filter has a column, an operator, and a value. Begin with a familiar symbol or a small set of conditions such as price, gap, and volume. For an intraday backtest, a filter should describe information that was available before the intended entry. An end-of-day metric is not a valid premarket input.
Choose dates and the In-range Band
In Analysis Parameters, set Start Date and End Date, or use the date shortcuts. Close position = (Close − Low) / (High − Low): 0 is the daily low and 1 is the daily high. A band of 0.50–0.80 classifies closes in that part of the daily range as in range. It is not a 50–80% return or a stop/target setting.
Read the sample counts first
Total found in history counts matching alert records. Included in calculation (n) counts usable, distinct symbol-date examples after duplicate symbol-days and unusable outcomes are removed. The counts can differ. A large sample can still be concentrated in a few names or market regimes; there is no universal sample-size cutoff that proves an edge.
Interpret buckets without calling them win rates
Above range means close position is greater than the upper boundary. Below range means it is less than the lower boundary. In range includes both boundaries. Within each bucket, Up, Down, and Equal compare the regular close to the regular open; their percentages use that bucket as the denominator. None of these buckets measures the profit of a specific entry and exit.
Inspect the Preview and validate on another period
Page through the underlying rows to check symbols, dates, opens, closes, and percentage changes. Investigate unusual observations and repeated names. Keep a separate date period to test whether the pattern persists. Repeatedly adjusting filters or the band to improve one historical result can overfit the data.
Save the full sample for execution testing
Give the context a descriptive name. Keep the save selector on All for an execution test. Saving only Above range or Below range conditions the sample on the eventual daily close and can bias an intraday simulation. The Export Data selector is separate from the save selector; exporting a subset does not change the saved context.
Your turn / about 5 minutes
Explain one result in your own words
Keep a brief note of your inputs and what you observe.
Open FILTRIX in a new tab ↗- Run the AAPL 2025 example with a 0.50–0.80 band, or use a clearly defined setup of your own.
- Record Total found, n, the band, and the three bucket counts. Check a few underlying dates and symbols.
- Write one sentence describing the distribution without calling it a trade win rate. Save All if you will test execution.
You're done whenYour notes identify the sample, the denominator, and what the buckets do not measure.
Pause before moving on
Check your understanding
Answer in your own words, then open each question to compare.
With a 0.50–0.80 band, where does a close position of 0.80 belong?
In range. Both boundaries are included. Above range starts above 0.80.
Is 60% Above range the same as a 60% trade win rate?
No. It describes where daily closes fell inside their ranges. A trade win rate needs explicit entry and exit rules.
Which rows should you save for an intraday execution test?
All. Choosing only Above or Below selects examples using the eventual daily close. Also check that the filter inputs were known before the simulated entry.
Get unstuck
If something looks different
The result is empty or much smaller than expected.
Check column names, units, operators, and dates first. All filter rows must match. Remove one condition at a time to diagnose the query, and record any change to the hypothesis.
Total found is larger than n.
Total found counts matching alert records. The calculation uses distinct symbol-days with usable outcomes, so duplicate alerts and unusable data can reduce n.
Save Analysis is unavailable.
Run the analysis successfully and enter a name. If the app reports a quota or query error, resolve that message first.
Common Mistakes to Avoid
- Calling Above range a trade win rate.
- Using future end-of-day information to select an intraday sample.
- Assuming Total found and n measure the same thing.
- Treating a large sample or a high percentage as proof of profitability.
- Changing filters repeatedly without validation on another date period.
Related lessons
How to Run Execution Analysis
Test entry and exit rules on a saved FILTRIX context. Follow the minute-data pipeline, interpret per-share PnL, and inspect individual trade charts.
How to Validate a Trade-Ideas Alert
Translate a Trade-Ideas alert into measurable FILTRIX filters, inspect the historical sample, and test a defined execution rule.
Ready to test the setup?
Use the practice task to record your inputs, inspect the sample, and explain the result.
Go to the practice task ↑