English · 한국어
This study plans to measure how often BTC price returns after an FVG forms. It is public protocol draft v0.9, not a results report. Venue, exact instrument, sample dates, baseline interval, and execution dataset remain unselected.
The question
Measure first touch, midpoint touch, and full fill within fixed horizons for FVGs detected consistently. Strategy P&L with entries and exits is outside this stage.
Separate detector and protocol versions
Detector v1.0 uses three consecutive completed bars C1, C2, and C3. Bullish: C1 high < C3 low. Bearish: C1 low > C3 high. Equal boundaries do not qualify; no body-size or strength filter is added.
The detector version does not mean the entire study is ready. The protocol remains draft v0.9 until data and quality rules are frozen. Confirmation is C3 close; outcome observation begins with C4.
Boundary measurements
| Outcome | Bullish FVG criterion |
| First touch | A later low reaches or falls below the upper boundary |
| Midpoint touch | A later low reaches or falls below the midpoint |
| Full fill | A later low reaches or falls below the lower boundary |
For bearish FVGs, use later highs reaching or exceeding the corresponding boundaries. A hypothetical bullish 110–120 interval has thresholds 120, 115, and 110. These OHLC boundary tests do not guarantee an executable fill. Flag jumps across the interval between bars separately.
Observation horizons and incomplete follow-up
Record cumulative rates at 5, 10, 20, 50, and 100 bars and elapsed time. The proposed complete-window analysis excludes a horizon when data ends early or required bars are missing, while separately reporting censored and excluded counts. Do not mix full-available-history rates with fixed-horizon rates.
Waiting time among completed touches does not represent all events. Report touch rates and untouched or incompletely observed counts together.
Overlapping events
Store every qualifying event as a separate row without claiming statistical independence. Retain overlap and shared-move cluster information. Freeze the clustering unit for uncertainty estimates before execution.
Before execution
Specify venue, instrument, interval, UTC boundaries, dates, missing and revised data handling, clustering, and code version before promoting the protocol to v1.0. Binance candle documentation is a candidate reference, not a venue selection.
The planned report includes counts by direction, every horizon, gap sizes, waiting times, and quality exclusions. Label additions made after viewing outcomes as exploratory. No measured revisit rate or profitability conclusion is reported here.
Sources and documentation
Binance Spot REST API — Kline/Candlestick data
Continue with common foundations
Position Sizing: Convert a Risk Budget into Order Quantity · Trading Expectancy: Calculate Average Profit and Loss (scheduled) · Open Interest: What an Increase in Contracts Means · Python Price Data Checks: Duplicates, Timestamps, and Returns (scheduled)
Research methods and reporting standards
For education and research. Numerical examples are teaching assumptions unless explicitly identified otherwise. Historical results do not guarantee future performance.