Trade Cluster Clustering
Ten percent of total profit attributed to a single intraday event can skew the entire statistical profile of an orb strategy. Every teardown orb trading backtest freshcornercafe has logged shows the same thing regarding trade cluster clustering. A single outlier during the first hour often masks a lack of consistent edge in the five minute range. This clustering analysis separates repeatable patterns from accidental windfall during a high volatility market open.
The Mechanics of Cluster Identification
Analyzing trade distribution requires mapping every execution against a specific timeframe. A tight cluster of winning trades occurring only during the first fifteen minutes suggests the strategy relies on initial volatility rather than sustained momentum. If the majority of gains happen during a singular expansion of the thirty minute range, the data lacks the density required for a robust model. One must look for a steady cadence of trades across regular trading hours to confirm the edge. A sudden spike in performance during a single session high does not constitute a repeatable system. The work involves stripping away these outliers to see if the remaining trades maintain a positive expectancy.
Detecting Outlier Dependency
Statistical noise often hides within a large sample size. An opening range breakout that captures a black swan move creates a false sense of efficacy. When the data shows profit concentrated in one specific window, such as a sudden move during the cash open, the model is fragile. The analysis focuses on the frequency of wins across different segments of the day. If the profit curve is a staircase with one massive vertical leap, the system is vulnerable to regime changes. A healthy distribution spreads gains across the opening range and through the middle of the session. A pattern that only functions during the opening bell is a momentum scalp, not a trend following system.
Measuring Distribution Density
The process involves segmenting the data into discrete blocks. Comparing performance in the fifteen minute range against the performance in the final hour provides a clear view of time based decay. If the edge evaporates after the first hour, the strategy is a volatility play. If the edge remains consistent from the premarket through the afternoon, the strategy is structural. Quantitative tools must isolate the specific time frame where the highest win rates occur. This prevents a single massive move from inflating the average trade size. A small sample overstates the edge when a single outlier is present.
Validation Through Systematic Filtering
Refining the data requires removing the top and bottom five percent of outlier trades. This reveals the true baseline performance. A strategy that remains profitable after removing the largest single day of gains is structurally sound. If the expectancy turns negative after filtering, the system is a lucky bet on a specific volatility event. The goal is to find a steady stream of trades that occur regardless of whether a single massive move happens. Consistency is measured by the lack of reliance on any single period of the day.