Sample Bias via Regime Filtering
Two equity curves that look identical on a chart can hide vastly different underlying mechanics, as the notes at orb trading backtest freshcornercafe describe. A profitable opening range breakout strategy might only function during periods of high volatility, meaning the backtest results rely on a specific market regime rather than a repeatable edge. This type of sample bias occurs when the data set selected focuses on a narrow window of intraday volatility.
The Mechanics of Regime Filtering
A backtest spanning only two years might capture a period of constant trend expansion. During these months, the price frequently breaks the high of the first fifteen minutes and continues toward the session high. This creates an illusion of consistency. If the data set excludes periods of sideways chop or low volume, the statistical advantage is overstated. The error lies in the selection of the time frame rather than the logic of the trade itself. A robust test requires data from various volatility environments, including low volatility regimes where price often stays within the initial range.
Identifying Volatility Clusters
Regimes are often defined by the standard deviation of price movement during regular trading hours. When a trader evaluates an orb strategy, the presence of a single massive trend can skew the entire expectancy. If the thirty minute range is consistently wide, the strategy appears to have a high win rate. However, if the data set lacks periods where the market open is quiet, the edge disappears in real world conditions. Testing must include periods of mean reversion to see how the system handles a lack of direction. A sample that only shows expansion is a filtered sample.
Timeframe Sensitivity and Edge Decay
The choice of the fifteen minute range as a measurement tool changes the profile of the data. A strategy optimized for a 15 minute candle may fail when the market shifts to a regime of micro-volatility. When the intraday movement is compressed, the breakout signals often become false entries. A backtest that ignores these compressed sessions fails to show the drawdown potential. Mechanical testing requires looking at the volatility of the opening bell relative to the rest of the session to ensure the edge is not tied to a specific level of movement.
Data Integrity in Backtesting
A small sample overstates the edge because it cannot account for the transition between regimes. If the testing period only covers a bull market, the results reflect market direction rather than the strategy. A mechanical audit of the data involves checking the average true range across different months. If the profitability correlates too closely with the ATR, the strategy is a volatility play, not a structural play. Comparing the performance across the first hour against the total daily range helps isolate this bias.