Linear Regression Slope

What it measures:

Measures the direction and rate of change of price over a rolling lookback window by fitting a least-squares regression line to historical prices. A normalized log-price slope expresses the fitted trend's approximate percentage change per bar, while the regression R-squared (R²) measures how closely prices follow a linear trend.

How to interpret it:

• Positive Slope: Indicates an upward fitted price trend.
• Negative Slope: Indicates a downward fitted price trend.
• Slope Near Zero: Indicates limited linear directional movement over the selected window, potentially reflecting consolidation or choppy price action.
• Increasing Positive Slope: Suggests the upward trend is strengthening.
• Positive but Declining Slope: Suggests the upward trend may be losing momentum, although the trend remains positive.
• Negative but Rising Slope: Suggests the downward trend may be weakening, although the trend remains negative.
• Increasingly Negative Slope: Suggests the downward trend is strengthening.
• High R²: Indicates that the fitted linear trend explains a large proportion of price variation within the window; it does not guarantee future trend continuation.
• Multi-Window Confirmation: Agreement between short- and longer-window slopes supports directional consistency, while divergence may indicate a transition in trend.

Used for:

• Classifying bullish, bearish, and neutral trend regimes.
• Assessing trend direction and changes in directional strength.
• Detecting potential trend weakening or transition when slope changes sign or its rate of change shifts.
• Confirming SMA/EMA alignment and other trend indicators.
• Combining with ADX, momentum, volatility, and Price Z-Score to distinguish directional trends from potentially extended prices.
• Generating normalized trend features for quantitative trading and machine-learning models.

Limitations:

• The slope describes historical movement within the selected window and does not reliably predict future returns by itself.
• Results depend on the lookback period, price transformation, and normalization method.
• Raw price slopes are not directly comparable across instruments with different price levels; normalized or log-price slopes are generally more suitable for cross-asset comparisons.
• Short lookback windows can be noisy, while longer windows may respond slowly to reversals.
• A linear regression line may poorly represent nonlinear trends, sudden price gaps, or changing market regimes.
• High R² measures historical linear fit, not predictive accuracy or the probability that a trend will continue.
• Slope thresholds for trend classification should be calibrated to the instrument and validated using out-of-sample or walk-forward testing.