Linear Regression

Start with 20 periods for daily charts. This balances responsiveness with smoothness. For longer-term trading, use 50-100 periods. For shorter-term, try 10-14 periods. Test different periods on your instruments.

Standard deviation measures how spread out prices are from the regression line. 2 SD bands capture approximately 95% of price action in normal conditions. Price at the bands is statistically 'extended' from the trend.

R² tells you how reliable the regression trend is. High R² (>0.7) means price closely follows the line - signals are more reliable. Low R² (<0.5) means price is scattered - regression signals become unreliable.

Yes, regression-based trading is suitable for TFSA accounts. The strategy typically involves moderate frequency trading appropriate for registered accounts.

Regression is a statistically optimal best-fit line; MA is a simple average. Regression has built-in slope measurement and R² for quality. MA lags behind price; regression line goes through the data optimally.

When R² is below 0.5, avoid regression-based trend trades - the trend isn't reliable. Options: 1) Wait for R² to improve, 2) Switch to range strategies (buy support, sell resistance without trend bias), 3) Use other indicators.

In strong trends (high R²), pullbacks to regression line are good entries. In moderate trends, wait for lower band touches for better risk/reward. Lower band entries have better risk/reward but occur less frequently.

When slope changes from positive to negative: 1) Exit longs immediately, 2) Don't fight the new direction, 3) Wait for new stable trend before re-entering. Slope change is a primary exit signal.

A close below the lower band in an uptrend is a breakdown signal. Options: 1) Exit immediately (conservative), 2) Wait for close back inside band (may be false break), 3) Add stop below break level if holding.

Use longer period (50-100) for major trend direction. Use shorter period (20) for entry timing. Both should be bullish for highest confidence. Enter when short-term is oversold but long-term is bullish.

Scan for: price near lower band (within 5%), slope > 0, R² > 0.6, RSI < 40. Rank by R² (higher = better) and distance to band. Most platforms require custom scripting or formulas.

Create an R²-adjusted position sizing: R² > 0.8 = full position, R² 0.6-0.8 = 75% position, R² 0.5-0.6 = 50% position, R² < 0.5 = no position. This automatically manages risk based on trend reliability.

Extend the current regression line slope forward, with parallel bands. Useful for identifying future S/R. Caution: accuracy decreases rapidly beyond a few bars. Re-calculate regularly as new data arrives.

Use polynomial regression when: 1) Linear R² is consistently low but price shows clear curved pattern, 2) Long-term trend clearly non-linear (e.g., exponential growth). Caution: polynomials can overfit and extrapolate poorly.

Well-designed regression systems (with R² filter, proper entries) typically achieve 50-60% win rates. Edge comes from risk/reward ratio at band entries (often 1:2 or better). Trend following approaches may have lower win rate but larger winners.