Trend Projection Overlays & Volatility Zones

Volatility and Trend Boundaries

Understanding trend boundaries is key to retail risk management. A "trend projection overlay" is a mathematically computed band drawn around the current price that represents where the trend would plausibly extend to, given recent volatility. The band widens when the underlying is volatile and narrows when it's calm — so the same overlay tells you both the trend direction and the volatility regime in one visual.

The most common implementation is a rolling ordinary-least-squares (OLS) linear regression computed on the last N closing prices, with a parallel "envelope" drawn at ±K standard deviations of the residuals. This is structurally similar to Bollinger Bands, but with a linear regression line as the centre rather than a simple moving average — which means the band tracks the trend rather than lagging behind it the way an SMA-based band does.

We call our implementation the Ghost Line — because it projects the trend forward into the future as a "ghost" extension of where the regression line would go if the recent trend continued indefinitely.

Actionable Breakout Zones

By overlaying these volatility bands, the terminal identifies areas of trend exhaustion, helping you avoid buying at the absolute high or selling at the absolute low. The two specific zones the Ghost Line surfaces are:

• Upper exhaustion zone — price has reached the upper band. Statistically, mean-reversion is more likely than continuation from here. If you have a long position, this is where you should be tightening your trailing stop. If you don't, this is where you should not be initiating a new long.
• Lower exhaustion zone — price has reached the lower band. Mirror image. Tighten stops on shorts, don't initiate new shorts.

The zones are not "reversal signals." They are "this trend is overextended" signals. The trend can absolutely continue past the band — strong momentum breakouts do exactly that — but the probability of mean-reversion is materially higher at the band than at the centre. That probability is what your position-sizing should reflect.

The math, briefly

For a regression window of N periods:

1. Compute the OLS slope and intercept: slope = Σ((xᵢ - x̄)(yᵢ - ȳ)) / Σ((xᵢ - x̄)²), intercept = ȳ - slope·x̄, where xᵢ is the period index and yᵢ is the closing price at period i.
2. Compute the residuals: εᵢ = yᵢ - (intercept + slope·xᵢ).
3. Compute the residual standard deviation: σ = √(Σ(εᵢ²) / (N - 2)) (N-2 because we fit two parameters).
4. Draw the centre line as intercept + slope·x extended forward, and the upper/lower bands as centre ± K·σ.

The choice of N is the only real tuning knob. Short windows (N=14) make the overlay responsive but noisy. Long windows (N=200) make it smooth but lagging. Strat Ai defaults to N=50 with a multiplier of K=2 — empirically this works well for intraday NSE index options, but you can tune it per underlying in the trend projections docs.

Why this beats static support/resistance

Static support/resistance lines drawn at "previous day high" or "round numbers" ignore the volatility regime. A resistance line at 22,000 on Nifty is a different proposition when Nifty's 14-day ATR is 80 points vs. when it's 200 points. The regression band automatically scales with realised volatility, so the same overlay tells you both where the trend is going and how confident you should be about it.

This is the same reason institutional desks use volatility-scaled bands rather than fixed-price levels. Fixed levels are a retail-trader simplification that ignores half the information in the price series.

Common pitfalls

• Don't treat the band touch as a reversal signal. It's an exhaustion signal. Reversals happen when additional confirmation shows up — typically a divergence in RSI or a shift in the conviction score. See our consensus decoder for how these combine.
• Don't use a single N. A 14-period Ghost Line on a 5-minute chart and a 50-period Ghost Line on a daily chart measure different things. Use both; they're complementary.
• Don't ignore the slope. A flat centre line in a sideways market makes the bands meaningless — the regression has no trend to project. The conviction score's regime classifier will tell you when the market is ranging, and the Ghost Line is much less useful in those regimes.

External context

The OLS regression band is structurally similar to the Keltner Channel and the Bollinger Band, but with a regression centre instead of a moving-average centre. John Bollinger's original Bollinger Band paper is the canonical reference for the volatility-band approach; the regression-centre variant is less commonly discussed in retail literature but is standard in institutional quant tooling. The specific N=50, K=2 default is empirical — it's what backtests well on NSE index options data over the 2021–2025 sample — not a theoretically optimal value.