SV · Foundations
Statistical indicators
Also called: statistical trading indicators, quantitative indicators
A statistical indicator estimates a defined property of returns, such as volatility, memory or a probability, rather than applying a fixed chart rule.
Definition
A statistical indicator is a chart study that estimates a well-defined quantity of the price process from data, with an explicit model behind it: for example a volatility estimate, a z-score against a fitted distribution, a Hurst exponent, an entropy measure or a probability band.
The distinction from a purely heuristic indicator is not complexity but interpretability: each output has a meaning that can be tested, and ideally an uncertainty attached.
Intuition
Many classic indicators are smoothing rules with arbitrary parameters. They can be useful, but their readings have no natural scale: a value of 70 means whatever convention says it means. A statistical indicator says something checkable, such as "price is 2.1 standard deviations above its 20-day mean under the current volatility estimate".
Because the claim is explicit, you can test whether it holds out of sample and whether the effect survives costs.
How it's calculated
Most statistical indicators follow the same pattern: choose a model of returns, estimate its parameters on a rolling window, and express the current observation in the model's terms. A simple example is the rolling z-score:
zt = (xt − μ̂t) / σ̂t
where μ̂ and σ̂ are estimated from the preceding window only. Other examples replace the normal model with an empirical percentile, a GARCH variance, or a regime probability from a hidden Markov model.
Good practice records the estimation window, the data used at each step (no future bars), and, where possible, a confidence interval for the estimate.
Example: if a 20-day window gives a mean daily return of 0.05% and a standard deviation of 1.2%, a day with a 3.0% return has z ≈ (3.0 − 0.05) / 1.2 ≈ 2.5: unusual under a normal model, but not rare for fat-tailed returns.
How to read it
Ask three questions of any indicator: what quantity does it estimate, how noisy is that estimate on this window, and has the effect been tested out of sample? An indicator that cannot answer the first question is a heuristic, which is fine as long as it is treated as one.
Compare readings against the indicator's own historical distribution, not just fixed thresholds.
Common mistakes
- Fitting parameters on the full history and then evaluating on the same data (look-ahead and overfitting).
- Testing many indicator settings and reporting only the best one without adjusting for multiple comparisons.
- Treating a statistically significant pattern as economically significant before accounting for costs.
- Assuming the return distribution is stable across regimes.
In Saqrid
SV ships a library of 100 built-in proprietary formulas, including momentum, Hurst, entropy, expected-move and probability-cone tools.
You can also write your own indicators in Python in a built-in code editor and run them on any chart.
Open SV.
Go deeper
Saqrid Institute lesson: Bands and Ranges: What Survives Statistical Scrutiny. Lesson text is open to read without an account.