Most price bands answer one question: how far is price from its average right now? Percentile deviation bands answer a sharper one: how unusual is that distance compared with how price has actually behaved recently?
That shift — from a fixed, formula-driven width to an empirical, observed one — is the whole idea. This guide explains what percentile deviation bands measure, how to read them on a chart, and where they part ways with the Bollinger Bands most traders already know.
The problem with standard-deviation bands
Bollinger Bands are built from a moving-average centerline plus and minus a multiple of the standard deviation of price (classically a 20-period SMA at two standard deviations). It's elegant, and it adapts to volatility: when price gets choppy, the bands widen; when it calms down, they contract.
But the standard-deviation approach carries an assumption that markets don't honor: that returns are roughly normally distributed and symmetric. In a normal distribution, two standard deviations would capture about 95% of observations. Index futures like NQ don't behave that way. Their return distributions have fat tails — large moves happen far more often than a bell curve predicts — and they're frequently asymmetric, behaving differently on the way up than on the way down.
Two practical consequences follow:
- Outlier sensitivity. Standard deviation squares each deviation before averaging, so a single violent candle inflates the band width well after the event has passed, distorting your read of "normal."
- A misleading probability. When price tags the upper Bollinger Band, the textbook reads it as "~2σ stretched." But that 95% interpretation only holds if the data is normal. On NQ, it usually isn't — so the band touch doesn't mean what the statistics class said it would.
What a percentile deviation band actually measures
A percentile deviation band keeps the idea of a centerline but replaces the width with something non-parametric. Instead of computing a standard deviation and multiplying it, it looks at the actual distribution of recent deviations from the centerline and asks where the current deviation ranks within it.
A band drawn at the 90th percentile sits at the level that recent price exceeded only 10% of the time. So when price reaches it, the reading is direct and literal: price is more stretched than it has been in roughly 90% of recent observations. No normality assumption, no multiplier to tune by feel — the band reflects the market's own recent behavior.
The TradeScorer band framework uses three reference levels for exactly this reason:
- 75 — the routine edge of value; price is stretched but ordinary.
- 90 — meaningfully extended versus recent behavior.
- 97 — a statistical extreme; price is at a level it rarely reaches.
Percentile bands vs Bollinger Bands: the differences that matter
Distribution assumption
Bollinger Bands assume a shape (normal) and derive width from it. Percentile bands assume nothing about shape and measure the distribution directly. On instruments with fat tails, the second approach describes reality more faithfully.
Robustness to outliers
Because percentiles are based on ranking rather than squared deviations, one extreme candle moves a percentile band far less than it moves a standard-deviation band. Your sense of "normal" stays steadier through volatility spikes.
Interpretability
"Two standard deviations" is only a probability if the data is normal. "The 90th percentile of recent deviation" is a probability by construction, whatever the distribution looks like. That makes a band touch easier to reason about under pressure.
Asymmetry
A symmetric standard-deviation band treats stretch above and below the centerline identically. Measuring signed deviation by percentile lets the upper and lower bands reflect the market's actual up-versus-down behavior — which, on NQ, is rarely symmetric.
How to read them on a chart
Percentile deviation bands are a context tool, not a signal generator. Read them in three layers:
- Value — price hugging the centerline is trading in its recent fair zone; there's no statistical stretch to fade.
- Stretch — price pushing past the 75/90 bands is extended relative to recent behavior. In a ranging or rotating regime, that's where mean-reversion setups tend to live.
- Extension — a tag of the 97 band marks a statistical extreme. It doesn't predict a reversal, but it tells you the move is unusual versus its own recent history.
The honest caveat: a stretched reading in a strong trend can stay stretched for a long time. Bands describe where price is, not what it will do next. They earn their keep as a filter and a context layer, paired with structure and regime — not as a standalone entry trigger.
Centerline and reference-method choices
Two knobs change how a percentile band behaves:
- The centerline. A simple SMA or EMA is enough to learn the framework and read stretch quickly. More responsive or session-aware centerlines — VWAP, session VWAPs, or smoother averages like HMA — change how tightly the band tracks price and suit different workflows.
- The reference method. The percentile method is the core idea, but related robust measures (MAD, IQR) and the classic standard deviation can each define the band's spread, letting you trade off responsiveness against stability.
If you want to learn the concept, start with an SMA/EMA centerline and the percentile method — that's what Forge PDB Lite is built for, and it's free. If bands become part of your daily TradingView routine and you need session VWAPs, robust methods, the 97 extreme band, and mean-reversion alerts, that's the job of the full engine on the Percentile Deviation Bands page.
FAQ
Are percentile deviation bands better than Bollinger Bands? Not universally — they answer the question differently. Bollinger Bands assume a normal distribution; percentile bands measure the actual recent distribution, which tends to describe fat-tailed instruments like NQ more faithfully. "Better" depends on what you're trying to read.
Do percentile deviation bands repaint? A band can update on the current, unclosed candle as new data arrives, like any indicator computed on live price. The percentile levels themselves are derived from confirmed historical deviations, so once a candle closes its reading is fixed.
Try it on your own charts
The fastest way to understand percentile deviation bands is to watch them frame stretch on a live NQ chart. Forge PDB Lite is free and gives you the core percentile-band concept on an SMA or EMA centerline. When you want the full engine — VWAP variants, robust methods, the 97 extreme band, and alerts — see the PDB page.
Related TradeScorer tool: Percentile Deviation Bands.
TradeScorer products are technical analysis tools, not investment advice. Trading futures involves substantial risk of loss. Read the full risk disclosure before use.