Every trader has had the moment: a candle stretches further than the ones around it, and the question arrives before any analysis does — is this normal, or is something happening?
Eyeballing it doesn’t answer that question. A 2% day looks dramatic on a quiet chart and unremarkable on a volatile one. “Normal” isn’t a fixed number; it depends entirely on how the market in question has recently been moving. The good news is that this is a question statistics answers well — as long as you frame it carefully.
A 3% move in a large-cap index is a very different event from a 3% move in Bitcoin. Even for the same instrument, a move that would have been ordinary in a stressed market can be extraordinary in a calm one.
So the useful question is never “is this move big?” but “is this move big relative to how this market has lately been moving?” That reframing turns a vague impression into something measurable: collect the recent percentage moves, measure how spread out they’ve been, and check whether today’s move sits inside or outside that typical spread.
The workhorse for this is the standard deviation of recent returns — a single number describing how much a market’s percentage moves typically vary. If a market’s daily moves have recently had a standard deviation of 1%, then a 0.5% day is routine, a 2% day is notable, and a 4% day is rare.
Two details matter more than they first appear.
Measure returns, not prices. A common mistake is to compute dispersion on raw price levels. In a trending market, price keeps pulling away from its own average, so a price-based band drifts off-center and widens for the wrong reason — the trend, not the volatility. Measuring in return space (percentage moves anchored to the prior close) keeps the yardstick centered on where price actually is and scaled to current conditions. This distinction is one of the design choices behind Oisigma’s approach; the How It Works page walks through it step by step.
Use a rolling window. Markets change character. A window of roughly the last 60 bars is long enough to be stable and short enough to reflect current conditions. As each new bar prints, the window rolls forward and the definition of “normal” updates with it — no manual redrawing.
Put those pieces together and you get an expected range: a center level implied by recent behavior, an inner band where price usually trades, and a wider outer boundary crossed only on unusual days. Each new close then answers the question directly. Inside the inner band: typical for recent conditions. Beyond it: unusual. Beyond the outer band: rare.
Note what this is and isn’t. The range describes; it doesn’t predict. A close outside the band tells you the move was atypical relative to recent behavior — it does not tell you what happens next, and it is not a buy or sell signal.
Most tools marketed for abnormal price move detection scan for something else entirely — volume spikes, unusual options activity, outsized prints. Those are legitimate things to watch, but they answer a different question — is participation unusual? — rather than is the price move itself unusual? The two often diverge: heavy volume can accompany a perfectly ordinary price change, and a genuinely rare price move can print on modest volume.
A dispersion-based range answers the price-side question directly, with no scanner and no data feed: the move either stepped outside the recent norm or it didn’t.
A range like this is only useful if it’s calibrated — if “usually inside” holds up when you check it against history. This is measurable, and Oisigma’s working paper measured it at scale.
On daily S&P 500 data from 1928 to 2024 — about 24,000 trading days — the next close landed inside a ±1σ expected range built this way about 71% of the time. Broken out decade by decade, that rate stayed within a narrow 68.7%–73.7% band through the Great Depression, the 1987 crash, the 2008 crisis, and COVID. Across 40 instruments spanning equities, FX, commodities, rates, and crypto, the cross-instrument average was roughly 72%, and the wider ±2σ boundary contained about 94% of closes. All of these figures are historical calibration measured in research; past behavior is not a guarantee of future results.
One number worth pausing on: in a plain, well-behaved (Gaussian) market measured with the same 60-bar construction, that inner-band figure would be about 67%. The extra few points are the fingerprint of fat-tailed real markets — many modest days, occasional extreme ones — reproduced in simulation to within a tenth of a percentage point. The full tables and tests are on the Proof page and in the working paper.
No. A calibrated range identifies that a move is unusual as it happens or after the bar closes — it does not forecast direction, and it does not say the move will continue or reverse. Anyone selling abnormal-move detection as a prediction engine is overclaiming. What detection gives you is context: whether a headline, an earnings gap, or a quiet drift actually moved the market more than a normal day would.
No historical yardstick is perfect, and the failure modes are worth knowing.
The calibration is an average property. It holds in the long run, not in every moment. In the first days of a fast crisis — when volatility spikes faster than a rolling window can register — the range runs too narrow, and historical containment on those onset days dropped to roughly 65%. The outer band is also slightly optimistic in the deepest tails. And the center of the range is a reference point, not a forecast: in testing it carries almost no information about which direction price goes next.
None of this breaks the tool; it defines it. An expected range is context for your own judgment, not a verdict, and it doesn’t remove uncertainty or risk.
The practical value of a calibrated expected range is consistency. Instead of a different intuition for every symbol and timeframe, you ask one question everywhere: is price inside or outside its normal range for recent conditions? A breakout, a pullback, an earnings gap, a quiet drift — all get read against the same yardstick, one that recalculates every bar.
That’s what the Behavioral Transform Model (BTM), Oisigma’s calibrated expected-range indicator for TradingView, puts on the chart: an expected-price reference line, a normal range, a wider abnormal-move band, and markers when price steps outside. No arrows, no signals — just structure, with the methodology published in full.
If you’d like to see how “normal” looks on the markets you actually trade, you can try BTM free for 30 days ($15/month after, cancel anytime). Watching the range recalculate bar by bar on your own charts is the fastest way to judge whether this way of framing price behavior is useful to you. Start your free trial →
Oisigma provides descriptive market analytics for educational use. It is not investment advice, does not predict prices, and does not provide buy or sell signals. Statistics referenced are historical and were measured in our working paper (not peer-reviewed); past behavior is not a guarantee of future results. Trading and investing involve substantial risk of loss, including the possible loss of all capital invested. Leveraged products (futures, options, margin) carry additional risk and can result in losses that exceed your initial investment. Bollinger Bands® is a registered trademark of John Bollinger; Oisigma is not affiliated with or endorsed by Mr. Bollinger. RiskMetrics® is a registered trademark of MSCI Inc.; Oisigma is not affiliated with or endorsed by MSCI Inc. Nothing in this article is a recommendation to use any particular strategy. Read the full Disclaimer →
Try it free for 30 days and see the range update as new bars print, on whatever symbols and timeframes you actually trade.
30 days free, then $15/mo. Cancel anytime from your account.
Pick up where you left off.