Does an Expected Range Work on Intraday Charts?

The calibration figures that get quoted for a volatility band are almost always daily-bar figures, and most of the people reading them are looking at something else: a 4-hour chart, an hourly chart, a 15-minute chart. So the obvious question is whether a range that held on daily closes says anything about the chart actually open on the screen.

For our own band the answer is yes, and it was measured rather than assumed. The more useful part of the answer is where the measurement stops.

The short answer, with its scope

The FAQ page states the published position in one sentence: containment was approximately scale-invariant across resolutions, with inner-band coverage staying within roughly 65–72% from monthly bars down to 15-minute bars, and the shortest cut drifting slightly under. That is a historical measurement; past behavior is not a guarantee of future results.

For orientation, the headline figures on the Proof page are daily-bar figures: across the instruments tested, the next close landed inside the inner band about 71% of the time historically and inside the outer band about 94%. The resolution check asked a narrower question — does that inner-band coverage survive on faster bars? — and it was run on one instrument, which is the first thing to keep in view.

What the resolution check measured

The working paper re-ran the construction on SPY, the S&P 500 ETF, at seven bar frequencies. Three were at or above daily and cover roughly 32 years of data: monthly bars contained the next close 67.5% of the time, weekly 69.4%, daily 70.9%. Four were intraday, run on a shared three-month window across the cuts so that they compare like for like: 4-hour bars 70.2%, 1-hour 69.5%, 30-minute 68.4%, 15-minute 65.1%. All of these are inner-band figures, and all are historical.

Two features of that list matter more than any single number. The first is the spread: across two orders of magnitude in sampling frequency, coverage moved within about seven percentage points. The second is what was held fixed: the paper is explicit that the model's parameters were derived on daily data and not re-tuned for the intraday cuts. The daily construction was pointed at faster bars and scored the same way.

The daily figure here, 70.9% on SPY over 32 years, differs from the century-long S&P 500 figure quoted elsewhere on the site because it is a different instrument over a different period; the resolution table is internally consistent, not a re-statement of the headline.

Why a return-space band carries across timeframes

The construction can be moved between timeframes without being rebuilt because it contains no timeframe. Each bar contributes one return — the percentage change from the prior close to this close — and the band is a rolling mean and standard deviation of the last several dozen of those returns, projected forward from the prior close. A return is a return whether the bar took a month or fifteen minutes to form; the construction never asks.

That is a consequence of building the band in return space rather than on raw price levels, which is the subject of return space vs price space; the mechanics of the rolling window itself are covered in what is rolling volatility. One standard deviation of recent returns is the unit of "normal" at every resolution, and the yardstick is walked through in is this price move normal.

The paper calls this property scale invariance. As a structural claim it says only that the same construction at a fixed window should describe "an ordinary next move" in the same terms at any bar size; the resolution check tested whether real intraday data honoured it. Between 4-hour and 30-minute bars, it did, to within a couple of points of the daily figure.

What changes at fifteen minutes

At 15-minute bars coverage fell to 65.1%, the largest departure in the table. The paper describes this as consistent with higher microstructure noise at sub-hour frequencies, and that phrasing is deliberately modest: it is the explanation the result is consistent with, not a mechanism the paper set out to isolate.

The intuition is simple to state descriptively. At daily resolution a close summarises a whole session. At 15-minute resolution a close is one tick among a few dozen in a bar, and a sequence of such closes carries bid–ask bounce and tick-size discreteness on top of the price movement the band is describing. That extra noise widens the realised distribution slightly relative to what the window anticipated, and it shows up as more closes outside than the daily figure would suggest. The published position is that the drift is small at 15 minutes and that nothing faster was measured.

One scope point belongs here. The resolution table reports the inner band only. The outer band's behaviour across intraday cuts is not in that table, so the outer-band figure quoted above should be read as the daily-bar figure it is.

What "60 bars" means on an intraday chart

The one setting that matters, the rolling window, counts bars rather than calendar time. The How It Works page describes the canonical window as roughly 60 bars; on a daily chart that is about a calendar quarter, on an hourly chart it is about nine regular sessions, a little under two weeks, and on a 15-minute chart it is roughly two and a half sessions. Same setting, very different amount of history.

This is why the same page lists timeframe alongside symbol, source and window as the things that change the structure drawn: a band on the hourly chart and a band on the daily chart of the same symbol are different objects, not one object at two zoom levels. The trade-off in window length — shorter is more reactive and noisier, longer is smoother and slower — applies at every resolution. What the published check did not do is prescribe a window per timeframe, and neither does this article; it reports what a fixed construction did on the bars it was given. A backward-looking window at any resolution still describes the recent past, and the first bars of a fast volatility shock are where it runs narrowest — a limit set out on daily data in when volatility bands fail.

Where the evidence stops

"Works on intraday charts" is easy to read as broader than what was measured, so the boundaries deserve their own list.

The intraday check was a single instrument, SPY, on a shared three-month window. The cross-asset evidence — currencies, commodities, rates, crypto — is daily-bar evidence. Nothing published reports intraday containment on a currency pair or a crypto asset by name, and the FAQ's own wording applies: results outside the tested universe carry the usual caveats.

The check reports the inner band. The outer band across intraday cuts is not reported.

The floor is 15 minutes. Five-minute and one-minute bars were not measured, and the direction of the 15-minute result — slightly under — is the only hint the published work gives about what faster bars might do; a hint is not a measurement.

And, as with every result on this site, the paper validates the range's calibration, not the profitability of any use of it. Some traders keep one expected range on an intraday chart as a consistent frame for reading whether a move was ordinary for recent conditions — one of the uses described on the How It Works page. Whether any such use delivers value after costs is an open question the calibration work does not address, on any timeframe.

The takeaway

An expected range built from returns does not know what timeframe it is on, and the published resolution check found that real intraday data largely respected that: on SPY, inner-band coverage stayed within roughly 65–72% from monthly down to 15-minute bars with the construction untouched, and drifted slightly under only at the fastest cut. That is a measured statement about one instrument's inner band, with a 15-minute floor, and it is stated that way on purpose. The honest version of "yes" is a yes with edges.

Frequently asked questions

Does the expected range work on 5-minute or 1-minute charts? The indicator draws on any timeframe TradingView exposes, but the published measurement stops at 15-minute bars, where inner-band coverage on SPY had drifted to 65.1% historically, slightly under the daily figure. Nothing faster was measured, so a 5-minute or 1-minute band is reporting a real condition on a timeframe the published record does not cover. The 15-minute drift, attributed to microstructure noise, gives a direction but not a number for faster bars, and past behavior is not a guarantee of future results.

Should the window be shortened on intraday charts? The published check did not do that: the model's parameters were derived on daily data and were not re-tuned for the intraday cuts, and the result is a statement about that fixed construction. The How It Works page describes the general trade-off — a shorter window reacts faster and is noisier, a longer one is steadier and slower — and it holds at any resolution, but no window per timeframe is prescribed anywhere in the published work, and no setting removes uncertainty. A different window produces a different band, and its coverage would need to be measured on its own terms.

Does the intraday result apply to crypto or forex, which trade around the clock? Not as a published figure. The intraday cuts were run on SPY only; the crypto and currency rows on the Proof page are daily-bar results. A 24-hour market also has no session boundary at which a bar's close becomes a daily settlement, so the close series feeding a 15-minute band there is defined by the data feed's session settings. Whether inner-band coverage on those instruments looks like SPY's intraday coverage is untested and is not claimed.

Why publish a result showing the band under-contains at 15 minutes? Because the alternative is a table that stops at the last flattering row. The value of the resolution check is the shape of the whole table — coverage moving within about seven points across two orders of magnitude with nothing re-tuned — and the 15-minute row is part of that shape. A reader deciding whether to trust a daily figure on an hourly chart is better served by knowing exactly where the drift began than by an assurance that it never does; the Proof page applies the same rule to crisis onset and the deep tails.

The quickest way to see what scale invariance looks like is to put the same band on a daily and an hourly chart of the same symbol and watch both recalculate. BTM draws a return-space expected range on any TradingView timeframe, and you can start a free 30-day trial to compare the two on the markets you already follow.