Volume spread · Method
Waiting for the Next Bar to Confirm
Confirmation filters out some wrong readings and gives away the first part of the right ones. Both effects are real, both are measurable, and almost every published version of the rule mentions only the first.
The rule, and the two things it does
A volume-spread reading is taken from a single bar: a wide range on heavy volume, a narrow range that absorbed a great deal of stock, a close at one end of the range. Single-bar readings are frequently wrong, so the standard discipline is to wait. The reading is acted on only if the next bar behaves consistently with it.
That does two things at once, and they pull in opposite directions. It removes readings that were noise, because noise is unlikely to be followed by a consistent bar. And it gives back the first part of every reading that was real, because the confirming bar is itself part of the move.
The parameters above are stated rather than fitted, and they decide the outcome, a stronger drift favours waiting and a noisier series favours it more, while a shorter horizon favours acting at once. That is the honest conclusion of the exercise: whether confirmation is worth it is a property of the data, not a general truth, so it has to be measured on the instrument and timeframe in front of you.
The failure that makes the rule useless
"Wait for confirmation" without a fixed window is unfalsifiable, and it is the most common form of the rule in print.
If confirmation may arrive on the next bar, or the bar after, or some time this week, then every reading that eventually works is confirmed and every reading that does not is retrospectively declared to have lacked confirmation. The rule then has a perfect record and no content, the same structure the method page describes for a parameter sweep reported as a test.
The fix is cheap and complete: state the number of bars, state what counts as confirmation, and state it before looking. A rule with a fixed window can be wrong, which is what makes its record worth something.
What confirmation should require
| Element | A weak version | A version that can be tested |
|---|---|---|
| The window | "The next bar or soon after" | Exactly n bars, stated in advance. After that the reading expires. |
| The threshold | "A close in the right direction" | A close beyond the signal bar’s extreme by a stated fraction of average true range, so the distance means the same thing on any instrument. |
| The volume condition | Not stated | The confirming bar’s volume above its own recent baseline. This is the addition that makes a volume method worth using over a price-only one. |
| The record | The occasions it worked | Every reading, with the unconfirmed ones counted, including the ones that would have worked. |
The last row is the discipline that turns the whole exercise into something informative. Recording the unconfirmed readings that went on to work is uncomfortable and it is the only way to know what the filter cost, and the cost is precisely the quantity the figure above computes.
The asymmetry between the two directions
One refinement genuinely follows from the rest of this reference rather than from the pattern literature: confirmation is not symmetric between strength readings and weakness readings, because the underlying market is not.
Selling can be forced and buying almost never is, so a weakness reading confirmed by heavy volume is describing a mechanism that operates quickly and visibly, the volume-at-turns page sets out why. A strength reading has no equivalent: accumulation is deliberate and spread out, so the confirming bar is less likely to arrive promptly and less likely to be heavy when it does.
The practical consequence is that a single confirmation window applied to both directions is measuring two different things with one ruler. If the rule is being tested at all, it is worth testing the two directions separately, and if the results differ, that is a finding about the market rather than a failure of the rule.
Why volume belongs in the condition
Of everything that can be added to a confirmation rule, a volume condition is the most defensible, and it is the specific reason a volume-spread method has anything to offer over a price-only one.
A confirming bar on expanding volume means the move was made with a great deal of stock changing hands; the same bar on contracting volume means very few participants took part. That distinction is checkable against published data, it requires no assumption about who was buying, and it is the honest core of what this section is about, reading a bar’s range against its volume rather than reading a shape.
What it does not do is turn the rule into a forecast. A confirmed reading on heavy volume is a better description of what has happened than an unconfirmed one, and the horizon over which that description remains useful is exactly what the figure on this page measures, which is a modest claim and one that survives being checked.
Frequently asked questions
What does "confirmation on the next bar" mean?
That a reading on one bar is not acted on until the following bar behaves consistently with it. A strength reading is confirmed by the next bar closing higher, a weakness reading by the next closing lower. It is a standard discipline in volume spread analysis and in most pattern-based methods, and it exists because single-bar readings are frequently wrong.
Does it improve the readings?
It improves the proportion that turn out to be right, and it reduces what each correct one is worth. Those are two different quantities and almost no published account states both. The figure on this page computes them together on synthetic signals: waiting raises the share of real signals among those acted on, and it hands back the first part of every real move, because the confirmation bar is part of the move.
So is waiting worth it?
It depends on the balance between the two effects, which depends on how strong the real signals are and how noisy the series is, both of which are stated openly in the computation here. That is the useful conclusion: the answer is not a general truth about confirmation but a property of the data, so it has to be measured on the instrument and timeframe you actually use.
What is the elastic-window problem?
That "wait for confirmation" without a fixed number of bars cannot fail. If confirmation may arrive on the next bar, or the one after, or within the week, then any signal eventually either confirms or is retrospectively declared unconfirmed, so every success counts and every failure is disqualified. Fixing the window in advance is the difference between a rule and an excuse, and it costs nothing to state.
What should confirmation actually require?
Something stated in advance and measurable: a close beyond the signal bar’s high, or a close a stated fraction of average true range beyond it, within a stated number of bars. Expressing the distance in the instrument’s own volatility units matters more than it sounds, a fixed number of points is a strict test on a quiet instrument and a trivial one on a volatile one.
Does volume belong in the confirmation?
It is the most defensible addition to it. A confirming bar on expanding volume means the move was made with a lot of stock changing hands; the same bar on contracting volume means very few participants were involved. That is a real distinction and it is checkable, which is more than can be said for most refinements to pattern rules, and it is the specific reason a volume-spread method has anything to add over a price-only one.
How does this apply outside volume spread analysis?
Identically, wherever a rule says to wait. A trendline break "confirmed" by a second close, a moving-average crossover confirmed by a follow-through day, a breadth divergence confirmed by a price break; all have the same structure and the same two effects. The figure here is about volume spread readings because that is what this section covers, and the arithmetic is general.
What would change the conclusion?
Evidence that the filtering effect is larger than the give-back on real data for a stated instrument, timeframe and confirmation rule, which is a measurable question and the right way to settle it. Anyone using confirmation should run exactly that comparison on their own series rather than accepting either this page’s framing or the literature’s recommendation, because the answer genuinely differs between markets.