Price levels that have traded heavy volume form stronger support and resistance — one of technical analysis's most widely held assumptions. We measured it. It cleared the randomness test comfortably. Then we added a single control variable and the effect evaporated.
A volume profile shows how much trading occurred at each price level. The level with the most volume is called the point of control (POC). The common reading is that a great many buyers and sellers changed hands there, so price reacts when it returns — support or resistance forms.
Made testable: if price is near a high-volume level, the subsequent move should be more muted. If the level holds, price should not pass through it easily.
The raw difference came out positive, and proximity to the point of control was stronger still.
| Measure | Value |
|---|---|
| Raw difference | +0.103 |
| Point of control (POC) proximity | +0.158 |
| Randomness test (null) | p = 0.000 |
The null test asks one question: if we shuffle the data and destroy its structure, would a result this strong appear by chance? The answer was a clear no — the p-value rounds to zero. What we were seeing was not noise.
This is where most analyses stop. The sentence "we found a statistically significant effect" describes exactly this table, and it is entirely true.
The second question we asked was: is this measure just restating something we already know?
There is an obvious relationship between volume-profile proximity and volatility. In a calm period price moves little, and because it is drifting in a narrow band it naturally sits near the high-volume region. Both are consequences of the same cause.
So we repeated the measurement holding volatility constant — asking, in effect, "if we compare only bars at the same volatility level, does volume profile still tell us anything?"
The effect did not disappear — it had never been there. What we had measured was not the support strength of volume profile; it was volatility itself, written in a different language.
Even the sign of the small remaining effect was not shared across symbols.
| Symbol | Residual effect |
|---|---|
| BTC | −0.058 |
| DOGE | +0.057 |
If a real market mechanism were at work, we would at least expect its direction to be consistent across symbols. Here BTC says one thing and DOGE says the opposite. That is separate evidence that the remainder is noise too.
A tautology explains a thing with itself. That is precisely what happened here:
The correlation is real. But "if volume-profile proximity is high the move will be muted" is a long-form restatement of "if price is calm, price is calm." It carries no new information. Used to generate a signal, it becomes a roundabout and noisier way of measuring volatility — while giving you the impression you have found something new.
This finding does not say volume profile is wrong. The observation that price slows down at high-volume levels is correct. What it says is: that observation adds no incremental information to a system already looking at volatility. Being true and being useful are different things.
The candidate was rejected, but this measurement changed every measurement that came after it. The rule it produced:
A null test alone is not enough. "Is this random?" and "is this a copy of information I already have?" are different questions requiring different tests. Volume profile passed the first at p=0.000 and fell to zero on the second. Since that date every measurement runs both: a null test and a partial correlation against the correct control variable.
Later candidates showed how much that rule was worth. Other measures fell into the same trap and were caught at the same gate — and some of them had produced far stronger numbers than volume profile did.
The rule has an inverse too: if the partial correlation comes out larger than the raw value, that is its own warning sign (a suppressor effect). That is the subject of a separate article.
So what happened to the candidate with the strongest numbers of all? Stablecoin supply produced the highest values we had ever recorded — and was rejected anyway. (coming)