4 / 6 · 11 min
Volume surge and impulses: catching an event while it is still running
Both numbers answer where something is happening right now. But a surge of the same size can mean completely different things — and what tells them apart is not the volume but what stands next to it.
Two numbers about the present tense
The 24-hour change tells you about something already finished: by the time a coin shows up among the day's leaders, the move is usually over. Volume surge and impulses answer a different question — where something is happening right now, while the event is still running. These are the only two numbers in the service that go stale within minutes, and they should be read accordingly.
What the word usual actually means
A surge is the ratio of volume over the last 5 minutes to what is usual for THIS instrument. Everything rests on the word usual, so it matters how it is computed: the last 24 hours are cut into 288 five-minute buckets, the current bucket is excluded, and the median of the rest is taken.
Median, not average: a single news bucket among 288 lifts the average so much that every ordinary five minutes ends up below normal. The cushion on both sides is there for illiquid instruments — their median sits near zero and a bare ratio flies to ×200 on a single trade. On a liquid instrument the cushion changes nothing; on a dead one it damps the outlier while preserving the ordering, so the column stays sortable.
Three times busier than usual — is that a lot?
That question cannot be answered in words, so here is a measurement across all 1217 instruments at one moment. The median surge across the market is 0.99 — the ordinary instrument trades exactly as usual, which on its own is a sign the yardstick is not skewed.
| Surge | Instruments | Share of market |
|---|---|---|
| ×2 and above | 60 | 4.9% |
| ×3 and above | 24 | 2.0% |
| ×5 and above | 15 | 1.2% |
| ×10 and above | 6 | 0.5% |
Hence the practical part: ×2 is ordinary, six dozen of them exist at once and there is nothing to be surprised at. ×5 is already a dozen and a half across the entire market — a list you can scan by eye in a minute. Knowing these shares matters more than knowing the definition: otherwise the threshold gets picked blindly and the feed is either empty or flooded.
Two large surges with opposite meanings
Measured at the same moment. COOKIE: surge ×28.56, up 24.1% on the day, daily range 42.4%, open interest at its yearly MAXIMUM — percentile 100. EVAA: surge ×18.57, up 1.3% on the day, daily range 3.68%, open interest at its yearly MINIMUM — percentile 0.2. Both surges are huge. But in the first case volume arrived together with the move and with new positions: money entered the market. In the second, price is standing still and positions did not grow — so the same volume means positions merely changed hands. One number, two different events.
A surge only reads in pairs
| Surge | Price | Open interest | What is happening |
|---|---|---|---|
| high | moving | rising | new money entering the move |
| high | moving | falling | old positions closing — such a move is finite |
| high | flat | rising | someone is building a position without moving price |
| high | flat | unchanged | turnover: positions changing owner |
Treating a surge on its own as a reason to enter
A surge says people came here, but not who they are or which side they took: every trade has both a buyer and a seller, and volume measures interest, not direction. A row reading ×12 without price and open interest beside it is half a sentence. That is exactly why the table above has four rows rather than two, and why the surge column sits next to the price change rather than off on its own.
Impulses: where the threshold runs and why it differs
An instrument enters the feed when price travels more than 3% in a minute or more than 6% over 5 minutes. For currencies, metals and commodities the thresholds are their own — 0.3% and 0.6%. The reason is arithmetic, not caution: the daily travel of EURUSD is about 0.4% and bitcoin's is about 1.5%, so on a common yardstick currency pairs would never reach the feed, not even on a Fed decision day. The difference in travel is roughly fourfold while the threshold is cut tenfold — deliberate slack for quiet days. Remember it exactly that way: the threshold is ours, not the market's, and the feed shows what is unusual BY THAT YARDSTICK. One coin enters the feed no more than once every 2 minutes, otherwise a single move crowds out everything else.
Thirteen entries are one event
A measurement of the feed at one moment: 60 entries and only 23 coins, of which 53 entries are one-minute and 7 are five-minute. A single coin — GLMR — took 13 of the 60 slots: −7.0%, then +6.4%, then −6.2%, round and round. Its daily turnover is $180 thousand and over the past hour it travelled +21.8%. Thirteen entries here are not thirteen opportunities. They are one instrument that has entered a high-volatility regime: at that turnover a three percent move in a minute is the normal course of its life, not an event. A repeat in the feed reads as a sign of a STATE, not as a queue of signals.
Trading everything that appears in the feed
An impulse is a fact, not a forecast: it reports that a move started and promises nothing about continuation. Roughly half of impulses retrace within an hour. Worse is something else: entering off the feed on a small instrument means buying at the worst available price — you arrive exactly when the near book has been cleared out, and you pay for it in spread and slippage. The feed is useful not as a source of entries but as a way to learn where to look; the decision is made on the chart and by your own rule.
Why bother, if the move is visible on the chart anyway
On a chart you see the move of one instrument, and only if you happen to be looking at it. We have 1217 instruments across nine venues, and nobody can scan them by eye within a minute. Both metrics exist for one purpose: to narrow 1217 down to a dozen and a half worth your attention right now. That is precisely the work a human should not be doing and a machine does for free.
The impulse feed sits next to the table and updates in real time, and the surge column is sortable. Sort by surge and look at the top ten: for how many of them is price standing still? Usually it is more than half — and those are more interesting than the ones that have already flown.
Impulses and surgesTwenty impulses and an hour later
Take twenty entries from the feed in a row and write down three things for each: the volume surge at that moment, the instrument's daily turnover, and the change in open interest over the hour. Exactly one hour later, note where price sits relative to the impulse level. Then split the twenty cases into two piles — those where open interest was rising and those where it was falling — and compare how many held. The difference will most likely be sharper than the one between a big and a small surge, and it will be your own measurement.
An instrument shows a surge of ×12 while price has not moved in five minutes. Is the data broken?
No, and this is one of the most interesting configurations. There is volume but no movement — so buyers and sellers are finding each other at one price, and somebody is absorbing the entire flow without pushing it. Open interest settles the question: if it is rising, somebody is building a position and trying not to move price; if it is flat, this is turnover — positions simply changing owner. The first is often more interesting than any impulse, the second means almost nothing.
Why is the impulse threshold for gold and currency pairs ten times lower than for crypto?
Because the threshold measures how UNUSUAL a move is, not how large. EURUSD travels about 0.4% a day and bitcoin about 1.5%: three percent in a minute does not happen on a currency pair for years, and on a common yardstick none of them would reach the feed even on a Fed decision day. Note that the travel differs roughly fourfold while the threshold is cut tenfold — the slack is deliberate, so the feed does not fill with ordinary daily bustle. The general lesson: any threshold has to be set against the instrument's own norm, otherwise it measures the size of a move rather than the event.