Alerts about big scheduled buys on-chain

The idea: you can see somebody place a large buy order split into equal slices, and the promised price impact is enormous. When we tested it, the real move over the order's lifetime was smaller than claimed by orders of magnitude, and there is nothing tradable in the alerts.

Why it sounded right

A scheduled buy on-chain is one of the rare cases where the future is literally visible. The order sits out in the open: total size, slice size, interval, how much is already filled. You know not only that somebody is buying but that they will keep buying over the next several minutes regardless of what price does. That is not a guess about intent, it is a timetable.

From there comes the calculation printed in every alert: the coin has some market depth, a buy of a known size should eat through it by so many percent. The figures came out impressive. And the logic feels airtight: if money is guaranteed to arrive in a thin market, price has to rise, and all you need to do is stand in front of that flow.

The sheer visibility adds to the case. Large addresses are visible, their behaviour is visible, and "follow the big money" sounds like common sense. It even has an obvious explanation for why it is not arbitraged away instantly: most market participants watch the exchange chart, not the chain, so they do not know another slice is coming in ten minutes.

How we tested it

We collected the whole available feed of alerts over several months — not just the last screenful the built-in view shows, but the full archive page by page — and matched each one against the coin's price on the exchange. First we checked the thing that matters most: does the claimed price impact match what actually happened while the order itself was filling. That is not about trading, that is about whether the mechanism exists at all.

Then we counted the result of entering at several hold lengths, collapsing repeat alerts on the same coin inside a day into a single trade — otherwise several messages about one and the same order pose as several independent confirmations. We split it by month to see whether the result repeats. Separately we checked the cleanest subset: alerts on one large asset, where the market is deep and there is the most data. And we ran a time-shift check — the same set of trades, but with entries rotated to other moments. If random shifts give you the same thing, you are not measuring the signal.

How it ended

There is no mechanism. A typical order fills over roughly an hour and a half, and the real price move over that same span is a fraction of a percent, while the alert promised a vastly bigger one. Off by orders of magnitude. The reason shows up the moment you look at where the buying actually goes: it hits a tiny pool on one network, against a wrapped version of the token. That pool re-levels with the global market instantly, and the pressure never reaches the main price at all. The calculation in the alert is arithmetically correct and meaningless in substance, because it uses the wrong depth.

The trading result matches. At every short hold length it is below the cost of getting in and out. More than half of all the profit over a day comes from a few trades on one coin, and some of those are alerts from the same day. Remove that coin and exactly nothing is left. The cleanest subset is negative at every hold length. It does not repeat by month: everything rests on two weeks inside one month — the same trap as in the other third-party signals we have tested. The time-shift check looked encouraging at first, but on inspection that too turned out to be one coin entirely.

One live thread is left, and it is inverted. On small coins that only trade on-chain, price after the alert does go the way of the buying at first — for the first half-hour or hour — and then turns down and ends the day meaningfully lower. So a large scheduled buy works as liquidity for somebody else's exit, and the side worth trading was the other one. Few cases, so this is a hypothesis, not a conclusion.

What this teaches

• Check the claimed mechanism before you check the result. If the promised price impact does not match the real move over the same span, there is nothing left to count.
• Look at which market the flow actually goes into. A wrapped token in a small pool is not the coin you are trading.
• Repeat messages about one event have to be collapsed, otherwise they look like independent confirmations.
• Check whether the whole result sits in one coin or one pair of weeks. Break it out by month and by ticker before you celebrate.
• A time-shift check is useful, but its result also has to be broken apart: it can fall entirely on one ticker.
• Alerts about "somebody is accumulating" point at somebody's exit more often than at somebody's entry. If you come back to this, test the other side.