How to read it

This page explains what the scanner shows and how to use it.

What the scanner does

It finds tokens that have already run up in price and have just become a topic of conversation on Twitter. For each one it opens a signal: the scanner expects the price to fall over the next three days.

We arrived at this through testing. The first versions tried to catch attention before a rise. Seven weeks of measurement with six different methods showed they guessed no better than a coin toss. Catching the end of a rally turned out to work, so this is what the scanner does now.

A token on this screen is a candidate for a short position. It is not a recommendation to buy.

When a signal appears

Three conditions have to come together within one hour:

• The price has already run. Over the previous day the token rose much faster than the market.
• People have just started talking. Several different accounts posted about the token in that hour. One account posting many times, or copies of the same phrase, do not count.
• The market is not surging. When everything rises at once, overheated tokens usually keep rising too, and short positions lose money. In those hours the scanner waits.

The exact thresholds are our know-how; we do not publish them.

Signal 1, Signal 2, Signal 3

All three streams follow the same rules and differ in one number: how far the price must have run before the scanner takes interest. Signal 1 has the lowest bar and fires most often. Signal 3 has the highest bar and fires rarely.

Accuracy is roughly the same across the three. The difference shows up in depth: the more a token had to rise to trigger the signal, the deeper it typically falls afterwards.

The countdown on each card

16,2 / 72 means 16.2 hours have passed out of 72. The scanner's claim covers exactly three days from the moment the signal appeared, and the entry price on the card is the price at that moment.

The fresher the signal, the closer the current price still is to that entry. Cards younger than six hours are highlighted in green. A card past three quarters of its window fades out: most of the predicted move may already be over. The bar under each card shows the same countdown as a shape.

Closed positions

The right column lists every position whose three days have finished, newest first. A row like −17,2% · fell means the token lost 17.2% against the market and the scanner was right. A plus and rose mean the scanner was wrong.

Losing signals stay in this list on the same terms as winning ones. We sort by time and by nothing else, so you can check the record yourself.

The rule every position follows

Each position in this record is run the same way. A short or a long opens at the signal price and closes exactly 72 hours later, whatever the result at that moment. There are no stop losses and no early exits: the number in each row is the full 72-hour outcome, with losses shown at their real size.

Risk is managed by position size instead. Each position takes 2.5% of the portfolio, so the worst a single token can cost is 2.5%. There is no cap on how many run at once: this is a speculative book, and all of it may be in the market. The arithmetic sets the only limit — forty positions of 2.5% fill the portfolio.

The figure at the top of the right column is the portfolio under that rule: every closed stake compounded in order of closing, 2.5% into each. Next to it stand the two risk numbers — the worst a single position can cost, and the deepest the portfolio has been below its own peak over the whole record.

The pace is given as two annual rates, because the answer depends on what you do with the profit. APR is the simple rate: you take the profit out as it appears and the original capital keeps working. APY is the compounded one: the profit stays in, and the next stake is taken from the grown capital. Our own accounting works the second way. Both are an extrapolation of a short record, not a forecast.

If the whole portfolio is already in the market, a new signal still appears on the screen and still goes out as an alert, marked no capital. The record does not count it, because there was nothing left to open it with.

A token often fires again while its position is still open. A repeat that comes more than a day after the position's last alert is an instruction to add: another 2.5% of the portfolio goes into the same position, held for its own 72 hours from its own hour and its own price. No more than three stakes go into one token, so a single name never holds more than 7.5%.

The row shows ×2 or ×3 for the number of stakes, and its percentage is what the entry and the add-ons made together. The alert for a repeat says «add to the open position» rather than opening a new one.

The whole record, from 26 July 2026, is counted under today's rules, add-ons included. Up to 11 September 2026 those rules were run over the data we had collected; from 11 September every entry and every add-on is a live signal.

Every signal still makes its own claim: down over the next 72 hours from its own hour and its own price. Open the token card and each signal of the position is listed there with the hour and the price it fired at, so a later entry knows its own numbers.

The score next to each stream

58 of 76 fell · typically −8.52% reads like this: out of 76 finished signals, 58 fell as predicted. The typical figure is the median result — half of the falls were deeper, half were smaller. The count covers every finished signal since launch.

Why every figure is measured against the market

Each percentage on this screen is the token's move minus the market's move over the same three days. The raw return of a small token mostly repeats what the whole market did; after the subtraction, what remains is the token's own move, and this is what the scanner is judged on.

An example. The token fell 4% while the market fell 6%. Against the market the token is up 2%, so this signal counts as a miss.

The chart

The blue line is attention, the yellow one is price. Each line is scaled to its own usual range: attention can double overnight, while a 3% day is a big move for a price, so on a shared scale the price line would flatten out. Compare each line with its own past rather than one line with the other.

The green zone above and the red zone below mark levels that are unusually high or unusually low for this particular token.

When a chart is missing

The chart starts on the day we began tracking the token. Until the window fills up, we draw nothing for the days when nobody was watching.

Which tokens can appear

The token must be listed and traded, with a readable price at both ends of the window. Otherwise there is nothing to measure.

What the screen leaves out

• Trading costs. Exchange fees and the funding paid on a short position are real and are not included in these figures.
• Slippage. How much the price moves against you when you place an order. We measure it separately; it is not in these numbers either.
• Any promise about the future. The record covers the time since launch, and that is not long yet.

Free and subscription

Finished results and the full record are open to everyone, with no key and no registration. Live signals, the ones still inside their 72 hours, go to subscribers. There is one subscription, and it opens everything: every live signal plus alerts to Telegram and to your devices.

How often it updates

Collection and scoring run every hour, and the screen is rebuilt on the same cycle. The countdowns keep ticking in between.