Chapter 7
Bot detection
Volume is easy to fake and a chart is easy to draw. The last tile on the risk strip is the card's answer to one question — was this tape written by people, or by a script?
7.1What the Real tile tells you
Real is the fourth tile on every card's risk strip. Two layers of checking feed it: lightweight signals that every token on the board gets, and a deeper look at the candles themselves for a limited number of cards each scan — plus any card you check by hand.
| Tile reads | Meaning | What to do |
|---|---|---|
| chart clean · 82% human | The candles were examined and nothing bot-shaped was found in them. When the ranking feed reports bot flow, the tile shows the human share instead; above 55% bots it turns yellow even with clean candles. | Proceed — but see the warning below. |
| 1 bot flag · 2 bot flags | Patterns were found — one is a warning, two or more a fail. The score already carries the cost of them. | Treat the volume as suspect and read the chart yourself. |
| wash traded | The ranking feed flags the pair as wallets trading with themselves. | Treat the volume as fake. |
| run check | Nobody has looked at this token's candles yet. | Press it. The tile is a button — dashed means no answer, not a pass. |
| 3/8 candles · checked: no candles yet | A check ran but could not reach a verdict — it needs eight one-minute candles. | Try again later; very new tokens often have nothing to analyse yet. |
7.2The patterns it looks for
When a pattern is found its name appears on the card, and in the chips under Why this score. They describe shapes in the tape that people do not produce:
| Pattern | What it looks like | What it usually is |
|---|---|---|
| IDENTICAL CANDLES | Candle ranges barely vary across the window. | A single script printing the tape. |
| DEV BUY + BOTS | One enormous first candle, then a run of near-identical small ones. | Deployer buy followed by automated fill. |
| ONLY HUGE CANDLES | Every candle is violent and volume spikes dominate. | No real order book; a handful of wallets throwing size. |
| STAIRCASE | Flat stretches punctuated by repeated jumps. | Timed buy-ins manufacturing a chart shape. |
The lighter layer works on the trading data already on the card rather than the candles, and asks the same question a different way: does the size and the timing of the trades look like a crowd, or like a clock? What it finds is charged against the token as an ordinary penalty and appears in the chips like everything else.
7.3Running the check yourself
When the tile has no verdict yet — run check, or a candle count — it is a button. Pressing it pulls the token's candle history and runs the detectors against the live card. It needs a tradeable pool and a signed-in account; see Chapter 3 — the risk strip for the rest of the strip.
A chart clean result is a statement about the candles that exist, not a guarantee about the token. A launch an hour old has barely any history to examine, and a patient script can draw a convincing chart. Clean here means “nothing caught”, and it is the weakest of the four tiles to lean on.
AlgoLens user manualChapter 7 · Bot detection