What data we use
- On-chain data from Robinhood Chain. Blocks, transactions, token transfers and liquidity-pool reserves, read through our own indexing and infrastructure providers.
- Public posts from people acting publicly. Token calls posted by callers, KOLs and founders on public channels. We don't publish private individuals' activity, and we never publish Telegram group membership lists.
- Published research, always attributed. For example, our wash-trading tracker quotes Bitquery's investigation until our own measurements go live.
Follower-achievable returns
Every grade on Amplified is based on what someone following a wallet, caller or signal could realistically have made, not on the lowest and highest price. See follower-achievable return.
| Assumption | Launch default |
|---|---|
| Entry | The price 60 seconds after the buy, call or alert, not at the same moment |
| Trade size | $500 |
| Costs | A 1% trading fee each way, plus price impact calculated from the pool's reserves at that size |
| Outcome | Measured at fixed times after entry (1 hour, 24 hours and 7 days), never at the peak |
Grading wallets and callers
- Minimum history. No grade until at least 10 graded trades or calls. Below that we show "Not enough history".
- History is always shown. Every grade states its sample size and date range, e.g. "n=38 · since 12 Aug".
- Misses count. Losses and misses are included at full weight.
- Recency. Recent behaviour counts more than old behaviour, because wallets change hands and strategies drift.
- Conflict flags. For callers, we flag buying before a call and selling into it.
Detecting linked wallets
Blockchains don't label who owns what, so links are inferred from behaviour. We link wallets using the patterns below.
- Shared funding. First ETH received from the same source, especially shortly before trading.
- Coordinated timing. Buying or selling within seconds of each other, repeatedly.
- Matched amounts. Identical or near-identical trade sizes, including one wallet buying and another selling the same amount.
- Repeated co-trading. The same wallets appearing together across many tokens.
Linked wallets form a group, and a group counts as one actor everywhere on Amplified, including in holder checks, volume estimates and Convergence. Links are probabilistic, so each group carries a confidence level. See linked wallets.
Convergence
A Convergence alert fires when two or more independent groups (graded wallets or callers) buy the same token within 15 minutes, and the token's danger level isn't High or Extreme. Groups linked to each other or to the deployer don't count as independent. See convergence.
Danger level
Token Check sets a danger level (Low / Elevated / High / Extreme) from the factors below. It always shows the reasons behind the level.
- Contract. Permissions (mint, pause, blacklist, upgrade), taxes and hook fees, honeypot simulation.
- Liquidity. Pool depth against market cap, lock or burn status, sell impact.
- Holders. Concentration measured by group (not wallet), early-buyer and bundle share.
- Activity. Share of volume from linked wallets trading with each other.
- People. The deployer's history, and callers' conflict flags.
Upside estimate
Some alerts show an Upside figure. This is the probability, estimated by a statistical model trained on graded history, that the token gains 100% before it falls 35%. It is always shown next to its base rate (how often that happens for all comparable tokens) and next to the danger level, never alone.
We'll publish a calibration chart, showing how often the stated probabilities came true, once there's enough Robinhood Chain history to make it meaningful. Models can be wrong, and a probability isn't a promise.
Sell impact
"A $1,000 sell moves price ≈2.1%" is calculated from the pool's current reserves and fee tier at the stated size. It ignores other trades arriving at the same moment, so treat it as a minimum.
The Proof Ledger
- Fingerprinted first. Every alert is recorded with a timestamp and a fingerprint, a hash of the alert's contents, before the outcome is known.
- Graded later. It's then graded at 1h, 24h and 7d on the follower-achievable basis above.
- Misses stay on the record, at the same visual weight as wins.
- Verifiable. You can check that an alert wasn't edited after the fact by recomputing its fingerprint from the published contents.
Limitations
- Wallet links are inferences and can be wrong in both directions.
- Robinhood Chain is young, so every record has limited history, and we say how much.
- We can't see off-chain deals, private groups or undisclosed payments.
- Nothing on Amplified is financial advice. Scores describe on-chain activity. They don't tell you what to buy.
Changelog
| Date | Change |
|---|---|
| 28 Sep 2026 | Methodology and launch defaults published |
Found something wrong? Use the corrections page.