first chart direction predictor for Robinhood memecoins. DIPLAB reads how top holders traded before to call where they will exit - so you see the best spots to enter, accumulate and exit.
Every scanner shows you how much a token's whales hold. None of them tell you what they are about to do with it. You are not losing to the rug you bought - you are losing to the whale who already knew his exit while you were still reading the chart.
DIPLAB answers one question: where does this chart go next, and where do the people holding it get out?
85% direction accuracy across 40 backtested scans (34/40 calls correct) - each token scanned before the move, every case reproducible. Full log: backtest.md.
Paste a contract address. DIPLAB pulls the token's top holders, opens each one's real trade history across every token they touched before, and learns the one thing that actually repeats - the multiple at which that wallet takes profit. Then it projects that habit onto the token in front of you.
The result is a map of walls: the price levels where holders are likely to sell, how much supply sits behind each, how deep the dip under it runs, and a single exit-risk score for the token.
SWARM / WETH · migrated V4 pool · bundled supply 9.51% · EXIT-RISK 76 / OK
wall supply dip read
──────────────────── ────── ──── ──────────────────────────────
0.64x (-36%) 1.8% ~43% 2 whales, habit 0.8x - fast flippers
0.33x (-67%) 1.5% ~38% 2 whales, habit ~0.4x - exit into any pump
0.12x (-88%) 0.9% ~26% 1 whale, deep bag, exits low
→ 4.1% of supply sits in whales who habitually exit below current price.
sell pressure is under you, not above. the scenario arrow points down.
This is a real read, captured live. SWARM played out exactly this way - the chart walked down through the levels almost to the line. DIPLAB has been backtested on real scans, and it reads the next few hours of a chart with high accuracy, because it is not reading the chart at all - it is reading the people who move it.
DIPLAB has been backtested on 40 real scans, one token at a time - each scanned before the price move, with no future data, and the called direction compared against what the chart actually did next on-chain.
- 85% direction accuracy - 34 of 40 calls correct.
- up-moves called right: 21 / 26 (81%). down-moves called right: 13 / 14 (93%).
Because the engine is deterministic - the same token at the same block always returns the same read - every case is reproducible: run the address through DIPLAB at the listed block and you get the same call. The full case-by-case log is in backtest.md.
This backtest measures chart direction. The DANGER verdict (bundled supply / sybil-cluster flagging) is a separate capability and is not scored in this run.
The hard part is not the math. It is that on Robinhood Chain you cannot see
who traded. A relayer submits the transaction, so tx.from is almost never
the person who bought. The router routes the buy to a recipient wallet that
isn't the sender either. A naive read attributes every trade to a handful of
infrastructure addresses and learns nothing.
DIPLAB identifies the real trader by the token itself - who actually received or gave up the coin, netted against the market side of the swap. On the bonding curve that trader is named in the curve event; in a migrated pool the swap only names the router, so DIPLAB reconstructs the trader from the transfer legs of the transaction. Getting this right is most of the work, and it is why the reads hold up.
A wallet's habit is the median multiple at which it closes a position - proceeds over cost, across every token it fully exited in the recent window. A whale that consistently sells at 2x has a habit of 2.00x. One that dumps into any green candle sits below 1.0x and is telling you it will do the same here.
The multiple is dimensionless, so it composes across the curve (priced in ETH) and the pool (priced in the pool's quote) without conversion. Wallets with too little history to read are marked unreadable rather than guessed - a habit built on two trades is noise, and DIPLAB says so instead of inventing a wall.
Robinhood Chain tokens are born on a Pons V2 bonding curve and graduate to a Uniswap V4 pool, and those are two different worlds. On the curve, trades and prices come from curve events and the depth is exact arithmetic. In the pool, trades are V4 swaps, price is the pool's own state, and liquidity is spread unevenly across ranges.
A migration detector routes each token to the right engine, so a fresh curve launch and a token that graduated three minutes after birth are both read correctly - and a token that graduated is flagged as past the curve game, not force-fit into it.
Each readable whale's entry into this token, multiplied by their habit, is a price level - where that wallet is likely to get out. Levels near each other group into a wall, tagged with the share of supply behind it. A scenario arrow threads the walls in order, touching each and dipping beneath it, to trace the path the chart is likely to walk.
The dip under a wall is how far that supply's sell-off pushes price, read from the pool's real liquidity. It is a mechanical estimate - a floor. A real dump, with panic on top, runs deeper, and DIPLAB says so rather than dressing the number up as a promise.
Before any of that, DIPLAB checks who bought and who was simply handed tokens. Supply held by wallets that received the token without ever buying it - bundlers, insiders, a dev's own spread - is the clearest tell that a chart is staged. Past a threshold, DIPLAB withholds the forecast entirely and flags the token as risk, because a map of exits means nothing when the holders never had to enter.
Refusing to answer is a feature. A tool that always produces a confident number is lying some of the time.
One number, 0 to 100, 100 clean and 0 danger. It weighs the exit pressure sitting below price, the bundled share, how concentrated the top holders are, the depth of the pool, and the psychological levels where crowds sell. The heaviest weight is on the thing DIPLAB uniquely sees - whether the people holding this token habitually sell into strength or bleed out below their entry.
No private keys. No signing. No transaction path, not for convenience and not behind a flag. The engine only reads the chain, and a CI job fails the build if a single signing primitive ever appears. The RPC endpoint lives in an environment variable and is never committed. DIPLAB cannot touch your money, which is the point of building it this way.
Being straight about the line is more useful than pretending there isn't one.
Solid. Trader attribution across both layers, habit reading, the wall projection, bundle detection, the dip estimate and the score - all run live against Robinhood Chain mainnet and return real data.
Rough. The score leans on wallets having enough readable history, and a freshly launched token whose holders are all new wallets gives a thin read - DIPLAB says "thin" rather than bluffing. The dip depth in migrated pools is a liquidity-based estimate, exact only on the curve. And the map is only as good as the window of history it reads; more history, more signal.
None of this is hidden behind a confident number. When DIPLAB is not sure, it tells you.
◆ SHIPPING NOW
quick wins on top of the live engine.
- deeper scans - read the top 50 holders, not just the top 10
- priority lane - your scan runs first, no queue
- scan history - save your checks and come back to them
- export - take any result out as an image or CSV
◆ NEXT
from a tool into a terminal.
- multichain - beyond Robinhood Chain: BNB, Solana, Ethereum and more
- Telegram bot - the same read from a contract address, in chat
- whale alerts - get pinged when a tracked whale enters or dumps a token
- watchlists - follow the tokens and the wallets you care about
- wallet profiler - paste a wallet, not a token: its habit, its average exit multiple, how smart the money really is
◆ THE LAB
the long build - where DIPLAB becomes a live edge.
- native app - the full DIPLAB terminal as a real application, desktop and mobile
- browser extension - scan any token straight from your browser
- live whale radar - a constant scan of the whole chain: "wallet X just entered token Y" as it happens, not on request
- smart-money index - every whale on the chain ranked by real profit, and what they are moving into right now
- predictive alerts - not "a whale exited" after the fact, but "by their habits, this token is near its exit wall" before the dump
- track record - a public log of the calls DIPLAB made and how they played out
the vision: the pre-trade check every Robinhood memecoin trader runs first - and the radar that shows where smart money moves next.