Onchain intelligence for Solana memecoins
An ML ensemble trained on 25,000+ historical tokens scores every new Pumpfun launch in real time. Wallet track record, KOL detection, rug filters — distilled into one 0–100 number.
ML score, real time
Every Pumpfun token is scored 0–100 within seconds. Calibrated probabilities for 2×, 5× and 10× outcomes — no false moonshot promises.
Wallet intelligence
127K wallets with historical track record. KOL detection, sniper flags, alpha wallets and dev career — readable at a glance.
Telegram alerts
Direct messages the moment a token matches your filters. Custom score threshold, KOL tracking, watchlist sync.
One honest number, from real data.
No hype, no cherry-picking. Every token goes through the same pipeline — onchain data in, a calibrated probability out.
ML ensemble
Four models (LightGBM, XGBoost, logistic regression) vote over 155 onchain features, trained on 25,800 historical tokens. Score computed walk-forward — honest out-of-sample, zero leakage.
Anti-rug system
A dedicated model plus structural filters — dev exit, whale concentration, insider clusters, thin liquidity — estimate rug risk at 0.62 AUC (out-of-sample) and cut real rug exposure by ~40% on high-score tokens.
2× · 5× · 10× odds
A dedicated booster per threshold: predicting a double is not predicting a moonshot. Calibrated probabilities, not promises — a “score 80” means what it says.
AUC out-of-sample · misurata su ~31.000 token con ATH maturo · capacità di ordinare, non “precisione”
Cosa significa davvero uno score alto
Non promesse: la frazione REALE di token che ha centrato il multiplo, out-of-sample, su token con esito maturo.
| Se entri su… | 2× | 5× | 10× | Rug |
|---|---|---|---|---|
| A caso | 34% | 10% | 5% | 26% |
| Score 85+ | 50% | 19% | 11% | 16% |
| Score 75+ | 46% | 16% | 8% | 16% |
| Score 65+ | 45% | 15% | 7% | 17% |
Score 85+ = 1 token su 2 raddoppia (vs 1 su 3 a caso). Numeri veri, aggiornati dal modello ogni notte.
Non indoviniamo le gemme
Le memecoin sono imprevedibili: il risultato dipende da hype e pump che al lancio non esistono ancora. Nessuno può prevederle davvero — e chi promette di farlo mente.
Il nostro AUC reale è circa 0.60. Non è magia: sposta le probabilità a tuo favore quel tanto che basta. È il numero onesto, non gonfiato — chi ti mostra 0.85 sta barando.
Funziona con disciplina: entra solo sugli score alti, evita i rug, usa una strategia di uscita. Il vantaggio è reale ma va usato bene — non è un bottone «diventa ricco».
Two blockchains, separate pipelines.
Same honest method, two independent engines. Each chain has its own dedicated pipeline and database — no mixed data, no mixed claims.
Pumpfun · Solana
The flagship engine. ML ensemble trained on 25,000+ historical tokens, score 0–100 in real time, Telegram alerts. All live stats and AUC on this page refer to Solana.
Robinhood Chain · Ethereum L2
New coverage: 🪶 tokens launched on Uniswap v2/v3/v4. 894 tokens analyzed so far — about 62% reached ≥2×, 74% on v2 pools. Onchain data only for now: the ML score for this chain is still under validation and not public.
Proof, not promises.
The latest tokens scored live by the engine — the same feed that powers the bot. When the outcome is known, we show it: wins and losses.