Paper accounts that behave like the real thing: live prices, the real exchange fee, fills that arrive late and sometimes fail. And a public record of every signal we publish, scored net of fees against simply holding BTC, losses included.
See the paper ledger How we score signalsMembers never trade real money through MagicMoneyMachine: no exchange keys, no wallet, no custody. Private and invite only today; paper accounts open to the public after our legal review.
One rule underneath all of it: measure every position net of the real fee, and compare it with doing nothing.
$10,000 of simulated money on live Coinbase prices. Orders pay the real fee, settle a few seconds late at the price then, and fail when the price runs away, so the paper result is close to what a real order would have done.
Every signal is timestamped when it is published and scored at +1 hour, +24 hours and +7 days: net of fees, against BTC held over the same hours. Every signal stays in the record, the losers too. Results are tracked forward from publication, never backfilled.
Clone a paper book, change one rule, and let the same replay judge both versions on years of history, before a single day of paper trading is spent on it.
Launch outcomes, creator history, new-listing results and the measured record behind each signal family: data we compute ourselves from public chain data and licensed sources, with the source named next to the number.
We publish what fails. Most signal families lose money once the round-trip fee is paid. The scorecard for everything: beat a weekly BTC buy, net of real fees, or say we do not.
| Signal family | What happened | Verdict |
|---|---|---|
| Eight classic entry engines (mean reversion, breakouts, divergences…) | About zero gross edge over 13,045 signals; −9.8% per position after fees; 0 of 4 walk-forward quarters positive | Loses |
| Buying what is already pumping on Coinbase | −2.7% at +1 hour and −4.1% at +24 hours after fees; the fee is the loss | Loses |
| Solana launch sniping at first sight | First sighting is the worst entry: −51% per trade across 20,000 simulated runs | Loses |
| Weekly rotation into altcoins | Every rule tried lost to holding BTC out of sample, mostly to fees | Loses |
| A launch model that ranks new tokens | Finds winners (AUC 0.86 on 25,575 unseen launches); keeping the money under a +25% exit fails | Open question |
| A weekly BTC buy | The benchmark every other rule has to beat | Benchmark |
The platform never places an order on a member's exchange account and never holds a member's funds or keys.
Every member sees the same published record at the same time. Nothing here is a recommendation for your situation, and nothing is investment advice.
No win-rate headlines and no earnings claims. A win rate counted per partial sale can look great on a losing position; we report per position, net of fees.
The operator runs a private book with a separate bot. That book is never presented as the platform's record, and it never trades against or ahead of a published signal.
Private and invite only today. Members sign in with Telegram or email. Public paper accounts open once our legal review is done. There is no waitlist and nothing is for sale.
Member login Open in Telegram