about & methodology

a bot that publishes its losses on purpose

MoonRadar isn't a signal group or a moonshot promise. it's a single experiment, run in public, testing one question: can a trading bot learn to stop losing money and prove it?

01

the story

I started MoonRadar because every memecoin tool i tried did the same thing: copy a wallet, hope it wins, never look back. none of them could tell me which of their own signals actually made money.

so i built one that keeps score on itself. it picked Solana because the memecoin pairs move fast enough to generate a real track record in weeks, not years and because the on-chain data is open enough to follow smart money honestly.

the first months were ugly. the bot chased momentum, bought every pump, and bled to in realized paper losses. i could have quietly reset it. instead i let the curator do its job.

on the bot auto-demoted its losing momentum combos no human override and the curve looked like it turned. then i checked whether it could actually sell its winners. it couldn't: ~60% of the "recovery" was tokens with fake liquidity the bot could buy but never exit. the price feed booked them as profits; reality couldn't fill the exit. honest P&L is still deeply negative. catching that in public, correcting the books, and rebuilding around the one signal that survives honest accounting smart money is the entire reason this site exists.

02

the stack

five components, each doing one job, each auditable from the live dashboard.

signal engine

watches every new Solana memecoin in real time for momentum spikes and smart-money wallet movement.

combo curator

scores every signal combination against its own forward P&L. demotes losers automatically.

wallet scorer

ranks tracked wallets by win rate and contribution. probation for new ones, weights for proven ones.

exit tiers

trailing stops, take-profits, and timeouts re-validated weekly against counterfactual exits.

position sizer

scales each trade to combo confidence and recent variance. small on probation, larger on proven edge.

caution system

tags market news as noise, real, or fear so the bot doesn't trade a headline it shouldn't.

sellability gate

checks every token can actually be sold before trading it the fix for the unsellable-token wins that faked the recovery.

honest reconciler

re-checks closed trades against real on-chain sellability and rewrites any unsellable "win" to its true loss. the books stay honest.

03

transparency

all trades public
every signal, open, and close streams to the live dashboard the moment it happens. nothing is hidden after the fact.
no backtests, only forward results
there is not a single backtested number on this site. the drawdown shown is real forward paper P&L. so is the current balance.
every decision logged
promotions, demotions, and the reasons behind them are published to the changelog. you can audit why a combo died.
code: partially open
the signal schema and combo-scoring logic are documented publicly. the live model weights stay private for now.
04

roadmap

live now

paper trading · public track record

every signal, trade, and demotion runs live on the dashboard paper-only, no real capital.

in development

real execution · auto tier

the bot trades a wallet you connect and control, on your risk rules. gated behind the live-trading criteria first.

next

public API

real-time signal and trade endpoints for Signal and Auto subscribers.

later

more chains

the curator logic is chain-agnostic. Base and Hyperliquid are the likely next stops.

exploring

community wallet voting

let subscribers nominate wallets for the tracker scored the same brutal way as everything else.

watch the experiment run

the live dashboard updates every 60 seconds. no login.

open live dashboard