🤖 How the SOAG Agent works
The whale playbook — find an old token about to break out, check the catalyst, size it 2–5x, set a stop — turned into a number the agent computes automatically, and the community approves. Experience, encoded.
SURVIVOR breaks out
│
▼
AGENT scores it ──▶ signal + DegenScore + flow ──▶ conviction 0–100
│
▼ (above the bar = a CALL)
📄 PAPER position opens automatically ──▶ tracked to WIN / STOP ← the scoreboard
│
▼ (optional, the real-money gate)
👍×3 in Telegram ──▶ ✅ group-approved ──▶ would fire a real buy (when armed)
The read — how a pick is scored
1
It only watches survivors
Tokens old enough to have rugged but didn't — they cleared the 7-day kill zone, still have liquidity, still trade two-sided. Other scanners chase fresh launches; this reads what already lived. You can't pay to be on this list.
2
Three inputs money can't fake → one number
Every survivor that's moving gets folded into a single conviction score:
SIGNAL is it breaking out / being accumulated right now? (the radar's read)
DEGENSCORE the catalyst proxy — survival age + organic (not painted) volume + LP depth + two-sided flow
FLOW buy-side pressure + momentum — the "if no catalyst, go with the flow" part
You can't pay to be old, you can't paint an organic move, thin LP betrays a fake. So the number is hard to game.
3
What a call carries
entry the market cap it triggered at
target 2x – 5x — higher conviction aims higher (the whale model: enter big, bank the move)
stop −30% — if it drops here the thesis is dead, the call is closed
The lifecycle
4
The agent calls it — and tracks itself, honestly
The moment conviction clears the bar, a paper position opens automatically (🤖). It's then forward-tracked on every refresh: a WIN if the peak ever reaches the target, a STOP if it hits the stop. Losers are counted. That's the scoreboard — proof, not "trust me".
5
The group approves the real ones
Each call also posts as a proposal. React 👍 three times in the $SOAG group and it's promoted to ✅ group-approved. Today that's the badge on the scoreboard; when the hot wallet is armed, that 👍×3 is the gate that fires an actual buy. The agent never pulls the trigger alone — the group does.
A real call, walked through
Here's an actual call live on the board — $清正 — scored step by step:
survivor? 93 days old · 🥇 GOLD · LP intact · two-sided → yes, it earned the look
signal 🌱 ACCUMULATING — quietly being bought → actionable
degenscore A (82/100) → credible, the catalyst is structural
flow buy-heavy book + up momentum
──────────────────────────────────
conviction 0.42·signal + 0.34·credibility + 0.24·flow ≈ 74 / 100
the call enter @ $69K · target 2.9× (~$200K) · stop $48K (−30%)
From there it's automatic: 🤖 a paper position opened, and it's tracked every refresh. If the peak reaches ~$200K it closes WIN 2.9×; if it falls to $48K it closes STOP at a real loss — and that loss stays on the scoreboard. No cherry-picking. Watch this one play out →
Reading the board at /agent
win rateshare of closed calls that hit their target
avg peakaverage best-X every call reached
⏳ proposalspicks awaiting the group's 👍×3
🤖 / ✅an agent paper call vs one the group approved
1.8x → 3.5xwhere it is now → its target
WIN / STOPa closed call — target hit, or stopped out
Real money (staged — currently OFF)
When you choose to arm it: the community chips in SOL, the bot buys the group-approved calls, and — by design — it never pays out. Funds can only ever leave via a swap or to one hardcoded wallet, under per-trade and daily caps. A pre-sign guard refuses anything else before signing.
Right now it's paper only. No SOL is accepted or moved. The track record comes first — the real wallet is a deliberate switch you flip later, not a default.
This is a tool for reading the board, not advice. Survival is history, not a promise. Memecoins go to zero. Never risk what you can't lose.