Narrative Scalper v4.0.0

---

name: Narrative Scalper slug: narrative-scalper version: 4.0.0 description: Research crypto narratives at speed and run disciplined scalp trades. v4.0 hunts two pools — fresh launches AND recent runners — with continuation-entry rules, trend-quality scoring, and the same mandatory safety gates. Experimental: no guarantee of profit, not recommended unless you're crazy like the author. category: research

---

Narrative Scalper — public playbook v4.0

A complete system for researching fast-moving crypto narratives and running disciplined scalp trades. Built from studying the best trading bots in the game (Freqtrade, Hummingbot, the Telegram snipers, aixbt-style narrative trackers) and rebuilt around one lesson: in memecoin scalping, the edge is not dying. Most scalps lose. The winners are whoever loses least on the ones that fail.

v4.0 widens the hunting ground: fresh launches AND recent runners are both first-class setups. Runner entries are continuation scalps with their own rules — trend intact, volume confirming, invalidation close by — and chasing vertical blowoffs stays banned, always.

This is a generic playbook. It holds no wallet, no balances, no positions, and no personal trading data. Keep your own private ledger.

It does not promise gains. Never trade money you can't afford to lose.

Purpose

Research crypto narratives at speed, track social posts for momentum, and run quick in-and-out scalp trades with strict risk rules. v4.0 keeps every defensive upgrade and the autonomous scheduled loop, and hunts two pools: fresh launches and recent runners. Momentum continuation gets its own entry rules and scoring — chasing vertical blowoffs is still banned.

No hype, no holding losers. This skill does not promise gains.

Authority gate (non-negotiable)

The skill does NOTHING live until the user has given, in this conversation:

1. A funded wallet you control, confirmed working.
2. Total bankroll for this desk.
3. Max risk per trade (default: 1–2% of bankroll).
4. Daily loss limit (default: −5% of bankroll, then desk closes).
5. Authority for the specific session/action — from whoever owns the money, in writing.

A wallet-setup approval is NOT blanket trade approval. When numbers are missing, research and paper-track only.

Standing authority (optional, for scheduled operation). The money owner can grant standing authority in writing — "trade autonomously within these guardrails until I say stop." The grant must lock, at minimum: max risk per trade, daily loss limit, max concurrent positions, permitted chain(s), and the suspicious-contract rule. With a standing grant, the desk can run on a timer (section 12) without per-trade approval. The grant never overrides the guardrails, and the owner can revoke it at any time. Without either session authority or a standing grant, research and paper-track only.

0. How the system thinks

Signal detection and execution are separate jobs. Detection finds narratives worth a look and hands execution a structured ticket: token, entry price, size, stop, take-profit rungs, time stop. Execution fills it mechanically. Never improvise mid-trade — the ticket IS the trade.

Every trade is tagged with the strategy version that took it (v4.0.0, v4.0.1, ...). No version tag, no trade. Without this you can't tell what worked when you change the system.

1. The narrative lifecycle

Every narrative moves through five stages. Each wants a different action:

1. Whisper. A few sharp accounts mention something new. Volume tiny. Action: research only. Score it.
2. Discovery. Posts pick up, unique accounts grow, price moves. Action: entry window — if it passes the checklist.
3. Momentum. Everyone's talking, price climbing fast. Action: momentum entries are allowed — but you're late to the move, so size it as a scalp, not an investment: tight ATR stop, full exit ladder attached, short time stop. "Already pumping" alone is never the reason to skip OR to enter; the gates and the score decide.
4. Euphoria. "Can't lose," rocket emojis, normies arriving. Action: sell into it. Euphoria is the exit sign, not the invitation.
5. Distribution. Post velocity falls while price chops. Action: be gone already.

The most expensive mistake is entering at stage 4 because the chart looks exciting. The chart always looks exciting at the top. Stage 3 is tradeable with a stop; stage 4 is where you take profit, not where you start.

2. Ignition detection — is this actually starting?

A narrative is tradeable at ignition, not at noise. Require all three:

1. Coincident independent mentions. Several unconnected accounts mention the same ticker inside a short window. One viral post is marketing; five strangers is ignition.
2. Attention acceleration. Don't count mentions — measure whether mentions-per-minute are speeding up versus the token's own baseline. Speeding up = ignition. Flat-but-high = already priced in.
3. On-chain cross-check. The social spike must line up with fresh-buyer or liquidity anomalies on-chain. Social without chain flow is just talk — skip it.

Track smart money by fresh-wallet funding fan-outs, not labels. When a known winner's new wallet makes its first buy into a contract under 24 hours old, that's the earliest readable signal. Measure your copy latency in blocks: more than ~10 blocks behind the smart wallet and you're their exit liquidity — skip.

Continuation entries (recent runners). For tokens already running, ignition is replaced by continuation — require all three:

1. Trend intact. Higher highs on the 5m/15m, no vertical blowoff in the last hour. A stair-step climb is tradeable; a single near-vertical candle is stage 4 — skip the entry.
2. Volume confirming. Volume supports the move — rising or steady on pushes, drying up on pullbacks. Price climbing while volume fades = exhaustion, skip.
3. Defined invalidation close by. A recent structure level (last pullback low, VWAP, breakout retest) sits within 1.5–2x ATR so the stop has a real anchor. No level = no trade.

Prefer pullback-to-structure entries over chasing a green candle. Chasing vertical IS the euphoria entry — banned, always.

3. Scoring (20-point check, trade only at 14+)

• Freshness (1–5): under 48 hours old scores high. For runners, score Trend quality instead: orderly stair-step with volume support scores high; vertical blowoff or fading volume scores low.
• Spread (1–5): crossing from crypto-native to wider audiences? One circle talking is an echo chamber.
• Story strength (1–5): explainable in one repeatable sentence?
• Liquidity (1–5): depth to enter AND exit fast without pushing price against yourself. Check depth, not market cap.

One freshness/trend slot per token — launches get Freshness, runners get Trend quality, never both. The binary gates in section 4 run BEFORE scoring. A token can score 20/20 and still be untradeable if a gate fails.

4. Contract and holder gates — binary, before anything else

Run these before scoring. One fail = no trade, no debate. Never touch a coin whose contract looks suspicious in any way that could stop a sell. Any doubt = automatic skip. No exceptions, no "just this once."

1. Buy+sell simulation. Simulate the buy AND the sell against live chain state before broadcasting. If the simulated sell returns less than ~50% of the expected output, it's a honeypot or sell-tax trap — skip. This one check blocks more losses than every other gate combined.
2. Holder concentration, cluster-adjusted. Merge wallets linked by shared funding first, then require: top-10 holders at 60% of supply or less, no single wallet over 10% (excluding burn). Reject tokens where launch-block buyers hold most of the supply.
3. Owner privileges + proxy scan. Flag mint(), blacklist, mutable tax. Resolve proxies to their implementation and check who controls upgrades — "ownership renounced" means nothing behind a proxy admin.
4. Tax ceilings at execution time. Re-check buy/sell tax right before sending — contracts flip taxes minutes before launch. If tax exceeds your max, wait for it to drop. Never buy into a punitive early-block tax. Skip block-0 when the contract punishes early blocks — wait out the tax window instead of racing it. Auto-shrink the buy to the contract's max-buy.

5. Execution — how orders go out

• Route through a private relay by default (MEV Blocker or Flashbots Protect RPC). Sandwich attackers live on your slippage; private routing is free and kills the biggest silent tax on small scalpers.
• Slippage per leg: tight on exits into liquid pools (~0.5%), wider only for fresh-launch entries. Never 5% on a thin pool — that's directly extractable.
• Split large entries instead of one fat market buy that moves the book against you.
• On Base there is no public mempool — watch mined blocks and factory PairCreated events, not pending transactions. On Ethereum, use the private relay for exits and normal flow; bundles only for truly speed-critical entries.
• The exit ladder is attached at entry (section 7). Never enter without the ladder programmed.

6. Entry checklist — one "no" = no trade

• [ ] Passes all four contract and holder gates.
• [ ] Launch: ignition confirmed (coincident mentions + accelerating attention + on-chain cross-check). Runner: continuation confirmed (trend intact + volume confirming + invalidation close by). Euphoria verticals are never entries.
• [ ] Scores 14+/20 (Freshness for launches, Trend quality for runners).
• [ ] Size from ATR math (section 8), within max risk per trade. Never "just this once."
• [ ] Exit ladder written into the ticket BEFORE entering.
• [ ] Portfolio heat after entry within max heat (5–10% of bankroll in total open risk).
• [ ] Open positions under the max (default: 2 concurrent scalps).
• [ ] Drawdown governor checked: not in cooldown, size not halved.
• [ ] Strategy version tagged.
• [ ] Session authority or a standing grant is in hand.
• [ ] Momentum note: "already pumping" is not a skip reason — the gates and the score decide. Chasing a vertical candle IS a skip reason.

7. Exit rules — the ladder, mechanical

• TP1 (+20–30%): sell a third to half. Move stop to breakeven.
• TP2 (+50%): sell another portion.
• Runner: trail with a Chandelier exit — highest high minus 2–3x the 1m/5m ATR, ratchets up only, never down. Let it run until the trail or the time stop takes it.
• Stop: 1.5–2x the 1m/5m ATR at entry, not a fixed %. A fixed 10% stop on a coin that swings 40% an hour is pure noise.
• Time stop 1–4h per ticket (shorter for momentum entries): hasn't worked = hope, not a trade. Out.
• Thin pools: exit in small chunks. A full-size sell into a thin or chaotic book reverts and costs more than patience.
• Euphoria: sell into it. Euphoria is the exit sign.

Never move a stop further away. Never average down. Never turn a red scalp into a "long-term hold."

8. Risk — sized by volatility, governed by drawdown

1. ATR sizing: position size = (bankroll x risk%) / (ATR x multiplier). Wilder coins get smaller size automatically.
2. ATR stops: 1.5–2x the trade-timeframe ATR. Calibrate with MAE journaling: if every winner went 1.2x ATR against you and still hit target, your stop multiplier is too tight.
3. Portfolio heat: cap total open risk at 5–10% of equity. Every entry checks heat-after against max heat.
4. Drawdown governor: 3 consecutive losses = halve size and take a 60-minute cooldown. Restore full size only at a new equity high-water mark. Daily ladder: −2% changes nothing; −3–4% halves size and caps the session; the daily loss limit closes the desk and needs a written return condition.
5. Single order gate: every order passes one gate function (authority, heat, governor, checklist). Plus a max-orders-per-day circuit breaker.
6. No leverage. Spot only.
7. Fractional Kelly (after 20+ live trades per setup): once you have measured edge, size up to a fraction of Kelly (start at quarter-Kelly, hard cap 5%) instead of flat 1–2%. Never before the data exists — Kelly on guesses is just leverage with math.

9. Ledger and weekly review — the calibration instrument

Keep a private ledger — one line per trade:

`{date, strategy_version, narrative, entry, exit, size, atr_at_entry, pnl_usd, pnl_pct, r_multiple, mae, mfe, minutes_held, close_reason, tx_hashes, notes}`

• R-multiple: profit or loss measured in units of initial risk. A +2R win and a −1R loss are comparable across coins.
• MAE (max adverse excursion): the worst the trade went against you. MFE (max favorable excursion): the best it went for you. MAE of winners tunes your ATR stop multiplier.
• Weekly metrics: expectancy = win rate x avg win − loss rate x avg loss (in R; above 0.2R is acceptable), profit factor = gross wins / gross losses (1.5–2.0 is good), SQN (above 2.0 is good), avg R after losses (catches revenge trading).
• Kill what doesn't work. No ledger, no edge.

10. New variants paper-trade first

New signal or execution variant? Paper-trade it live for 20+ trades before sizing up. Memecoin backtests lie: dead tokens vanish from the data (survivorship bias), historical prices you couldn't have gotten, look-ahead bias. Live paper data doesn't lie. If you must backtest: tick data including dead tokens, with fees + slippage + one-bar delay baked in.

11. Alerts

One alert channel (Telegram). One message per event: signal ticket created, TP1/TP2/stop hit, daily P&L summary, governor trip, kill-switch trip. De-dupe with cooldowns. A scalper you can't see is a scalper you can't trust.

12. The desk loop — running it on a schedule

With standing authority (see Authority gate), the desk runs itself on a timer — every 30 minutes is a good cadence for memecoin scalps. Each run, in order:

1. Read the ledger tail. Reconstruct open positions, today's realized P&L, consecutive losses (governor state).
2. Check bankroll. Confirm free quote currency before anything else.
3. Manage open positions first. Price every open scalp and apply its exit ladder mechanically: TP1 → move stop to breakeven; TP2; runner trails highest-high minus 2–3x ATR (ratchets up only); hard stop at the ticket's ATR stop; time stop per ticket — not working = out. Never move a stop further away. Never average down. In thin or chaotic pools where full-size sells revert, exit in small chunks.
4. Scan for entries only if: open positions under the max, daily P&L above the loss limit, not in governor cooldown, enough free bankroll. Hunt two pools: fresh launches (<48h, real volume) AND recent runners (strong 24h/6h, volume confirming, trend intact, no vertical blowoff). Gates first, then ignition (launches) or continuation (runners), then the 20-point score (Trend quality replaces Freshness for runners). Euphoria verticals are never entries. One "no" anywhere = skip. Never force entries to look busy — "always be trading" means always be scanning.
5. Size, tag, log. ATR sizing within max risk; tag the strategy version; log the ticket.
6. Report. A private, detailed briefing after every run: positions managed, entries/exits with prices and P&L, every candidate scanned and why it was skipped or taken, bankroll state. Public reporting stays honest and minimal: one daily recap with the day's numbers — no individual calls, no wallet addresses, no balances, no holdings. Never invent fills; log only what the tools returned.

13. Deliberately excluded

• Latency arbitrage / colocation: the edge is physical fiber and FPGAs. Not us.
• CEX market-making: needs two-sided inventory and quote infrastructure. On-chain memecoins are AMMs anyway.
• Funding-rate arbitrage: needs perps venues and size. What we steal from the institutions is their risk discipline — heat caps, governors, kill-switches. Free, no colocation needed.

14. The ways scalpers lose

1. Entering at euphoria.
2. Skipping the simulation gate.
3. Moving the stop loss.
4. Averaging down.
5. Trading without the user's risk numbers.
6. Revenge-trading past the governor.
7. Running an unversioned, unlogged variant.
8. Forcing entries when nothing passes the gates.
9. Chasing a vertical blowoff as an entry (stage-4 euphoria).

Check before every session.