---
name: rotation-watch
description: Rotation Watch, the trench-rotation early warning Ben commissioned 28 Sep 2026 - what it measures, the rule it trades, and where it runs. Read before any "is money rotating into or out of the trenches" question, before touching /opt/rotation, and if a Rotation Watch alert is questioned.
sources: [cowork]
aliases: [rotation watch, rotation, trench rotation, trenches alert, rotation dashboard]
---

# The ask, 28 Sep 2026
- [stated] Ben pasted a Telegram post from **James Bull** claiming money rotated out of the trenches into Bitcoin and Top100 this month because of the BTC pump, and that it will flow back once BTC slows. He asked me to check whether that was the case.
- [stated] He then asked for an **alerting dashboard that warns him about major rotations BEFORE they happen**, delivered both as a Telegram message and on a dashboard, mapping past flows, what triggered them, and the patterns, "go as wide and deep as you need".

# The verdict on the claim: it is backwards
- Trench volume ROSE through September. Launchpad volume a day, week by week: 113M, 125M, 162M, 180M. September average +20% over August.
- Alts beat BTC over the same stretch (SOL/BTC +22% since 1 Aug), which is the opposite of a rotation into BTC.
- The only true part: trench share of DEX volume dipped in the first week of September, then reversed within two weeks.

# The pattern, measured over 894 days to 27 Sep 2026
- BTC and trench share move TOGETHER, correlation **+0.476**. Hottest trenches (4.9% share) came with BTC near $112k in Sep 2025; deadest (0.76%) with BTC at $62k in Jun 2026.
- A BTC rally LEADS trench volume by **4 to 6 days**, sign positive.
- **The rule the service trades:** when BTC's 7-day return crosses above +10%, launchpad volume over the next 14 days rose a median **+19%**, positive in **11 of 14** occurrences since Apr 2024.

# What is built
- **`/opt/rotation/rotation.py`** on the kookwatch droplet, hourly cron at :20, SQLite `rot.db`.
- **Dashboard: http://159.223.88.168/rotation.html**, reading `rotation.json` beside it.
- Action words until 2 Oct 2026: BUY TRENCHES / HOLD / STAY OUT / NO CALL (retired, see below). Telegram muted by /opt/rotation/QUIET.

# Trench correction, 2 Oct 2026 (Ben: make the trenches rotation correct, informed by all trading info)
- Test Apr 2024 to Sep 2026, 14 day forwards, launchpad volume vs Solana+Base meme basket prices. Raw: http://159.223.88.168/hx-c9ois/edge/chain-baskets-2026-10-01/ (rotdb_daily.json, trench_words.json, trench_test.py).
- Old BUY TRENCHES fired on 61% of days; meme baskets after it did the same as any day. KILLED as a buy signal.
- BTC +10% in 7d (16 events): launchpad volume up 75% (median +15%), meme baskets up 50% (median -1%). BTC -10% (17): pad up 47% (median -14%), memes up 65% (median +6.2%).
- Trench share 30% above its 30d low (39): pad up 62%, memes 46%. 30% below its 30d high (38): pad 50%, memes 32% (median -6.7%). Any day: pad 52%, memes 40% (median -3.8%).
- Lesson: busier launchpads do not mean higher meme prices. The earlier 11 of 14 rule was about volume only.
- Since 2 Oct evening: rotation.json `word` comes from basket rule v3 (below); `activity` still written, not shown. Trenches panel and the Rotation Map launchpad tile were removed from the page.

# Basket rule v3, 2 Oct 2026 evening
- [stated] Ben: "I don't care about busy launchpads. I care about prices of tokens going up, and the tokens are in baskets." He wants to know when each basket turns bullish or bearish, enters a buy or sell zone, which baskets are risk on or risk off, and which are good rotations; three iterations informed by all Cowork trading insights.
- Rule (rule3.py, chains.py chains-3): RISK ON/BUY = daily close above prior 30 day high closes, BTC above 100d SMA, majors basket (10 OKX majors equal weight) risk on by the same rule. TAKE HALF first time HOT (z>=2, top 20% of 90d range). RISK OFF/SELL = close below prior 15 day low. ROTATE INTO = 3 tested risk on baskets with biggest 30d move.
- Test: 14 point in time CMC baskets (dead coins count), Jan 2023 to Sep 2026, 1% cost: 118 trades, +8.1% avg, median +3.5%, 58% wins, PF 2.65; Sep 2024 on: 77, +4.0%, PF 1.81; 2025 flat (PF 1.06). Weekly clustered 40 weeks PF 2.92. Rotation 3 strongest: +353% vs BTC +400%, worst drop -40% vs -53%, ~26% money in; 3 weakest +112%.
- Killed for baskets: depth buy zones (40 to 80% under 90d high kept falling), MA/momentum states (only cut exposure), lone breakouts (majors off since Sep 2024: 16 trades -10.4%), BTC gate alone (2025 PF 0.41), stops. Hot = warning since Sep 2024 (median -6.3% 14d with BTC on), so half.
- Live 2 Oct close: rotate into Solana apps (DeFi), Bonk.fun, Base AI (all half size, ran hot); also risk on Solana AI, TON, Base apps, majors, Pump.fun; risk off Solana older memes, Base memes; RH/Ink/Arc young, untested. Live scorecard from 3 Oct close (ledger3 in state.json); Telegram BUY / TAKE HALF / SELL / CANCEL / rotation change via chain basket bot; chain-2 BUY alerts off.
- Raw: http://159.223.88.168/hx-c9ois/edge/basket-rule-v3-2026-10-02/ ; handover dash-h4 http://159.223.88.168/hx-c9ois/rotation/dashboard-handover.html . Backups *.bak-v3-* in /opt/rotation and /opt/rotation/chains.
- Sources, all free from the container: OKX daily closes, DefiLlama DEX and launchpad volume, DefiLlama stablecoin supply. It deliberately avoids BTC dominance, which a new listing can move without money moving.
- Under the existing watchdog as `rotation`, staleness 7200s, so a stopped cron gets caught.

# Rotation Map (the CASH/BTC/ALTS basket book), 28 Sep 2026
- [stated] Ben has the Rotation Map framework reviewed by another AI. His standing decisions: keep v2 frozen as the comparison baseline, do not tune new thresholds on the same 2020-2026 periods, treat an early-entry idea only as a shadow experiment judged on future data, and deploy no changes to it without a separate explicit request from him.
- [stated] 28 Sep 2026: Ben wants his Robinhood indices integrated at the bottom of the rotation dashboard (major rotations stay on top) as a basket drill-down: is rotation into Robinhood Chain happening, what on Robinhood is doing well, and whether a basket is running too hot so it is better to scale out.
- [stated] He wants AI Projects and AI Company Pairs merged into one AI basket: "We don't need both."

# Chain baskets and candle closes, 30 Sep to 1 Oct 2026
- [stated] Ben cares mainly about alts, above all the indices on Solana and Robinhood; he wants broad basket coverage on Solana, Base, Telegram/TON and Ink (Kraken's chain), adapted to each chain's own launchpads, so he can rotate early into baskets marked bottomed and turning bullish.
- [stated] He wants the dashboard to flag weekly and monthly closes: heads up and countdown (hours and minutes) 1 day before and 1 day after a weekly close, 2 days before and 2 days after a monthly close; read out about 10 majors one by one, a majors basket, and every index, each with a clear bullish/bearish category, plus an overall verdict on top (very bullish to very bearish), with drill-down charts and a Telegram heads up.