Rotation dashboard: 5 checks and 100 ideas

4 Oct 2026, Bangkok. Dashboard: rotation.html. Raw coin pick tests: coin-picks-v3.

5 check rounds

Data: every basket's last daily move recomputed from raw CoinMarketCap prices: 12 of 12 match to 0.01%.
Rules: sell lines, 15 day lows, "so far" and buy lines re-derived from each price line: 0 mismatches. One entry (Bonk.fun 11 Sep) is a BUY (BOTTOM), as the rule allows.
Coin picks: found a $1.00 "USD1" meme and a staking token (tsTON) in the lists. Both are blocked now.
Reliability: nothing restarted a stuck job, and a missing coin would have cost a basket -50% every day. Both fixed: self-healing every 10 min from 2 schedulers, and the daily self-check.
Alerts: the old rule could still send a SELL for Bonk.fun that contradicts the new rule. Silenced.

Scoreboard: 21 DONE today, 43 NEXT (worth testing), 36 NO (tested before or not worth the clutter).
Research used: StockCharts RRG and SCTR, IBD market exposure, Weinstein stages, Altcoin Season Index, CoinGecko and Sharpe sector trackers, Artemis, Santiment, FTSE Grayscale sectors, plus papers on crypto momentum (short 2 to 6 week horizon) and trend following (Zarattini et al. 2025).

Reliability

DONE

Self-healing job (rotfix) every 10 min: restarts a stale or hung engine, closes, rotation, drill or map run

Before: nothing restarted the cron jobs; a hung run blocked every later run

DONE

Run rotfix from two schedulers (systemd timer watched by the supervisor, plus cron)

Ben's redundancy rule

DONE

Hard time limits on the hourly engine (50 min) and closes (20 min)

A stuck network call could freeze the feed for hours

DONE

Known-good copy of the page, restored automatically if the page breaks

Page can never go blank

DONE

Daily independent self-check at 08:20 BKK: basket moves recomputed from raw CoinMarketCap, sell lines and words re-derived from the price line

Catches wrong numbers, not just stale files; Telegram only on a failed check

DONE

Message Ben only when the automatic fix failed for 60 min (through alertgate)

Quiet rule

DONE

Red line on the page when a data source failed

Only mark what is off

DONE

Fix: a coin missing from CoinMarketCap counted -50% EVERY day after day 3

Latent bug; tests counted it once. Would have dragged a basket ~3% a day

DONE

Fix: old chain-2 rule could still send a SELL (-12% stop / 14 days) that contradicts rule v4

One open old trade on Bonk.fun would have fired by 13 Oct

DONE

Coin picks: drop dollar look-alikes (USD1 'drip' meme at $1.00) and staking receipts (tsTON)

Found in the live pick lists

DONE

Faster pick filling (8 baskets per run instead of 3)

New baskets waited hours for coins

Engine writes a heartbeat file the supervisor can watch directly

rotfix covers it today

Second data source for daily closes (CoinGecko) to cross-check CoinMarketCap

Only if CMC ever disagrees in the daily self-check

Replay test: rebuild the last 120 days from scratch weekly and compare words

Good, but the daily check covers the live day

Alert dedupe across sessions (same BUY twice)

Ledger already keys on basket + entry day

Coin picks

DONE

10 coins spread from big to small instead of the 5 biggest

Leaning biggest cost 3 to 4 pts per trade in both eras (2019-22 +33.9 vs +38.2%, 2023-26 +7.7 vs +10.6%)

DONE

Live 24h move next to each coin in the pop-up, flag 'up 30%+ in 24h: skip for now'

Applies Ben's skip rule at the moment he buys, not only on the BUY day

DONE

Show which coins were left out and why

Transparency

Quiet on X as a pick filter

Quiet coins lose less often but made the same money; quiet tilt worse than random in 2025-26 (+4.6 vs +12.4%)

Lean to the biggest coins

Costs 3 to 4 pts per trade, both eras (CI crosses 0, same sign as the other session)

Long-wick skip from hourly candles

Needs a candle source per coin (GeckoTerminal, rate limited); today: check the chart

Coin's 1st or 2nd new-high only

Held in the other session on hourly data; needs a per-coin ATH history

Buy the retest (limit order at the breakout level +3%)

Better than chasing on coins; needs a per-coin breakout level

Size by volatility (smaller in wild coins)

Standard in trend following (Zarattini et al.); untested here

One-tap 'copy all 10 contracts'

Handy if Ben buys through a multi-buy tool

Pool size next to each coin

Ben's rule: size and depth are never an objection

Re-pick weekly while HOLD (coins that left the category drop out)

Picks are fixed on the BUY day today

Hide coins with no DexScreener pool

Rare on CMC-listed coins

Holder growth per coin (GMGN)

Paid calls; test first

Smart-money wallets buying the coin

Wallet copy tests lost money (rounds 8 and 9)

Rules

Blend several breakout lookbacks (10/20/30/60 day) instead of one

Zarattini, Pagani, Barbon 2025: blended lookbacks Sharpe 1.5+ on crypto

Two-day confirmation before a BUY

Buying 2+ days late lost money in 2025-26 (rule v4 tests)

Graded exposure (0/25/50/75/100% like IBD)

Sizing ideas only won by having more money in (round J)

Weinstein stage label per basket

Already inside the words (BASING, HOLD, EXTENDED, SELL)

Relative strength vs BTC gate (only buy baskets beating BTC 30d)

RS vs BTC/majors tested in 101 breakout ideas, did not beat

Volatility target per basket

Vol target tested in round 4, no gain in both halves

Monthly momentum rotation (top 3 baskets by 30d)

Holding the strongest daily lost 30-70% in 2025

Stablecoin inflow gate

+2289 vs +2256%, too small

Funding rate / open interest filter for majors

Data free on CoinGlass/Binance; untested

Social dominance spike = top warning per basket

Santiment style; our own X tests found 'loud' mixed

Google Trends / search interest

Liu and Tsyvinski: attention predicts weekly returns (in sample)

DE-RISK MEMES threshold sweep monthly

Already robust across 7 to 15% and top 10/20/30%

Short side (sell basket perps on SELL)

Ben trades spot memes; not requested

Re-entry after a SELL if it breaks out again within 5 days

Covered: any new 30d high close is a new BUY

Weekend filter

Weekend tested for breakouts, no edge

BTC dominance gate

Tested round 4, no gain

Fear and Greed gate

Tested; flips with threshold

Page

DONE

Gold 'Coins to Buy' button

Ben's ask

DONE

Bigger layout on wide screens (1480px, bigger words, taller charts)

Ben's ask; phone unchanged

DONE

'Changed at the last close' line (HOLD → SELL)

Change log, shown only when something changed

Relative Rotation Graph (baskets as dots: leading, weakening, lagging, improving, with 5 day tails)

Best visual from StockCharts; descriptive only, so it must not compete with the words

'vs BTC since buy' only when a basket is behind BTC

Marks only what is off

Days in the current word ('HOLD, day 15')

The 'since' date already says it

Sort cards by action (BUY, SELL, TAKE PROFIT, HOLD, WAIT)

Ben wants chain grouping

Grey out baskets that are WAIT

Less to read; check with Ben

Heatmap tiles sized by market cap

Adds a second way to read the same thing

Altseason gauge (75% of top 50 beat BTC over 90 days)

Breadth gate already does this job

Dark/light theme switch

Not asked

Install as a phone app (home screen icon)

Cheap: web manifest

Tap a chip to jump to its card

Small win on phone

Show the next close time in Bangkok (07:00)

Tells when words can change

Track record page: every live signal, result at 7/30/90 days, hit rate

Live ledger starts 4 Oct; worth a page once 10+ trades exist

Equity curve 'follow every card' vs BTC

Same, after live trades exist

Tooltips on every number

Ben: no tap/hover hints

Explain each word in one line under it

Rules box already does it once

Search box for a coin ('which basket is WIF in')

Useful with 13 baskets

Show basket members count and top coin weight

Detail behind the ?

Sparkline vs BTC ratio

Clutter

One-line 'why' on each card

Removed on Ben's request

Alerts

DONE

BUY alert carries the 10 coins with full contracts

Picks v4 in the BUY text

DONE

Old rule alerts off

No contradicting SELL

Morning digest only when a word changed

Telegram already sends each change; digest may be noise

Alert 1 day before a likely SELL (close within 3% of the sell line)

Heads-up; must not become noise

Alert when the 24h move of a pick passes +30% (skip)

Shown live in the pop-up instead

Weekly report card (signals, results)

After live trades exist

Separate chat per chain

One chat, fewer places

Voice or push notification

Telegram is enough

Baskets

DONE

Solana memes CMC top 15, Solana infrastructure, Base infrastructure

Ben's ask (Robinhood-style set)

DONE

Solana stock memes (stonk.fun)

Hidden until 2 Nov (needs 30 real days)

Base stock memes

Only 1 tagged coin, no roster source

Remove LP receipts from baskets (JLP in Solana DeFi)

JLP tracks SOL/ETH/BTC/USDC and dampens the basket; needs a re-test of the DeFi basket

Hyperliquid ecosystem basket

Big 2025 narrative; CMC tag exists

BNB memes (four.meme) basket

Was in the never-used test set; behaved like the others

Sui and Aptos memes

Check liquidity first

Prediction market tokens

Narrative; small set

AI agents on Solana vs Base split

Already separate baskets

Square-root market cap weights

FTSE Grayscale style; equal weight is what was tested

Tokenized stocks basket (xStocks)

Not crypto rotation

Robinhood chain baskets on the slow rule once they have 120 days

Automatic later

Research

Out of sample on the 3 new baskets (2019-22 data where they exist)

Confirms the rule on infra

Rule on 4h data for faster exits

Not testable on daily basket history

Cross-check daily closes with Binance for majors

Cheap sanity check

Does the pick list still beat nothing after costs (DEX fees, 10 coins)?

10 buys cost more fees than 5; check with $100-500 sizes

Basket breadth: share of members above their 20d average

Warns when one coin carries the basket; needs member prices

Time of day of the daily close (07:00 BKK) vs a 19:00 BKK close

Close time can matter for breakouts

Kaiko style monthly momentum index as a benchmark

Benchmark only

Seasonality (month of year)

Too few cycles

Correlation cap (skip a BUY if 80% overlap with an open basket)

sol_memes and topmemes share most coins; tested in round 4 as no gain, re-check with new baskets

Ladder in over 3 days instead of all on the BUY close

Could lower entry risk; costs if it runs

Use the X 'new voices' warning from the coin studies on baskets

Worked on single coins at ATH retests