Rotation plan: basket method, ranking, positions, cycle
4 Oct 2026, Bangkok. Dashboard:
rotation.html. Every result: rule v4.7, 3 baskets at a time, tested on 2023-26 (new) and 2019-22 (old, never used to design). Raw files below.
1. Baskets (5 iterations)
| Choice | 2023-26 | 2019-22 | Verdict |
|---|
| Size: 15 biggest coins (live) | +1849% | +4382% | keep |
| 5 / 10 coins | +251% / +671% | +669% / +1214% | too noisy |
| 17 / 20 / 30 coins | +1409% / +1022% / +869% | +3536% / +4592% / +3234% | 15 to 20 is the sweet spot |
| Weights: by size (square root / capped) | +621% / +226% | +1517% / +3171% | equal weight wins |
| Min daily volume $250k / $1M (live $25k) | +343% / +606% | +2496% / +1477% | keep $25k |
| New coins wait 7 / 14 / 30 / 90 days | +1680 / +1700 / +1782 / +1128% | +3479 / +4368 / +2595 / +1478% | no wait for the signal; the wait belongs to the coins you buy (8+ days) |
| Refresh monthly / daily (live weekly) | +968% / +685% | +3264% / +3979% | keep weekly |
| Number of baskets (random subsets, median) | 5: ~+170%, 8: ~+480%, 11: ~+1000%, 14: +2256% | 5: ~+815%, 8: ~+1300%, 11: ~+2700%, 15: +4427% | more baskets = more chances: added ETH memes, ETH AI, BNB memes, BNB AI |
Also fixed: staking, wrapped and LP tokens and $1 look-alikes are kept out; every Monday's joiners and leavers are logged and shown in each basket's ? pop-up.
2. Which basket first (50 more ways)
None beat "strongest 30 day move first" in both cycles (live: beats 19 of 20 random orders in 2023-26, 15 of 20 in 2019-22). Closest: % above the 50 day average (19 / 18) and 30 day move per worst drop (19 / 15). Joining a basket that is already running: worse in both cycles. Swapping weekly to a stronger basket: worse. File: rank50_*.txt, latejoin_*.txt.
3. Breakout windows 10 / 20 / 30 / 60
Live (30 day high, 20 in a hot market) +2256% / +4427%. 10 day: +464% / +336%. 60 day: +520% / +1233%. Blends of several windows (money split): +878% to +1163% / +1357% to +2657%. Verdict: keep the live window; blends lost in both cycles. File: windows_*.txt.
4. How many positions
3 baskets at a time is best (2023-26: 1 slot +1674%, 2 slots +1636%, 3 slots +2256%, 4 +1288%, 6 +571%; worst drop -24% with 3). Inside a basket, fewer coins = more luck: 2 coins trail the full basket by 42 points in the worst tenth of trades, 6 coins by 26. Best use of 6 positions: 2 coins in each of the 3 baskets. File: kslots_*.txt.
5. Trading cost at ~$12k per coin
Real quotes (Jupiter, KyberSwap, STON.fi): big coins 0.1 to 0.7% to buy, mid coins 1 to 2%, small coins 4 to 30%+. The coin pop-up now shows the live cost at your size and stars the 2 cheapest. Platform fees: fomo 0.5% per trade, BasedBot 1%, Jupiter 0.1%, 1inch 0%, Binance/Bybit 0.1%. TON coins (NOT etc.) are far cheaper on Binance/Bybit than on chain.
6. Cycle
- BTC: top $126.2k 6 Oct 2025, low $57.8k 1 Jul 2026 (-54%, 9 months, the shallowest and shortest bear on record), now ~$84.9k.
- Last bull runs: 2020-21 runs lasted a median 43 days, 2023-25 only 18 days; tokens went from ~11k listed (2021) to 20M+ tracked (2025): money spread thin, rotations fast.
- Strong BTC was not a problem for alt baskets: in 2020-21 and 2023-24, basket buys while BTC was stretched did better; the weak period was the chop after the top (2025-26). File: btcext_*.txt.
- One lead to test next: when ETH/BTC jumped in the 30 days before a basket buy, the buy did worse in both cycles.
7. Monitoring and alerts
Trading Telegram is off (file /opt/rotation/MUTE_TRADING). Self-healing (rotfix) every 10 min from 2 schedulers; independent self-check every hour from 2 schedulers; a failed check first triggers a fresh run and only messages if it still fails 40+ minutes later.
Files
basket_variants_*.txt, varbuild.py, compact.py, rank50.py/_*.txt, windows.py/_*.txt, latejoin.py/_*.txt, kslots.py/_*.txt, btcext.py/_*.txt, baskcount_*.txt, rank_test.py, idle_test.py, idle2.py, bench_eval.py
Version 1, 4 Oct 2026. Latest: http://159.223.88.168/hx-c9ois/edge/rotation-plan-2026-10-04/