DMA/FGI Portfolio Rules v2 — reference strategy
The pinned reference identifier, ordered rules and hypothetical backtest method.
Live Reference strategy: reference/dma_fgi@2#ffc3614028fb. This is an advisory target, not an instruction to trade.
Live 200-day moving averages, crypto and US-equity Fear & Greed, and ETH/BTC, shown in the app next to each decision.
Rules, in priority order
Live 1. Cross-down exit (cross_down_exit). The first eligible rule in priority order sets the advisory target; the others stand down.
Live 2. Cross-up equal-weight (cross_up_equal_weight). The first eligible rule in priority order sets the advisory target; the others stand down.
Live 3. ETH/BTC ratio rotation (eth_btc_ratio_rotation). The first eligible rule in priority order sets the advisory target; the others stand down.
Live 4. ETH/BTC deviation DCA (eth_btc_deviation_dca). The first eligible rule in priority order sets the advisory target; the others stand down.
Live 5. DMA overextension DCA sell (dma_overextension_dca_sell). The first eligible rule in priority order sets the advisory target; the others stand down.
Version boundary
Planned Versioned strategy configs you can hold fixed. Today the config is global and can change in place.
Planned An adapter for the S&P 500 sleeve the reference strategy models. Deposits keep it at 0% until one exists.
Research Version 1’s deployed research slice uses a rule-wide exit cooldown; the version 2 reference strategy uses independent per-asset exits.
Backtest method
The frozen story recording has 500 days, from 2025-05-29 to 2026-10-10.
Signals use daily closing prices; simulated decisions fill the following day, with 0.3% assumed slippage and 3% stablecoin APR. It excludes separate gas and protocol fees.
Planned An adapter for the S&P 500 sleeve the reference strategy models. Deposits keep it at 0% until one exists.
Hypothetical backtest: it fills each trade the day after its signal, assumes a yield on stablecoin balances only, and includes an S&P 500 sleeve Zap Pilot can't execute yet. Past performance does not guarantee future results. Backtest window: 2025-05-29 to 2026-10-10, as of 2026-10-10.