thesis-fund-strategy

name: thesis-fund-strategy description: >- THESIS FUND v1.0 — one configurable engine, many one-tap macro bets. You bring the market VIEW; the fund expresses it with discipline. The THESIS env var selects a preset from thesis-presets.json (e.g. risk_off = bet against the Trump economy, recovery = US risk-on, war_escalation / war_recovery = Iran/US/Israel, hype_vs_market, gold_over_btc / btc_over_gold). Each preset is a long basket + a short basket that together express the view, held in ONE wallet — but it is NOT a blind hold: each name is only pressed when the market is CONFIRMING the thesis direction, and the DSL + drawdown gate de-risk when it isn't. NOT a copy-trader; the runtime owns the LLM gate (pass-through), DSL exits, and all risk.guard_rails. license: Apache-2.0 metadata: author: jason-goldberg version: "1.0.0" platform: senpi exchange: hyperliquid requires: - senpi-trading-runtime>=1.1.0 - senpi_runtime_helpers

🎯 THESIS FUND v1.0 — bet your view, run with discipline

The Thesis Fund is one engine that expresses any of several macro views. Instead of picking a trading style, you pick what you believe will happen — and the fund trades the basket that expresses it. One producer (thesis-producer.py), one wallet, driven by the THESIS env var which selects a preset from config/thesis-presets.json.

Disciplined conviction, not a hope trade. The fund holds the preset's long/short basket, but it only presses a name when the market is confirming the thesis direction (4h/1h trend + 24h momentum aligned). A long-basket name in a fresh downtrend is skipped; a short-basket name in a fresh uptrend is skipped. The DSL + drawdown gate de-risk when the thesis stops working. You express the view; the engine enforces the discipline.

Presets (the one-tap bets)

Each preset = a THESIS value. Opposing presets are just flipped baskets.

THESIS	The bet	Long	Short
`risk_off`	Bet against the Trump economy / risk-off	gold, silver	SP500, XYZ100, BTC
`recovery`	U.S. recovery / risk-on (the mirror)	SP500, XYZ100, BTC	gold
`war_escalation`	Iran/US/Israel quagmire deepens	BRENTOIL, CL, gold	SP500, XYZ100, BTC
`war_recovery`	De-escalation / recovery	SP500, XYZ100, BTC	BRENTOIL, CL, gold
`hype_vs_market`	HYPE keeps outrunning the majors	HYPE	BTC, ETH, SOL
`gold_over_btc`	Real gold beats digital gold	gold	BTC
`btc_over_gold`	Digital gold beats real gold	BTC	gold

Add or edit presets in thesis-presets.json — no code change needed.

How it scores (confirmation, not blind hold)

Each tick, for every name in the active preset's basket, the producer fetches 1h+4h candles and scores how strongly the market is confirming the thesis direction for that name (the direction is fixed by the preset):

Component	Pts
4h structure aligned with the thesis direction	+3 / +1 neutral / skip if opposed
1h confirmation	+1 aligned / −1 opposed
24h momentum aligned	+2 (≥`momThresholdPct`) / +1 (≥0)
RSI room (not overbought for a long / not oversold for a short)	+1

minScore (4) is the confirmation bar. A name whose 4h structure opposes the thesis is skipped entirely — the fund waits for the view to start working rather than fighting the tape.

Execution & exit

slots 6 (the basket), margin_pct 12%, tick 300s, strict 5x clamp (then venue max)
DSL balanced let-it-work: phase1 max_loss 14% / retrace 9 / 1 breach; weak_peak_cut ON (8h @ 2.0) + dead_weight_cut ON (16h) recycle a name that stops confirming; hard_timeout 5d; phase2 ladder 10%→0 / 22%→45 / 40%→65 / 70%→80 / 120%→90
drawdown_halt_pct 20% halts the whole fund if the thesis is broadly failing.

XYZ handling

Many baskets include XYZ (indices/metals/energy) — candle fetches route with dex="xyz" (the producer keys off the xyz: prefix). XYZ trades 24/7. The main/xyz clearinghouse sections are two VIEWS of ONE cross-margined wallet, so get_positions() takes accountValue ONCE via max() — never sums.

Files

File	Purpose
`runtime.yaml`	Single-wallet runtime spec (DSL, risk, LLM gate)
`scripts/thesis-producer.py`	Preset-driven producer daemon
`scripts/thesis_config.py`	Thesis resolution + preset load + SenpiClient wrapper
`config/thesis-presets.json`	The thesis baskets (edit/add presets here)
`config/thesis-config.json`	Shared tunables (sizing, scoring, leverage)

Operator install

See README.md — set THESIS=<preset>, fund one wallet, launch one daemon.

Hard rule for user-conversation Claude sessions

User-conversation Claude sessions MUST NOT call any of: create_position, close_position, edit_position, ratchet_stop_add, ratchet_stop_edit, ratchet_stop_delete, cancel_order, strategy_close, strategy_close_positions.

These tools are reserved for the producer daemon (entry path) and the DSL ratchet engine (exit path). User-conversation sessions are read-only. The producer daemon handles real signals on its next tick.