The system, engine by engine.
Shishin runs one coordinator, not four products. Each morning a breadth-driven classifier reads the market and hands the day to whichever of four engines fits the regime — or to cash when none does. Below is the single backtested track everyone gets, coloured by the guardian on duty, with each engine broken out in full.
Four engines. One track.
One coordinator, not four products. A macro classifier picks which engine is deployed each morning; the chart below is the single NAV everyone gets — $100k forward-walked over the backtest, log scale — with each day coloured by the guardian on duty. Roughly a third of the time no engine qualifies and the system sits in cash. Select a guardian to highlight its days.
Colour marks the guardian on duty each day. Positions are held past the handoff, so each engine’s economic footprint runs longer than its coloured stretch — switch to Holding period to shade Suzaku’s full window: the 468 days it held a position versus the 383 it was on duty.
Small-cap aggressive momentum breakouts. The workhorse in trending and choppy markets.
Share of gains is this engine’s contribution to the coordinator’s total return. It holds positions longer than it is on duty — 468 days versus 383 coloured on the chart — so its economic footprint runs past its coloured segments.
When the market regime is unreadable — choppy, conflicted, or outside any engine's mandate — Shishin stays in cash. No list is forced. The four engines fire only when their conditions are met; otherwise the prudent position is no position.
One engine at a time, by regime.
A breadth read decides which single engine is deployed each morning — or whether to trade at all. The others stand down.
The classifier reads how many stocks are actually participating, not just where the index closed, and deploys the one engine built for that tape. Positions are held past the hand-off, so an engine’s economic footprint runs longer than its days on duty. And when the regime is unreadable — choppy, conflicted, or outside every mandate — nothing is forced: the system held cash on 457 of 1.258 days (36%) of the backtest. The mechanics are in the regime classifier and market breadth.
Each guardian, in full.
Four engines, four regimes. What each one owns, what it hunts, how it behaves, and what it actually contributed over the five-year backtest.
How the risk is held.
Per-trade and regime-level — not a portfolio panic button. The risk model, not just the curve.
Sized by conviction
Positions scale with the composite score, capped at 25% of NAV — a deliberate four-position floor. A tighter cap concentrates returns but runs too thin a book, so 25% is where it’s held. See Position sizing by conviction.
A locked stop on every entry
Each position carries an ATR-based stop, set at entry and held good-till-cancelled — so the downside is defined before the trade exists, not improvised after.
A concurrent-position cap
The book holds a bounded number of names at once — diversification by construction within whichever engine is deployed, rather than a sprawl of thin positions.
No portfolio kill-switch — by design
Risk is held per trade (the stops) and at the regime level (the cash gate), not by a global drawdown halt. We tested de-grossing into drawdowns and rejected it: the choppiest, worst-looking stretches are among the system’s best earners, so a panic switch cuts the wrong exposure. See the experiments that failed.
$100K → $6.30M over a five-year, survivorship-bias-free backtest — hypothetical, net of fees, and positive in every full year including the 2022 bear (+70%). Past performance does not predict future results. See the full track record →
Research, not advice.
Shishin is a research publisher. We hand you the ranked read and the regime behind it — we never touch your money and never tell you to act. The four engines run the same whether anyone is watching; the decisions are entirely yours.