This article explains how the Shishin system allocates a single shared portfolio across its four engines. It is educational and general only. It is not personalised investment advice, and nothing here is a recommendation to buy, sell, or hold any security. Live results will differ from any description of the mechanism.
People often assume Shishin runs four strategies at once. It does not. There is one book, capped at about twenty positions, and on any given day at most one guardian is on duty inside it. A macro read decides which one, or decides that none qualifies and the idle cash goes to the ETF sleeve. This is how four regime-specialist engines share a single portfolio without cannibalising each other: they take turns.
The short version
Shishin is one shared book of at most about twenty positions. A breadth-driven regime read deploys exactly one guardian per day, Genbu, Suzaku, Byakko, or Seiryu, or deploys none when no engine qualifies, which happens on roughly a third of days. Idle cash, on those days and in the gaps between positions, goes to a trend-filtered ETF rotation sleeve, unless the sleeve’s own off-switch has it in cash too. The engines never all fire at once. Diversification comes from rotating across regimes over time, not from blending four books together.
One book, one guardian, or cash
The whole system points at a single portfolio. There is no Genbu account and a separate Suzaku account; there is one book, and the four engines are candidates to run it. Capacity in that book is finite, capped at about twenty positions, so the allocation question is not “how much do we give each engine” but “which engine is allowed to fill the slots today.”
Each trading day resolves to one of three states:
- A guardian is on duty. One of the four engines is active and is the only engine touching the book. It selects, sizes, and manages positions up to the shared cap under its own rules.
- A different guardian is on duty. The regime read has flipped, and a different single engine now owns the book. The prior engine’s open positions are handed into the new regime under the same shared cap, not run in parallel.
- No one is on duty. No engine qualifies, so the system adds nothing and lets existing positions run down under their own exits. The freed cash drifts into the ETF sleeve, or stays in cash when the sleeve is switched off.
Publication is near 9:35 ET, shortly after the open, once the regime read for the day has resolved and the on-duty guardian (if any) has produced its list. What the board shows is whichever guardian is active, and the names it is working. The naming of the four is covered in the four-guardians framework; the exact map from regime to guardian lives in the regime decision table.
Why turn-taking, not blending
The obvious alternative is to run all four engines simultaneously and blend their picks into one portfolio, giving each a quarter of the book, or weighting them by some confidence score. We do not do this, and the reason is structural rather than aesthetic.
The four engines are regime specialists. Genbu is built for patient quality accumulation, Suzaku for expanding-breadth momentum, Byakko for defensive selectivity when the broad tape is hostile, Seiryu for the transition off weakness. These are not four flavours of the same idea; they are designed to be right in different worlds. In any single regime, at most one of them is actually in its element. The other three, if forced to trade, would be deploying capital into conditions they were never meant for.
Blending has two failure modes that turn-taking avoids: slot cannibalisation, where four engines split too few slots, and averaged conviction, where a strong regime signal is diluted by three weak ones.
- Slot cannibalisation. A blended book has to split about twenty slots four ways. That leaves each engine too few positions to express its edge, and it forces out-of-regime engines to consume slots that the in-regime engine could have used better. Turn-taking gives the one correct engine the whole book.
- Averaged conviction. Blending dilutes a strong regime signal with three weak ones. The point of the macro read is to identify which single posture the market is rewarding; averaging four postures together throws that information away. Turn-taking preserves it: the regime read is a switch, not a dial.
Taking turns means the engines never compete for the same slots on the same day. There is nothing to reconcile, because only one of them is ever holding the pen.
The cash state
Sitting out is a first-class outcome, not a failure. On roughly a third of days the regime read does not resolve to any guardian: breadth is ambiguous, no engine’s conditions are met, and the honest answer is that none of the four has an edge worth deploying. On those days the system adds no new stock positions, and the idle cash sits in the ETF sleeve (or in cash, when the broad market is below its 200-day average and the sleeve’s off-switch is on).
This matters because a system that always trades will always find a reason to trade, and most of those reasons are noise. Reserving the right to do nothing is how the book avoids forcing an out-of-regime engine into the field just to stay busy. The cash state is also the reason there is no discretionary override pulling the book out of the market in a panic: the exits and the regime read already produce a cash-heavy posture when conditions warrant it, without a human hand on a switch. We describe why we run no portfolio-level circuit breaker in no kill switches.
How the book is sized inside a regime
Once a guardian is on duty, the shared book still has to be filled position by position, and not every name gets the same dollar weight. Sizing is driven by two things, described here qualitatively rather than by any parameter.
- Risk. A position’s size scales down as the name’s own volatility rises, so that a wide, jumpy stock and a tight, orderly one contribute a comparable amount of risk to the book. This keeps any single name from swinging the portfolio out of proportion to the others.
- Conviction. Within the on-duty engine’s own ranking, higher-scoring candidates can carry more weight than marginal ones. The book leans toward the setups the active engine likes most, without letting any one of them dominate the cap.
The net effect is a book that is concentrated where the active engine has the most conviction and the risk budget allows, and thinner where it does not, all inside the same roughly twenty slots. The full logic of sizing off risk and conviction is its own topic in position sizing by conviction.
Shishin is the orchestration layer
There is no separate program arbitrating between the engines. The orchestration, reading the regime, choosing the single guardian or choosing cash, handing the shared book from one engine to the next, is what Shishin is. When the board shows a guardian and a list of names, it is showing the output of that one decision: which engine owns the twenty-slot book today.
This is worth stating plainly because it is the crux of the whole design. Shishin is not a wrapper around four independent products that happen to share a brand. It is a single allocator that owns one portfolio and rents it, one day at a time, to whichever of its four specialists the regime currently favours. The board is a live readout of that rental.
The honest limit
Turn-taking buys coherence at the cost of concentration. Because exactly one guardian is on duty, the book on any given day is a single regime bet: if the macro read is wrong, or the regime turns faster than the read updates, the entire twenty-slot book is exposed to that miss, with no offsetting engine running the other side. A blended book would hedge some of that intra-day; a turn-taking book does not. What offsets it instead is spread across time rather than across engines: the rotation from one guardian to the next as regimes change, and the no-engine state on roughly a third of days when no engine qualifies, which keeps the book from staying pinned to a single wrong bet.
The diversification this design offers is across time and regimes, not across simultaneous strategies. Over a long enough window the book has been Genbu, then the ETF sleeve, then Suzaku, then Byakko, and the smoothing comes from that rotation, not from holding four uncorrelated sleeves at once. On any single day, the reader should understand the book for what it is: one engine’s best expression of one regime, and nothing else.