This article walks through what the Shishin system does on a typical trading morning. It is educational, not personalised investment advice, and the ranked board it describes is research, not a recommendation to buy, sell, or hold any security.
Every trading day, before most people have finished their coffee, the same sequence runs from start to finish: thousands of tickers go in one end and a single ranked board comes out the other, published by roughly 9:35 ET. Nobody decides which names deserve to be there. The pipeline decides, the same way, every day. This is the guided tour of that pipeline, stage by stage.
The short version
Each morning Shishin pulls thousands of US-listed tickers, computes roughly forty indicators per name, reads market breadth to fix the macro regime, deploys exactly one of four guardians for the day, scores every candidate from 0 to 100, ranks the field, and publishes the board by about 9:35 ET. The whole sequence is deterministic and rules-based. The board that appears is the one the system itself trades from.
First, pull the universe
The day starts with the raw material: thousands of US-listed tickers. Not a hand-picked watchlist, not yesterday’s winners, but a broad, standing universe of names that clear basic listing and liquidity conditions. The point of starting wide is that the system should never be in the position of finding only what it went looking for. If a name has the shape of a candidate today, it gets measured today, whether or not anyone was watching it yesterday.
This pull happens pre-market, on settled prior-day data plus the opening context. Running it before the bell means the rest of the pipeline works from a stable, complete picture rather than chasing a moving intraday tape.
How many stocks does Shishin scan each morning?
With the universe in hand, the scanner computes a battery of indicators for each ticker: roughly forty measurements per name, all of them arithmetic on the price and volume history. Trend and position relative to the moving-average stack, several reads on volatility, momentum and thrust over multiple horizons, the tightness and shape of the recent base, distance from the highs, and liquidity. No single one of these decides anything. Together they let the system see each name in the round rather than through one keyhole.
This enrichment step is the same for every name, every day. It is the subject of how a pre-market scan works, which covers what a real scan measures and why breadth of indicators beats any single filter. For the tour, the thing to hold onto is scale and sameness: forty-ish numbers, thousands of names, one recipe.
Read breadth, fix the regime
Before any name is judged good or bad, the system asks a prior question: what kind of market is this? A strong-looking candidate in a deteriorating tape is worth less than the same candidate in a broad advance, so the context has to be settled first. Shishin answers this from market breadth, how much of the universe is participating in the trend, not from the headline index, which a handful of mega-caps can distort.
The breadth read resolves into a macro regime, and the regime is what tells the system which engine belongs on the field today. How breadth becomes a regime call, and why it is built to switch neither too often nor too late, is the subject of the macro regime classifier. The mapping from a given regime to a specific guardian is laid out in the regime decision table.
Deploy exactly one guardian
The regime read selects one, and only one, of four guardians to run that day into a single shared book of at most about twenty positions:
- Genbu, the quality engine, for the regimes where durable, higher-quality names lead.
- Suzaku, the momentum engine and the most-deployed workhorse, for broad risk-on advances.
- Byakko, the defensive engine, for the windows where the momentum engine has no edge.
- Seiryū, the recovery and transition engine, for the turn out of stress back toward participation.
On roughly a third of days, no engine qualifies and no stock engine is deployed; the idle cash goes to the ETF rotation sleeve, unless the sleeve’s own off-switch has it in cash too. That is not a failure of the pipeline; it is the pipeline working. The single most important thing a rules-based system can do in a market it has no read on is to decline to guess. One guardian or none, never a committee, and never a discretionary override of the day’s regime call.
Score, then rank
Now the deployed guardian looks at the measured universe and turns the forty-odd indicators per name into a single number. Every candidate is scored from 0 to 100, a composite that blends the distinct feature families so no single indicator can carry a name to the top on its own. A high score requires several things to be right at once. The mechanics of that blend, why several weighted features beat any single indicator, and why the weights are frozen rather than re-fit, are covered in composite scoring.
The score is what makes the field comparable. Once every candidate carries one, the whole universe can be sorted into a single ranked list, strongest at the top, with the drop-off visible as you read down. The ranking, not just the pass/fail of a filter, is the output that matters.
When does Shishin publish the board?
The ranked list is published as the board by roughly 9:35 ET, minutes after the open. That board is not a marketing artefact assembled separately from what the system does. It is the exact same ranked list Shishin itself trades from: the names on your screen are the names in the book, in the same order, with nothing cherry-picked in or quietly dropped out. That identity between what is shown and what is traded is deliberate, and it is the subject of why the board equals the watchlist. You can check the published record against the live account on our track record and verification pages.
And then the pipeline stops. The list lands, and the hand-off is complete. Shishin has done the measuring, the regime read, the deployment, the scoring, and the ranking, the same way it does every morning. What happens next is not the system’s call. The board is research, ordered and dated and public; what any reader does with it is their own decision.
Why deterministic, and where that bites
Everything above is reproducible, and it lands on the clock: the board is published by roughly 9:35 ET, minutes after the open, every trading day. Feed the same universe and the same data through the same rules and you get the same board, every time, with no human in the loop nudging a name up or a regime call sideways. That is the whole design: a record you can trust is only possible if there is no discretionary layer quietly redrawing it.
The honest limit is that a rules pipeline is only ever as good as its data and its regime read. Bad or late data corrupts the indicators, and every downstream stage inherits the corruption. More consequentially, the regime read sits upstream of everything: it decides which guardian runs, and therefore which universe is even eligible to be scored. A wrong regime call does not fail loudly. It cascades quietly through the entire day, deploying the right engine into the wrong market, and the board will still look clean and confident while it does. Breadth-based regime reads are built to be slow to fool for exactly this reason, but no regime call is right every day, and on the days it is wrong, the tidiness of the pipeline is no comfort at all.