What Shishin is.

Shishin (四神) is a quantitative research publication. Before each US market open it publishes a ranked list of US-equity setups generated by a four-engine system that changes behaviour with the market regime. Shishin publishes research. It is not a broker, not an automated bot you hand your account to, and not personalised investment advice. You read the work and decide for yourself.

What we publish

Every trading day: a ranked list of setups, each with a composite score, a setup-state read, and the indicator context behind it; a one-line read on which market regime is in force; and, on a regular cadence, long-form research explaining the method. Everything is an observation, not an instruction. The responsibility for any trade is yours.

Four engines, four regimes

Markets are not one environment. They are several, interleaved. A breadth-driven macro classifier reads the market each day and hands the book to whichever guardian fits. Only one is awake at a time; they never trade at once.

• Genbu (玄武, North): quality leadership when the trend is broad and strong.
• Suzaku (朱雀, South): momentum and breakouts when participation is healthy.
• Byakko (白虎, West): a defensive posture when breadth deteriorates.
• Seiryū (青龍, East): selective recovery setups as a bear exhausts.

The full architecture is in Four engines for four regimes.

There is a fifth guardian. In the classical scheme, the four directional beasts surround a center held by the Yellow Dragon, the beast of the ground itself. That is Koryu, our crypto measurement house: the same discipline, published formulas and a daily attested record, applied to a fifth market. It measures crypto; it does not predict it.

How it began

Shishin began with a loss: a discretionary bet that unravelled, and the decision never to trade on impulse again. What replaced it was a system: rules instead of feelings, a multi-year backtest instead of a hunch, and a hard commitment to run live exactly what was validated on history. The full origin lives in the research library.

What we hold to

• We judge the system on three numbers at once (total return, risk-adjusted return, and worst drawdown), never one in isolation.
• We run live what we backtested. No quiet retuning between the study and the trade.
• We are honest about the hard parts, survivorship, the long boring middle, and the drawdowns that are the price of the edge, not a flaw in it.

The track

Over a five-year backtest (2021-05-12 to present), the four-engine stack held a Sharpe of 2.44 with a worst drawdown of -15.94%, across the whole window, not a cherry-picked stretch. We lead with the risk-adjusted figures because they’re what survives contact with live markets, and we run live exactly what we tested: one locked revision, no retuning. The hypothetical compound return over that window was 137.2% CAGR, net of fees. Past performance does not predict future results.