Where Signals Earn Capital.
ShikumiBot's agentic workflows capture edge by shepherding signals through a validation pipeline to live sleeves.
Our trend signal proof of concept graduated the validation pipeline.
The next version benefits from a regime filter and upgraded validation pipeline.
yielding more robust signals over time.
Become a development partner.
Deploy capital, develop signals, or rent access to the Signal Factory.
Robustness at every stage.
Each signal only advances by proving robustness in each phase.
- Pre-registered hypothesesThe hypothesis, the variants and the result that would kill it are committed before any result exists. Every variant tried is counted.
- Causal backtestsStanding tests prove every booked price was knowable on the bar it was booked.
- Point-in-time dataDelisted names stay in the data, and every universe is defined as of its date.
- A true out-of-sample testA reserve window stays sealed through research and is opened once.
- Research and live code agreeA forward test runs the frozen code on new data and must make the same decisions as the research code, event for event.
- A time-stamped recordThe forward test records every decision when it is made, so the record stands as written.
- Proven account mechanicsA paper sleeve runs sizing, slot limits and order logic against a simulated account before any capital is at risk.
- Execution matched to the modelEvery fill is reconciled to its booking, so slippage and fees are measured by exit path.
- Bounded riskEach sleeve starts private, inside a heat budget, with a retirement condition fixed in advance.
- Capacity before scaleCapital grows only after market impact is measured at size.
- Traceable numbersEvery public number traces to a frozen artifact and names its evidence class. Thin samples show as locks.
- Current claimsLabels follow the current record, and superseded figures are retired rather than quoted.
Independent audit
Auditing workflows operate independently from operating workflows.
Rules set in advance
A person decides at every stage, against rules written down before the results came in.
One signal factory. Two ways to trade.
Every signal comes out of the same validation pipeline. How it is configured to trade live decides where it goes: a discretionary workflow a trader acts on, or a systematic sleeve that runs on its own.
Signals built and tested in the validation pipeline, ready to trade.
Discretionary workflow
For signals configured to trade with a person in the loop, often rarer setups that benefit from context. The system does the watching and the paperwork. The trader makes the call.
- 01The setup is detected and an alert fires.
- 02An agent drafts the trading plan: levels, scenarios and the point where it is wrong.
- 03You decide, size and execute.
- 04The journal scores the trade against its plan.
Systematic workflow
For signals configured to trade unattended, often ones that fire frequently or harvest tails of a distribution.
- 01The frozen engine is pinned by digest at go-live.
- 02At each close, the walker reads the bars and decides.
- 03Positions are sized by slot count and risk per slot, under a notional cap per sleeve.
- 04Every fill is reconciled to its booking, by exit path.
One signal, several sleeves.
A signal decides what to buy and when. A sleeve decides how much. Sleeves of one signal take the same entries and exits; they differ in how many positions they hold and how much each one risks.
- Each sleeve is registered before go-live, with its configuration and a retirement condition.
- Live results report as a time-weighted index. No wallet addresses or balances are published.
- Configuration sets how much capital sits on the exchange. A higher risk per slot can carry similar dollar exposure on a smaller deposit, reducing counterparty risk and freeing the rest of the capital for other uses.
Same entries, same exits, same venue and notional cap. New sleeves follow the next release once it clears the pipeline.
Research workflows for disciplined discretionary traders.
Agentic workflows with a human in the loop increase operating leverage and ensure the trading plan adapts as market conditions evolve.
Configurable risk posture and trading system rules are available to our development partners.
Development partners benefit from early access and exclusive edge.
Three ways to partner.
Deploy capital
Rent signals from the systematic trading book, deployed in sleeves configured to your mandate.
Develop signals
Bring a signal. We take it through our validation pipeline, and if it earns capital, we deploy it in our book.
Rent access
Use our data, validation pipeline and execution to develop and deploy signals with your own capital.