Signals in. Disciplined orders out. Everything on the record.

Every strategy DocStox runs is chosen and configured by its operator: which signal sources to read, which filters apply, how much capital each strategy may use, and the risk limits nothing may cross. The platform's job is to execute those choices exactly, every time, and to write down what it did, and why, including when it did nothing.

The sources form a menu rather than a single bet. DocStox can watch public disclosure filings from members of Congress and from the large institutional investors whose reported positions move markets: the whales and the movers. It can watch government activity, such as federal contract awards landing on a company. It can watch the options market for unusual behavior: a rising count of open options contracts, or trades far larger than a stock usually sees, measured over a trailing window of roughly eighty to ninety days. And it can watch price itself, ranking stocks by momentum against their peers. No single source is the star; the discipline around them is.

Take one of those sources, congressional disclosures, as an example. Members of Congress are required by law to report their trades; the filings are public and arrive up to 45 days late. Most attention paid to them is noise: screenshots, hot takes, hero-worship. Configured as a source in DocStox, they become a dataset: disclosed amounts stay the ranges they really are ($1,001-$15,000 is a pair of bounds, never a fake point estimate), and candidates are scored by how many distinct members bought the same stock inside a rolling window.

A lone purchase is weak evidence; four members converging on one name in a month is a different kind of fact. But that score is a ranking input only: it never triggers execution and never bypasses the risk layer. The same shape holds for every source the operator turns on: signals propose, rules dispose. Each source exists to hunt for edges the market has not priced in, and none earns the right to trade until it proves itself in honest, survivorship-audited backtests and paper trading. Ambition is welcome here; it queues behind evidence.

Every order originates in a strategy, is ranked by the strategy engine, passes the risk layer, and is built by the execution module. Nothing else may place an order: not an alert, not a monitoring job, not a manual shortcut. If a signal is good enough to trade, it belongs in the ranked pool where the risk layer sees it. If it is not, it is not good enough to trade.

DocStox is designed around the assumption that its own process can die at any moment, and that the most dangerous code path is the one added "just this once." The protections below are structural: enforced in code and infrastructure, not in documentation.

Nothing goes live before it has a reproducible backtest, and the backtest engine is built to make lying to yourself difficult. No numbers from those backtests appear on this page, and that is deliberate: publishing them would invite exactly the inference the system is built to resist.

DocStox is also built to improve with use. Every decision it records becomes evidence, and strategies are periodically re-tested against that growing record. When a refinement earns its place, it earns it the same way everything else here does: by proving itself through the same honest backtesting gate before it may trade.

Disclosure following. Buys on report dates only, whether the filing came from a member of Congress or a large institution, ranked by disclosure clusters; exits are governed entirely by a trailing stop resting at the broker that ratchets up and never down. After a stop triggers, the stock goes on a cool-down list so it cannot be bought and re-bought into the ground, and sale proceeds must settle before the cash is redeployed.

Cross-sectional momentum. Long or flat only. Standard, boring, well-documented relative strength with a volatility filter, rebalanced on a schedule. Deliberately uncreative.

The wheel, an options-income cycle implemented as an explicit state machine. Sell a cash-secured put on a stock worth owning; if the shares are assigned, sell covered calls above cost basis until they are called away; repeat. Every stage is saved and every transition is written to the record before any order is placed, so a crash mid-cycle recovers to a known point.

The trading core answers to no one's browser. Brokerage credentials are encrypted at rest and never leave it; the dashboard sits behind authenticated access with a second factor; and the dashboard itself is a read-only observer: if it goes offline, trading continues uninterrupted, because it was never a dependency. It exposes exactly one action that can change anything: the kill switch, which is built to keep working even when almost nothing else does. And no one has to sit watching any of it. The platform monitors itself: scheduled jobs check the checkers, and anything that needs attention is flagged on the record without a human having to hunt for it.