ClawStreet

Ships in validation mode, orders disabled by default

A trading strategy you can actually read.

ClawStreet runs an intraday mean-reversion strategy where the rules live in a plain document rather than buried in code. Every entry needs three signals and has to clear four filters. Every decision is written down with its reasoning. And it ships with order placement switched off, so it proves itself on your own market data before it ever risks a dollar.

Ask a question See how it works

Validation mode is the default. Going live is a deliberate edit that you make, not something the system can do on its own.

3 of 3signals required before any entry

4filters that can veto a trade

3:40 PMhard flatten, never held overnight

Every tradelogged with its full reasoning

The problem

Most trading automation asks for blind trust.

You are asked to hand money to a system whose logic you cannot read, whose decisions you cannot review, and whose risk rules you cannot verify.

The strategy is a black box

Rules are scattered across code you did not write. Changing a threshold means a code change, and understanding why a trade happened means reading a stack trace.

Risk rules are assumed, not proven

Position sizing, stop placement, and settlement handling are usually invisible until the day one of them fails. By then the failure has already cost something.

Nothing is validated before real money

Most setups go straight from installed to trading. There is no period where the strategy has to demonstrate, on live market data, that its signals were worth acting on.

How it works

Rules first, evidence second, money last.

The sequence is deliberate and the system cannot skip a step.

Read and set the rules

Entry conditions, filters, stop sizing, position limits, and settlement handling all live in one readable document. Adjusting the strategy means editing prose, not shipping code.

Run in validation mode

The default mode logs every qualifying signal and the outcome it would have produced, without placing a single order. You get a record of how the strategy actually performed on live data.

Decide whether to go live

Only you can promote it, by making a deliberate edit after at least ten trading days of logged signals and a per-trigger win-rate audit. The system has no mechanism to promote itself.

What changes for you

What you can actually inspect.

The point of this design is that every one of these answers is available to you, in writing, before you commit anything.

Question	How it works today	With ClawStreet
Why did it take that trade?	Reconstructed after the fact from logs and price history, if at all.	Written down at the time, with the three signals and four filter checks that led to it.
What is my risk on this position?	Implied by a fixed percentage somewhere in the configuration.	Stop distance scaled to the instrument's own volatility, within a floor and a ceiling you set.
Can I change the strategy?	Find the right file, change the code, hope nothing else depended on it.	Edit one readable document. Dashboard edits are bounds-checked and can only tighten risk.
What happens if a job fails?	You find out when a position is still open the next morning.	Two independent verify jobs run behind it, at 3:45 and 3:50, before the 4:00 close.
Is it safe to run today?	Only one way to find out, and it involves money.	Validation mode logs everything and places nothing, for as long as you want.

The risk model

Where the numbers actually come from.

These are the constraints in the strategy document, not marketing figures.

Entry

Three signals must agree at once: a lower band breach, volume at 1.5 times the time-of-day average, and a momentum divergence. Any two of the three is not a trade.

Veto

Four filters can cancel it regardless: earnings within two trading days, an overnight gap above 5 percent, a spread wider than 0.10 percent, or the broad market down more than 0.75 percent.

Exit

The stop is 1.5 times recent volatility, floored at 0.8 percent and capped at 2.5 percent, with the target set at twice the stop distance before entry.

Everything is flattened at 3:40, with two independent verify jobs at 3:45 and 3:50 behind it. No position is ever carried overnight.

What you get

Risk controls that are visible, not implied.

Every one of these can be read, checked, and changed by you before anything runs.

Three signals, or no trade

An entry requires a band breach, a volume surge, and a momentum divergence at the same time. Any one of them alone is not enough.

Four filters with veto power

Upcoming earnings, an oversized overnight gap, a wide spread, or a falling broad market will each cancel a trade outright regardless of how good the signal looked.

Stops that adapt to the instrument

Stop distance scales with the instrument's actual volatility, within a floor and a ceiling, so a quiet name is not stopped out by noise and a violent one is not held through a collapse.

Nothing is held overnight

Positions are closed before the bell, with a second independent verification job behind the first in case the primary run is delayed or blocked.

Settlement rules built in

Cash account discipline is encoded directly: buys draw only on settled funds, with capital rotation and good-faith violation avoidance treated as hard rules.

Reviews can only tighten

The post-market review tracks performance per signal type and may tighten thresholds. It is structurally unable to loosen them, and every change it makes is logged.

Who it is for

For traders who want to audit the machine.

This suits people who want to understand and control the rules, not people looking for a hands-off return.

Systematic traders

A complete intraday framework with readable rules, so your effort goes into the strategy rather than into building order plumbing.

Editable rule document
Full decision logging
Per-signal performance review

Cautious automators

Validation mode means you can watch a strategy behave on live data for as long as you want before anything is at stake.

Orders disabled by default
Manual promotion only
Complete signal history

Cash account traders

Settlement discipline is a first-class part of the rules rather than something you have to remember at the moment of entry.

Settled funds only
Capital rotation handling
Violation avoidance rules

Under the hood

What you are installing.

Small, readable, and dependency-light by design.

Default modeSignal validation with order placement disabled

Strategy definitionA single readable document, backed up with a timestamp before every change

DashboardLocal web dashboard showing trades, logs, and live rules

Parameter safetyDashboard edits are bounds-checked and can only tighten risk, never loosen it

SchedulingStandard scheduled jobs for pre-market, intraday, and post-market sessions

DependenciesStandard library only, with one optional extra for a desktop status indicator

Broker accessConnects through your broker's own approved integration

HostingNothing hosted and no account required, it runs entirely on your machine

Questions

Read these before you install anything.

Does it trade real money out of the box?

No. It ships with order placement disabled and stays that way until you deliberately edit the strategy document to promote it. That edit is manual by design and the system cannot make it.

Is there a track record?

No, and nothing here should be read as implying one. The project has stayed in validation mode since release and the trade log contains no closed trades. Any performance figures would have to come from your own validation period, not from a claim on this page.

What can go wrong?

The usual and the specific. Strategies that validate well can still lose money live. Scheduled sessions can fail to authenticate depending on your setup, which is why a smoke test through the scheduler is documented as a required step before trusting it. Slippage and fills will differ from what validation assumed.

Can I change the strategy?

That is the point. Every rule lives in one readable document. Edits made through the dashboard are bounds-checked and can only reduce risk, and every version is backed up with a timestamp before it is written.

What does it actually require?

A brokerage account with the relevant automated-access integration enabled, a machine that can run scheduled jobs, and a standard Python install. There is no service to sign up for and nothing is hosted.

What stage is the project at?

Actively maintained and running in validation mode. There is no automated test suite, only a manual smoke test that verifies read access without placing orders. Treat it as a framework to evaluate carefully, not a finished product to switch on.

This is not financial advice and no performance is promised. In live mode this software places real orders with real money, and losses are possible. Validation mode is the default for a reason. Read the full disclaimer supplied with the software, and only promote to live trading using capital you can afford to lose.