Skip to content
FAQ

Straight answers

Everything below is the same answer you would get from a person. If your question is not here, send it and we will answer it and probably add it.

56 questions across 7 categories. The uncomfortable ones are in here too: what has never been tested against a real venue, and what is switched off while we wait on legal advice.

The product

What PTQ is, who it is for, and what the one subscription holds.

8 questions

What is PTQ?
A workbench for testing trading ideas, built for people who trade their own account around a day job. You write your trading rules. PTQ tests them against real past market data. It then runs the same overfitting checks professional investors use. Overfitting means a strategy that fits the past too closely and then fails on new data. The ideas that survive go to paper trading, where the orders are simulated. Live trading happens only on your own computer, and only after you arm it yourself.
Who is PTQ for?
People with a day job who trade their own account. You want your ideas tested hard before you risk money on them. PTQ is software, not advice. We are not a broker, and we never manage your money.
Is this a signals service?
No. PTQ never tells you what to trade. You write your own strategy. PTQ backtests it, checks it, and paper trades it. Live trading runs on your own computer, and only after you arm it yourself.
Do I need to be a programmer?
No. You build a strategy from plain rules, not code: when to buy, when to sell, how big the position is, how much you risk. Built-in templates show you the shape of a complete strategy. The one you run is still the one you wrote.
What does the subscription include?
Everything PTQ does, on one plan. In the browser you get the strategy builder, backtesting, the validation gate, and paper trading with every fill labelled simulated. You also get the PTQ OS for your own machine, the Claude connector, the Pine export, and live trading behind the arm step. Compute and support chat run on a credit allowance that is part of the plan.
What are the two halves, hosted platform and local OS?
The hosted platform runs in your browser and does the research: building, backtesting, validation, paper trading. The PTQ OS is a copy you install on your own computer. It is the only place broker keys and live execution exist. Both come with the one plan, and the licence covers the machines you actually use rather than a single seat.
Why are there no testimonials on the site?
Because we do not have real ones yet. When members say things worth quoting, with permission, they will appear. Nothing on this site is invented. That includes the praise.
Where do I start once I have paid?
The how it works page walks the route in order. Download the OS. Check the file against the published checksum. Run the installer. Activate the device with your licence key. Connect a broker if you want one. Then build a strategy or copy one. Nothing later in that list works until the step before it does. That is on purpose.

Quant mechanics

The maths inside: the validation gate, the optional module library, simulation mode and the review loop.

12 questions

Will PTQ make me money?
We make no promises about performance, and you should distrust anyone who does. PTQ's job is the opposite. It uses the same statistics professional investors use to show you when a backtest is likely overfit, meaning it fits the past too closely to trust. Most strategies fail that test.
What is the probability of backtest overfitting?
A published statistic (Bailey, Borwein, López de Prado & Zhu, 2017). It estimates the chance that your backtest looked good by luck rather than by finding something real. Try enough versions of a rule and one of them will look brilliant by chance alone. PBO puts a number on how much of that luck you are looking at.
What is the deflated Sharpe ratio?
The Sharpe ratio scores a return against how bumpy the ride was. The deflated version (Bailey & López de Prado, 2014) corrects that score for how many ideas you tried, for picking the best-looking one, and for returns that are not evenly spread. A raw Sharpe from a backtest flatters you. The deflated one asks whether the score still stands once the flattery is taken out.
What is the out-of-sample holdout?
PTQ hides part of the price history while you design your rules. Your strategy is then scored on that hidden data. Doing well on the part you tuned against counts for nothing here. Every run shows you where the split was.
Why do most strategies fail the gate?
Because most backtests are tuned to the past by accident. Three common causes: using data the strategy could not have seen at the time, testing only on the companies that lasted, and settings fitted to random noise. The gate catches these before money is involved. When a strategy fails, PTQ tells you why. Most of them failing is the product working, not a defect.
Can I trade a strategy that failed the gate?
The gate's verdict is always worked out and always shown. You see it before anything trades. The whole design pushes you down the honest route: paper first, then live only on your own machine, after you arm it yourself. The verdict exists so that the choice you make is an informed one.
What is the quant mechanics library?
A set of optional modules. Each one implements a published piece of quant research, and each one acts on YOUR strategy: testing it harder, or sizing it more carefully. The first batch: purged cross-validation with an embargo, a walk-forward degradation gate, a trading-cost budget measured in Sharpe units, error bars around your Sharpe ratio, a mean-reversion half-life measurement, a stationarity check, and a Monte Carlo drawdown distribution with a ruin bound. Each module names the research it comes from. Its panel states what it costs you as well as what it gives you. None of them is a strategy, and turning one on is not a promise about returns. Several exist precisely to make your results look worse and more honest.
How does a module get into the product?
Through one registry, not through a page of copy. Each mechanic is a single entry. It names its settings, their floors and caps, the research it implements, and its plain-English account of the upside and the downside. The settings screen and the learn panel are both built from that entry, so a panel cannot describe something the code does not do. Every new mechanic ships switched off. You turn it on and see the difference yourself.
What is Simulation Mode?
You run your SAME strategy through backtest and validation under two or more rule settings. Say, with and without volatility-targeted sizing. Then you see the honest difference on the same past data: equity curves overlaid, every metric and validation gate side by side, and a "what changed" list built from the actual firing journal, so every claimed difference traces back to a recorded event. Nothing is sorted by return. A lower return with better validation is shown as the trade-off it is. Every page of it is badged as a historical simulation. It shows what a rule setting did to the past. It does not predict what it will do next.
What is the review loop?
A report on your own trading record. It reads your order journal for the window you ask about and answers four questions, from real rows only. What happened. Why it happened: the engine's per-trade attribution of signal, market regime, slippage, and which rules fired. What the checks said at the time. And what you might try next. The suggestions arrive as ready-made Simulation Mode comparisons you can run. The report never adopts anything itself. The module has no write path, so producing one leaves your strategy and your standing arm byte-identical.
What happens to the review if the engine is offline?
It tells you. The journal half still renders, because that part is local. The attribution and drift halves come back marked unavailable, carrying the engine's own reason. No statistics block is composed to fill the gap. The same rule holds across the product: real, or honestly empty.
Do the modules ever change my strategy on their own?
No. Nothing changes itself. The drift monitor flags a problem. The Kelly module, which suggests a position size, refuses to suggest one at all when your own data cannot rule out having no edge. Acting on either is an edit you make yourself, to your profile or your strategy. Any such edit switches a standing arm off, so a running loop can never inherit a change you did not agree to.

Execution and brokers

Paper versus live, the honest status of the live path, and which venues exist today.

10 questions

How does paper trading work?
In your browser, PTQ simulates practice orders on US stocks. No broker receives those orders and no money moves. On your own computer, you can connect Alpaca's paper API for US equities. Both use pretend money, and every fill is labelled simulated. Practice helps you check how your rules behave beyond a backtest.
How does live trading work?
Live trading runs through the PTQ OS on your own computer, using your own broker keys. Nothing goes live until you arm it, one strategy at a time. Be clear on what exists today: the live path is an experimental single-shot Alpaca order path, never tested against the live venue. Automated execution, support for more than one broker, and production execution safety do not exist today.
Has a real live order ever been placed through PTQ?
Not yet, and we will not pretend otherwise. The Alpaca adapter is written and its refusal paths are tested, but no order has been sent to the venue in live mode. Paper trading is the part we can show you working today. When the first live order goes through against the venue, this answer changes and says so.
Why is live trading local-only?
Because your keys should never leave your machine, and because a live order should need a human. The hosted platform never holds broker keys. The arm step happens in front of you.
What checks does an order pass before it reaches a broker?
One choke point, the same for paper and live, and the same from every surface: the app, the command line, the Claude connector. In this order: the kill switch, your explicit confirmation, the per-trade risk verdict, which can block the order or trim it, your profile's live permission, whether a broker is actually configured, then the arm state and its one-time token. Next the portfolio limits: gross exposure, the per-symbol cap, open positions, the daily loss halt, the max-drawdown stop, orders per day. Last, a fresh quote. If the price has moved past the plan's limit, the order refuses rather than chases. Every attempt lands in the journal, refusals included.
Which brokers and markets?
Browser practice supports US stocks and simulates fills inside PTQ. To send paper orders to Alpaca, connect its paper API from PTQ on your own computer. Crypto exchange connectors are built but switched off pending a compliance answer. They are not available for trading while that switch is off.
Does PTQ trade unattended while I sleep?
On paper, yes. You can arm a strategy and let the loop run it on a schedule. Every order still goes through the same risk checks a manual order does. Unattended live trading is switched off in this build, so the loop runs paper only until that changes. We would rather tell you that plainly than sell you a phrase like set-and-forget.
What is the autopilot control panel?
One screen for the unattended loop. It lists every armed strategy, with its scope, its expiry, and whether each order still needs your say-so. It shows the state of the loop. And it shows each cycle's outcome per strategy, with the engine's own reason: what it opened, held, or blocked, every paper fill labelled simulated. From that screen you arm a strategy, disarm one, pause the loop, or hit a global kill switch that stops everything server-side. Stopping is always one click, from anywhere, with no confirmation. Arming never is.
How do I arm a strategy for the loop?
Deliberately, and on your own machine. You pick the scope and an expiry. You see the exact risk profile the loop will hold you to. Then you type a confirmation tied to that one strategy. A web page cannot do it for you, because the arm step runs locally only, behind that confirmation. Unattended live stays gated off in this build either way, so an armed strategy runs paper until the compliance switch is turned on.
What does the kill switch actually do?
It stops every submission. Engaging it needs no licence check at all, because a safety brake you can be locked out of is not a safety brake. While it is engaged, every order, paper or live, is rejected at the choke point. Releasing it does go back through the licence gate, because turning the brake off is the direction that widens what can happen.

Connector and routines

The Claude MCP connector, scheduled routines, charts and the TradingView export.

6 questions

What is the Claude MCP connector?
It lets Claude talk to the PTQ OS on your machine, over the Model Context Protocol. So you can ask about your strategies, runs and results in plain English. It reads what you ask it to read. It cannot arm a strategy, place a live order, change enforcement, or reach a broker key. Those tools do not exist on the connector's surface at all, so there is no argument you or anyone else could pass to reach them.
Can a message trick the connector into unlocking the product?
No, and it has been attacked deliberately to check. Entitlement is read from the signed licence on the device, never from anything a tool call carries. So hostile arguments get stripped by the schema and ignored by the handler: a fake lease, a claimed admin token, a credit count, an instruction to ignore earlier instructions. The paper-order tool has no live setting to switch. The activation tool cannot mint a licence, because the licence server is the only thing that can issue one.
Can PTQ send me scheduled updates?
Yes, through routines, and only the ones you say yes to. A routine is a real scheduled job on your machine. It reads your actual portfolio, journal or validation results, then sends the update as a notification. The wording is composed with the Claude CLI on your machine, on your own Claude usage. Without that you still get a plain notification. Every routine is listed, and switching one off or deleting it removes its scheduled job completely.
What happens to a routine when I switch it off?
The scheduled job goes away, not just the row in the list. Switching a routine off, or deleting it, removes the entry from your operating system's own scheduler. Check the scheduler afterwards and nothing is left behind. Switching it back on recreates it with the run time your device chose, not a new one invented on the server.
Does PTQ have its own charts?
Yes. Candlestick charts of real market bars, with your strategy's entries, exits, stops, targets and in-position periods drawn on top. The same engine that runs your backtests computes them, so the chart and the backtest can never disagree. If the engine is offline the chart says so. Nothing is ever simulated to fill the gap.
Can I see my strategy on TradingView?
Yes, one way. PTQ draws its own charts, and you can export a Pine v6 study to see your signals on a TradingView chart. Pine runs only there. Execution and risk stay in PTQ. The export is a study, not a strategy script: it plots signals and never places an order. A strategy the engine itself would refuse is refused at export too, with the reason, rather than approximated.

Billing

Price, what is issued when, and availability.

5 questions

What does it cost?
$997 a year, one plan, everything in it. Nothing is issued before payment: no licence key, no credits, no access.
Is there a cheaper tier or a trial?
No. There is one plan and no tier beneath it. Nothing is issued before payment: no licence key, no credits, no product access. What we do instead is spell out on this site exactly what the software does and does not do, before you spend anything.
Why annual only?
Because the work is slow on purpose. Build, validate, then paper trade for weeks before anything is live. Monthly billing sells impatience. A year matches how the product is meant to be used.
Is PTQ available where I live?
Checkout confirms that from your jurisdiction, meaning the country whose rules apply to you. If we cannot serve your region, checkout says so before any money moves.
How does payment actually work?
One payment of $997, in US dollars, through Stripe, covering a year. Card details go to Stripe and never touch our servers. Your licence key is issued when Stripe reports the payment, not when you sign up. That is why creating an account gives you nothing until the payment clears.

Security and data

Where broker keys live, what the licence gate does, what we hold and what you can take away.

8 questions

Do you hold my broker keys or my money?
No and no. Broker keys live on your machine and never reach our servers. Your money stays at your broker. The hosted platform holds neither.
Where exactly are my broker keys stored?
In your operating system's own secret store, on the machine that runs the OS: the login Keychain on macOS, DPAPI on Windows, the Secret Service on Linux. If none of those is available, the keys fall back to a file in your PTQ home directory that only your own user account can read. The product tells you which store it actually used, not the one it would prefer. Secret values are never logged, never returned by a status call, and never placed on a command line where another process could read them.
How do you make sure a key is never sent to you by accident?
The hosted service refuses it. Its credential screen stores metadata only: the provider, a label, the scopes, and a short mask of the key. Send it a field that looks like a secret and the request is rejected outright, with an error telling you keys stay on your machine. The routes that do handle real keys exist only on your own machine, and they refuse any request that did not come from that machine.
What do you actually hold about me?
Your account details, through our authentication provider. The strategies you create and how they are set up. Risk profiles and settings. Paper trades, which are simulated fills. Backtest and validation results. A usage log of which features you used and what credits you spent. Your declared jurisdiction and the acknowledgements you ticked. And support tickets or chat conversations. From Stripe we get payment records: name, email, amount and their identifiers. That is the list, and it is the same list the privacy policy carries.
How does the licence check work?
Your device activates against the licence server with your key, and gets back a short-lived signed lease: Ed25519, signed with a key that only ever exists on the server, valid for a few days, tied to that one device. The OS checks that signature and its expiry on every gated request, so an edited or hand-made lease simply fails. Metered work like a backtest is charged against a credit balance the server holds and decrements. So a copy of the software running away from the server cannot spend anything.
What can PTQ never do?
Move your money, trade on your behalf without you arming it, or see your broker credentials. No route on the hosted platform can arm live trading. No connector tool can. No request field grants it. Arming happens on your machine, behind a confirmation you type, with a scope and an expiry you set. Any edit to your risk profile or your mechanics switches a standing arm off.
Can I export or delete my data?
Yes, and neither is behind the paywall. Export hands you your account, strategies, profiles, routines and arm records. Deletion removes them, and it cascades through everything attached, so nothing is left orphaned to be quietly switched back on later.
Is the download signed?
Not yet. Code signing needs a developer account we have not set up. Until we do, the honest integrity story is the published checksum, which is a short fingerprint of the file. You check the download against it before you install, and if it matches, the file was not tampered with. The install notes also say what your operating system will tell you about an unsigned build. We would rather write that here than let you find it at the Gatekeeper prompt.

Not answered here?

Send the question. A person reads it and answers it. If other people will ask it too, it ends up on this page in your words rather than ours. Before you buy, that includes the questions we have to answer with a no.

PTQ

Still deciding?

Everything the product does is on this site, before you spend anything. Including the parts that are not finished yet.