Frequently asked questions

Software that turns a trading goal written in plain English into a validated strategy for US index futures — ES, NQ, RTY and YM. Two language agents propose and interpret; an engine in code produces every number; you approve before anything is deployed. You bring your own broker account, so Quant24 is never in the path of your money.

The software is real and it runs. The markets are not. This build is pre-launch: no broker connection, no capital and no exchange market data. Every chart and every metric you can reach is computed on synthetic data and carries a SYNTHETIC badge on the number itself. Where something cannot honestly be computed, the page says "not available in this build" instead of showing a figure.

A Strategy Package is what a run produces. It is versioned and executable: entry and exit rules, the risk envelope the Executor has to respect, the strategic context, and the falsification conditions declared before anything trades.

It carries its own evidence with it — the probability of reaching your goal with a

90% credible interval, the deflated Sharpe, how many hypotheses were tested to get there, and how far the strategy degraded on the sealed holdout. Every number in it is produced by code on synthetic data and labeled SYNTHETIC.

Packages expire. An old one is a record of what was true then, not a standing recommendation.

The Evidence Ledger travels with every Strategy Package. It lists each decision and the reasoning behind it, the alternatives that were discarded and why, every source with its robustness score and timestamp, and the known unknowns the agents declared instead of quietly rounding off.

It also carries what you need to run the whole thing again: seeds, skill versions, and the hash of the data feed.

The point is narrow and deliberate. You should be able to audit how a conclusion was reached, not only read the conclusion. Results in this build are simulated and labeled SYNTHETIC.

The ladder decides how much capital a strategy is allowed, and it reads that from the evidence rather than from how good the idea sounds.

Below a synthetic probability of 55%, a strategy stays in PAPER: it runs in simulation and accumulates evidence without a dollar exposed. Between that and 70% it can be assigned a small share of the capital you allocate. Above it, and only if the stress test is survivable, it can scale. The top level is never granted on simulated evidence alone; it needs a stable posterior from real execution.

The ladder moves down as readily as up. In this build every level that touches real capital is blocked with the note "requires broker connection".

One word of caution about names: PAPER here is a rung of the ladder, not the Paper plan on the pricing page. The rungs are a product mechanism, and none of them is something you buy.

Yes. That is copilot mode. You describe the rules you already trade and the Strategist turns them into a Goal Spec instead of inventing a thesis for you.

From there the run is identical: the same six validation engines, the same sealed holdout, the same auditor checking for look-ahead, survivorship and data snooping. You get the whole diagnosis, including the parts that fail — a mediocre result is delivered with its numbers, and only a dishonest process gets sent back to design.

In this build the run happens on synthetic bars, so read the output as a check on your method, not as a verdict from the market.

Minutes, not weeks — but read that as the shape of the product, not as a measurement. It is the design target: work a person would spread over two weeks of spreadsheets should fit in one session, because the language agents decide what to test and an engine in code runs it.

Nobody has timed it, because there is nothing to time. The engine is not built yet, so any number here would be invented, and an invented duration is exactly the kind of figure this site refuses to print. The product document declares a headline figure; we do not repeat it.

What we will do instead: when the engine exists, each run measures its own phases and shows the duration it actually took, next to the result it produced. That is the only duration you will ever read from us, and it will say whether it ran on synthetic data or against a real feed.

The trader we build for is pragmatic about AI: glad to use it to inform a decision, unwilling to hand it the wheel.

Account sizes make the case for discipline on their own. The average retail trading account holds US$ 25,000PROXY — an indirect figure standing in for one we do not have^F16F17F18, and around 70% of retail traders open with US$ 2,000 or lessPROXY — an indirect figure standing in for one we do not have^F16 — at that size, position sizing decides more than raw horsepower does.

If you already trade a system, copilot mode audits it. If you are starting out, Sandbox lets you test an idea before it costs anything. Quant24 is software you operate. It is not advice, and it is not somebody managing money for you.

Yes. Futures are leveraged instruments, and a loss can exceed what you deposited. The record of the segment is blunt: around 70% of day traders lose moneyPROXY — an indirect figure standing in for one we do not have^F16F17F18, and validating a strategy does not reverse that. What the ladder does is narrower: it keeps the amount at risk tied to the evidence that exists, the guardian caps exposure before every order, and nothing is deployed without your approval.

Yes. Futures are leveraged instruments: the margin you post is a performance bond, not the size of your exposure. A fast move against an open position can produce a loss larger than the balance in the account, and your broker can require you to make up the difference.

Stops reduce that exposure. They do not remove it, because gaps and thin liquidity fill you where the market is, not where your order was.

Quant24 is designed to size positions from the evidence behind a strategy and to enforce daily, weekly and monthly loss limits. None of that is running in this build: there is no broker, no capital and no market data, so no limit is protecting anything today. The account, the leverage and the obligation stay yours in any case.

No, and treating it as though it were is the most expensive mistake in this field.

A simulation is prepared with hindsight over data that already happened. It cannot reproduce the fill you would have got in a thin book, the spread that widened for eleven seconds, or a regime the sample never contained. It also does not charge you for hesitation: hypothetical trading involves no financial risk, so it never tests whether you would have stayed with the plan through a drawdown.

That is why validated results are labeled SYNTHETIC, why the holdout stays sealed, and why the ladder starts in simulation and moves up only on evidence from real execution.

It is the risk manager's escalation, and it sits above every strategy.

Three nested limits, measured on assigned capital: 2% in a day, 4% in a week,

8% in a month. The first stops new entries for the session. The second cuts size. The third takes the strategy off the book until you review it.

The Executor cannot argue with any of them and no layer can route around the risk manager, which holds a veto. You can also revoke an approval at any time — that is your control, and it is not the same thing as the kill switch, which is the automatic drop to execution plus risk when the strategic agent stops adding measurable value.

Limits are configuration, not physics: they cap intent, and a gap does not ask permission.

It means the number beside it was produced on simulated data, not on real market data and not on real fills.

In this build that is every number the product shows. No market data feed is licensed, no broker is connected, and no order has ever been sent. SYNTHETIC and REAL never appear in the same view, and a quantity that cannot honestly be computed is shown as "not available in this build" rather than filled in with something plausible.

When real execution begins, those results carry the REAL badge and the change is recorded in the build ledger. A number with no badge at all is a bug — tell us and we will fix it.

Because it is the honest frame for everything else on this site.

Around 70% of day traders lose money.PROXY — an indirect figure standing in for one we do not have^F16F17F18 We publish it with its evidence level and its sources, and we read it for what it says rather than for what it could be made to suggest. It is a PROXY: an industry statistic standing in for a measurement we do not have. It describes a segment.

It is not evidence that Quant24 changes that outcome for anyone. What it does justify is the discipline: test on data the design never saw, size positions by probability, and put a person in front of every deployment.

A stop is an instruction to leave. It is not a promise about the price at which you leave.

When a market gaps over your level — around an economic release, at the session open, in a liquidity hole — you are filled where the book is, and that can be materially worse than the number you typed.

We mitigate it with historical slippage cushions by instrument and hour, and with circuit breakers that stop trading when conditions stop resembling the ones the strategy was tested in. The residual risk is never nil.

Anyone who describes a stop as protection has told you something about their marketing, not about the market.

Sandbox is free and stays free: no card, no broker account, no waiting. Paper costs $29/mo, and the Live plan starts at $79/mo with a waitlist instead of a checkout, because the broker connection does not exist yet. Those are the published prices; no plan is on sale yet. Founding cohort pricing locks in for life.

Three reasons, in order of how much they bind us.

First, the legal one: taking a share of a client's profits is the structure that turns a software company into a registered adviser, and US rules keep it away from most retail clients anyway. Quant24 is not a registered investment adviser, and the way to stay honest about that is not to charge like one.

Second, incentives: a cut of the upside pays us when you take more risk. A subscription pays us when the process is worth keeping, which is the thing we actually control.

Third, arithmetic: you can read the price before you sign up, and it does not change because a month went well.

It means the price you join at is the price you keep on that plan, for as long as your subscription stays active — including after the public catalog opens at a different number.

It is not a discount that expires and it is not a coupon. It is the compensation for joining a product with no track record and telling us, every week, what is wrong with it.

Two limits. The lock follows the plan you joined on: move up later and the new plan costs whatever it costs then. And it lives in the founder contract, which a lawyer reviews before anyone operates under it — the binding version is that document, not this page.

Plugins are the dials, sold one at a time, on whatever plan you are already on: an extra active strategy, extra creations, a faster review cadence, around-the-clock coverage. They exist so that a single missing number never forces an upgrade.

Each one is priced against what it costs us to run, so the discount is generous where our marginal cost is near zero and disappears where it is not — the fastest review cadence is sold at no discount at all.

What you cannot buy as a plugin is access to real capital. That is the fence between plans, and it stays a plan-level decision on purpose.

Nothing breaks and nothing is deleted. Strategies that are already running keep running under the same risk limits; the action that crossed the quota is the only thing that waits.

You then have three ways out: buy the plugin that raises that particular dial, wait for the monthly counter to reset, or move up a plan if what you actually need is the other side of the fence.

What does not happen is an overage charge. Crossing a quota is a decision you make in the app, with the price in front of you — never a line you discover on an invoice.

Only on Paper (Paper plan, annual: $290/yr). No other plan has an annual option, and that is on purpose.

Governance rule four of our own pricing document says an annual plan does not open until we have at least three months of our own churn data. Asking someone to commit to twelve months on a product with no track record is a promise we cannot back yet, and taking the money first would be the easiest way to stop noticing.

When the founding cohort has produced that data, annual billing opens on the rest of the catalog. Until then the answer stays no, including for people who ask for it.

No. Quant24 Inc. is not a registered investment adviser, a commodity trading advisor, a broker-dealer or a futures commission merchant.

We sell software that you operate. Nothing on this site or in the app is personalized investment advice, and no output is a recommendation to buy or sell anything — the Strategist session carries a notice saying it is a simulation for educational purposes. We do not custody funds and we are never in the path of your money: the account stays at your own broker.

A consultation with a futures lawyer on how the CTA definition applies to this product is open and unresolved. We will publish the answer when we have it.

No. Every page under /legal is marked DRAFT — pending legal review, and that label is literal rather than decorative.

They were written by the team from the product as it actually works, and they are waiting on a qualified lawyer. Until that review lands they describe our intent accurately, but they are not a reviewed agreement, and they carry noindex so they do not circulate as though they were.

Nothing is on sale in this build, no account holds capital, and no data reaches a broker, so the practical exposure is small. We would still rather say this out loud than let a draft pass for a contract.

As few as the site can work with. Non-essential cookies stay off until you accept them: the banner opens in the rejected state, and "Reject non-essential" sits next to "Accept all" in the same weight.

What runs regardless is what the site cannot function without — the internal access cookie on private routes, the anti-abuse check on forms, and your session once sign-in exists. Site analytics is cookieless and loads only after consent. Payment pages run on the processor's own domain under their cookies, not ours.

The full table, marked with what exists today and what a later phase adds, lives on the cookie policy.

Only what you hand us. A form asks for an email address, a name, and whatever the specific request needs — a country, a channel, a message. It is stored so that a person can answer you, and it is not sold.

We never ask for and never accept bank details, card numbers, broker passwords or account credentials. When broker connections exist they will run through the broker's own authorization, so your keys stay with your broker.

You can ask us to delete what you sent by writing to b.caldera@quant24.io. The privacy policy describes the rest — in draft, pending review by a lawyer.

Three things.

Secrets, first: passwords, API keys, broker or bank credentials, card numbers. We do not accept them and a form is the wrong channel for them in any case.

Second, anything that needs an answer in minutes. Forms are read in business hours by a person, not by a rota.

Third, anything legal or regulatory — a subpoena, an inquiry from a regulator, a formal complaint, a data breach report. Those are not answered by whoever opens the inbox; they go to a named person and, where relevant, to outside counsel. Write to b.caldera@quant24.io, say what it concerns, and we will route it instead of improvising.

CFTC Rule 4.41 prescribes the wording that has to accompany hypothetical or simulated performance results.

Every number this product is designed to produce is simulated, so the notice accompanies all of them, on the site and in the app, reproduced word for word rather than paraphrased.

What the prescribed text says, in short: these are simulated or hypothetical results, and they carry certain inherent limitations. Unlike an actual performance record, they do not represent real trading. Because the trades were never executed, the results may have over- or under-compensated for the impact of market factors such as a lack of liquidity. Simulated programs are also designed with the benefit of hindsight. And no representation is made that any account will achieve, or is likely to achieve, results similar to the ones shown.

That summary is here to make the notice readable; it does not replace it. The rule also requires the prescribed text itself to be disclosed prominently and in immediate proximity to the results it qualifies, which is why you meet it beside the numbers rather than only here. The full text has its own page, with a link to the regulation itself.

Two, and their jobs are narrow on purpose.

The Strategist reads your goal, turns it into a measurable specification, decides what is worth testing, and interprets the results. The Executor runs the approved package with machine discipline and adapts execution to conditions.

Between them sits a risk manager written in code, holding a veto neither agent can route around, and an auditor that reviews the process rather than the outcome.

Two language agents propose and interpret; the engine in code produces every number; a human approves before anything is deployed. That is the whole architecture. There is no third agent quietly trading in the background.

The code, always.

A language model decides what is worth testing and reads the outcome. It never reports a metric it produced itself. Backtests, walk-forward passes, Monte Carlo reorderings, stress tests and the Bayesian posterior all run in a deterministic engine, from seeds and versions recorded so the run can be repeated by someone else.

A backtest that is narrated rather than executed is not a backtest.

If a quantity cannot be computed honestly, the interface says "not available in this build" instead of showing something plausible. That rule makes the product slower to impress, and much harder to fake.

It does not move money. Nothing reaches a broker without your explicit approval, and in this build there is no broker to reach.

It does not set the risk envelope. The daily, weekly and monthly limits, the position caps and the kill switch are configuration in code, and no agent can widen them.

It does not decide that a strategy has earned capital. That comes from the validation engine and the ladder, and it can be revoked as easily as it was granted.

And it does not advise you. It is a tool you point at a question, not a person with a duty to you.

We assume they happen and build so that they cost less.

A language agent proposes hypotheses and interprets output, and both are checked. The engine computes every number in code. The auditor runs a deterministic checklist for look-ahead, survivorship, unrealistic costs and a compromised holdout. A second model runs an adversarial pass against weak mechanisms and against analysis dressed up as evidence.

Sources carry a robustness score and a timestamp, and five outlets repeating one wire are deduplicated into one piece of evidence. A hypothesis with no declared economic mechanism does not enter the package.

What survives is auditable. A confident sentence is not.

Because using AI and delegating to it are different things, and the gap is measurable.

62% of surveyed retail investors use AI to inform investment decisionsVERIFIED — checked against a primary source^F20, while 31% fully trust AI for financial adviceVERIFIED — checked against a primary source^F19 and 26% would let AI manage their investmentsVERIFIED — checked against a primary source^F19.

We designed for the majority sitting in that gap rather than against it. You define the goal, you see the evidence, and you approve or refuse — per deployment, not as a blanket permission granted once at sign-up.

It is also the honest answer to the question of who is responsible for the decision.

Not for Sandbox or Paper — those run on simulated markets. For real capital, yes: you bring your own account at a futures broker or a prop firm, which is what BYOK means here, and Quant24 never holds or moves your funds. Broker integrations are planned, not built: no provider has been contacted yet, so the connection screens are waitlist screens.

Bring your own account.

You keep the account you already have at your FCM or prop firm. Quant24 connects to it and is never in the path of your money: we do not custody funds, we do not hold a balance for you, and we do not take a cut of a trade.

Market data reaches you through your own broker entitlement rather than being redistributed by us. That is the licensing rule, and it is also the cheaper answer for you.

What we sell is the software subscription. Close your Quant24 account and your broker relationship is untouched, because it was never ours.

None today.

The shortlist is Rithmic and CQG first, then Interactive Brokers by authorization and Tradovate, because between them they sit behind most retail futures accounts in the United States.

Every one of those is a planned integration, marked planned wherever it appears, and the honest state is that none has been contacted, no vendor agreement exists, and no certification has begun. We would rather publish that than a wall of logos implying something we have not signed.

Broker connect is designed as a complete interface in a "coming soon" state. It is built in a later phase, and it will be declared as a mock in the build ledger when it exists. When a route becomes real, the changelog will say so.

From the broker connect screen in the app, or from the contact form if you have not signed up yet.

Tell us which FCM or prop firm you actually trade with. That is the only input that changes our order of work, and it is worth more to us than a general vote of interest.

Joining costs nothing and commits you to nothing. It is a list of email addresses and broker names, not a queue with a price attached.

We will write when an integration genuinely works, not when one is close. If a name on the list is one we cannot support, we will say why.

That is the intent. A large share of retail futures activity now runs through evaluation accounts, and an evidence-gated ladder maps onto their rules better than most tools do.

Two caveats, stated before you count on it.

Some prop firms require that orders originate from your own device, which rules out a server-side connection and forces a different design.

And their drawdown rules are theirs. A firm can close an account for breaching a trailing limit at a moment when our own risk manager is perfectly comfortable.

Until an integration works and we have read the specific rulebook, treat this as planned rather than supported.

The markets.

"US regulated futures" describes where the contracts trade. US index futures are listed on CME Group exchanges, and those exchanges — along with the brokers and clearing firms around them — answer to the Commodity Futures Trading Commission. That is a fact about the venue, not about the vendor, and it is what the phrase on our home page means.

It is not a statement that Quant24 is registered. We are not a registered investment adviser, a commodity trading advisor, a broker-dealer or a futures commission merchant, and we hold no licence or approval from any regulator. Software that points at a regulated market does not inherit the market's status. There is more on that in are you a registered investment adviser?.

Whether you can trade those contracts from outside the United States is your broker's question, not ours. Access depends on the rules your FCM or prop firm applies to your country of residence, and on whether they will open an account for you. Quant24 connects to an account you already have: we cannot create one, and we cannot widen who is allowed to hold one.