Describe a trading strategy in plain English. QuantumEdge builds it, tests it on billions of real one-minute candles with the frictions real trading has, paper trades it with the very same engine, trades it through your own broker, and lets you license it to others with the verified record attached.
The code that backtests a strategy is the code that paper trades it and the code that trades it for real. Nothing is re-implemented between the test and the money, so the live version is the thing that was tested.
The AI asks only what the algo cannot work without, shows the strategy exactly as it understood it, and writes code that is checked and self-tested before you spend a run.
Fills only after the signal, capped by volume, with slippage, fees and borrow. Delisted stocks included. No information from the future. The market over the same dates beside every result.
One click runs the engine on live market minutes. Replay a past day and it matches the backtest to the cent. Every fill says Simulated.
Link your own account and let the engine trade it through your broker’s rails. Real results sit next to the backtest and the paper run on one chart.
No coding. Type the idea the way you would say it to a friend. The builder asks for the five things an algo cannot work without, only if you left them out, then shows “Here’s your algo” in plain sentences. Change anything in one line until it is right, then build it.
Then it writes the code, checks it against the engine, fixes what the checker finds, and runs it on a small sample in a sandbox. You see the sample trades, every check, and the code itself before the real run.
The first time a strategy exists you name it once. That is the algo: it appears on the Dashboard with every backtest it runs filed under it. The engine also runs on a laptop from the command line, same honesty, your own disk.
“Only S&P 500 stocks.” “All NASDAQ stocks over $5.” “Over $50M cash.” “Gapped up 10% at the open.” “Stocks that file an 8-K.” “Stocks doing a reverse split.” Price, volume, market cap, financials, distance from the 52-week or all-time high.
A 2018 test uses the 2018 S&P 500, including companies later removed. Membership history back to 1996. Company financials are dated by the day they became public; SEC filings by the exact public minute.
Whole-market screens run every day on the data network. A field it cannot screen yet gets a plain answer, never a silent guess, and the result says when a screen left companies out.
Every way a backtest flatters itself is closed by construction, not by a warning label. Every run setting is your own choice, typed in plain words, and the result names the ones you did not choose.
Stocks and crypto in one backtest, each on its own calendar. Long, short or both, with percent, dollar or share sizing and a maximum number of positions. Margin from 1x to 10x. Verified by hand against raw data and by 32 automated test suites.
Every result sits next to the S&P 500 with dividends over the same period (Bitcoin for crypto), with its own worst fall, and a quarter-by-quarter breakdown.
“Too few trades to tell skill from luck.” “One trade made most of the profit.” “Borrowing was not checked.” On the result, never buried.
Every assumption behind the number, costs in dollars, refused and unfilled orders with reasons, and the settings you did not choose. Older results are tagged “old rules, re-run”.
The full trade log and four downloads: trades, summary, parameters and the exact strategy file that ran. Every run stays in your history. Change one thing and run it again.
Nothing is re-implemented. The paper run is the backtest engine fed one minute at a time, live from the market, with the same fill rules as the backtest it came from. Replay a past day through the live path and it comes out identical to that day’s backtest: the same trades, the same equity, the same costs, to the cent.
From the Dashboard’s Paper tab: run it live, or replay any past day while the market is closed. Decisions land one to two minutes after each minute closes.
On every page and in every trade log. A paper result is never a broker execution and never a forecast. Late and missing bars are counted and said on the run.
Crash, stop or reboot: the run resumes from its own journal and tells you it did. Stop asks you to type STOP, closes every position at the last bar, and files the result next to the backtests.
Investors can follow a paper run, always labelled Simulated. An optional mirror echoes each simulated fill to a broker paper account and shows the broker’s fills beside ours.
Everything a strategy produces lives under that strategy. Four views answer the only question that matters: does the real thing behave like the test?
Positions with live prices, every fill with entry, exit, hold time and dollar result, filters for wins and losses, biggest gain or loss, and a date range.
Quarter by quarter, never one blended number. Projections compound the average trade at the strategy’s real pace, labelled illustrative, and project nothing when there are no trades.
Backtest, paper and real on one chart, each its own most recent period of the same length, with a stats table side by side. Click a line to hide it.
S&P 500, Nasdaq, Dow and NYSE; inflation, unemployment and the Fed rate, each explained, with the direction it moved and when it was published.
Rename or archive strategies, move or rename backtests. New algos start from a strategy in the builder, so none is ever empty.
Slide-out menu, scrubbable charts, two cards to a row, sideways tables, light and dark mode shared across both tools.
Every measure recalculates for the window you pick and the view you are in, so the Sharpe you read in Paper is the paper run’s, not the backtest’s. A hypothetical-results notice sits on every Backtest and Compare page.
QuantumEdge never holds a credential that can trade real money. Orders go through your broker’s own rails, under your authorisation, and can be stopped by you alone. We read, compare and show.
Interactive Brokers, Alpaca and 30 more brokers across 7 regions, with a guided setup tested before it is saved. Statements import on link, the rest comes by CSV with the format detected and every problem explained line by line.
Synced every five minutes with up to a year of history on link. Time-weighted returns, deposits kept out of drawdown, live prices between statements, every fill counted exactly once.
Attach a strategy to your linked account and the same engine trades it through your broker’s rails. It starts on paper, goes live when you say so, and a kill switch that is yours alone flattens everything in one tap.
A small script on your own machine reports fills to your workspace over a write-only link that can never place an order. Readable Python, revocable at any time, with a health light.
License an algo without ever selling the code. Subscribe to one and watch it paper trade under your own account before a cent is at risk. The platform produces the record, so nobody can game it.
List an algo. It runs in our sandbox for every subscriber; the code never leaves. The listing shows the platform’s own verified backtest and paper record, every warning included, under a visible hypothetical label. You earn per subscriber; the platform keeps a small fee.
Subscribe, watch it paper trade under your account, then attach your own broker. You keep custody, the creator keeps the code, the platform keeps the record honest.
No performance claim anywhere: not in a listing, not in a blurb, not in a message. Returns appear only on a result page, next to the warnings that qualify them.
Private workspace per account. Nobody sees your strategies, backtests or linked accounts unless you invite them, and every new account is approved by hand.
Name your fund and invite up to 250 investors by email. They land straight in it: real trading in full, backtests hidden, summarised or full as you choose, charts coarsened so trades cannot be reverse-engineered.
Investors never see strategy code, settings or notes. An access log shows who looked at what, and when.
An eight-character code, valid thirty minutes. The holder sees results with their quarters and two backtests side by side, plus a masked trade log: “Stock 1”, no prices, no sizes.
A view counter, and “revoke all” in one click.
A throwaway user, no network, nothing writable, no secrets, memory and time limits. A strategy that tries to phone home simply fails, and the result says how it was run.
Isolation is written into the code and enforced by automated tests; every route’s access rule is tested on every change. Your email never appears in a file path.
Rate limits per address and per account, no hints, hashed passwords, encrypted secrets, and “sign out everywhere” in one click.
Services are watched and restarted, queues survive restarts, paper runs resume from their journal, and a public health check says so.
Automatic certificates on every hostname; services bound to localhost behind the web server, with memory caps and no default secrets.
A privacy policy and terms of use written to be read, linked from every sign-in page and every shared view.
No. Describe the strategy in plain English; the builder asks only for what is missing and writes the code, which you can read and download. If you do code, upload your own Python file instead.
No. The builder, backtests and paper trading need nothing but an account. Linking a broker is for seeing your real results next to the test and, when you choose, for trading through it.
No. Paper runs use simulated money and simulated fills, produced by the same engine as the backtest from live market minutes. Every paper page says Simulated, and a paper result is never a forecast.
No. Fills happen only on minutes after the signal, capped by volume. Slippage, fees and borrow are charged. Delisted stocks are included and nothing from the future leaks in. When one minute touches both your stop and your target, the stop wins. The result tells you how much to believe it.
Your strategy file, your settings and your results are stored against your account and are not readable from anyone else’s. We do not trade off your logic and we do not publish it. Subscribers to a listed algo run it in our sandbox and never see the code.
No. The platform holds no credential that can place an order. Real trading happens at your broker, under your authorisation, with a kill switch that is yours alone.
No. Backtests and paper results are hypothetical and simulated. Nothing on these pages is a recommendation to buy or sell anything, and past or projected results do not predict the future.
Ten minutes from an idea to a tested algo with a record you can share.