A backtest is only useful if you know how far to trust it.
Quantis Trade is a daily-bar, multi-factor, regime-aware backtesting platform for retail traders who think like quants. The differentiator is not a longer indicator list. It is that the platform tells you when its own answer can't be trusted, and holds back a number when the maths doesn't hold up.
Below: the honest-failure checks behind every result, the data we actually have, and the four surfaces you will spend your time in.
1. Describe
Tell the Strategy Agent what you want in plain English. Intent parsing is explicit: unsupported universes or unparseable phrasing return a clear error, never a different strategy swapped in without telling you.
2. Generate and simulate
The agent compiles your idea into a backtestable DSL and runs it through the Rust engine on real daily bars. The TypeScript verifier independently recomputes the Sharpe ratio and CAGR and flags any disagreement on the result card.
3. Read the warnings, then iterate
Every result ships with its caveats attached: sample-size gates, a time-slice stability check, a luck-adjusted verdict. A run with no real price data is refused, not estimated. Iterate on parameters, compare runs, then follow a strategy in paper trading.
The ten checks on every result
Every result card runs the same ten checks, and says so on the card when one fires. The methodology page explains each one, with the thresholds it uses. If you find a result where one of them should have fired and didn't, that is a bug we want to know about.
- 1. Engine integrityEngine-recompute mismatch detection
A backtest engine can get its own arithmetic wrong. If nobody re-checks the headline numbers, a wrong Sharpe ratio looks exactly like a right one.
- 2. Sample sizeInsufficient-history suppression
A backtest with a handful of bars and one closed trade has almost no statistical power. A Sharpe ratio computed from 12 data points comes with a confidence interval so wide it tells you very little.
- 3. Sensitivity to settingsHow far the Sharpe falls when the settings change
A strategy that looks great with a 20-day lookback but falls apart at 24 days is probably fitted to its parameters rather than capturing something real.
- 4. Data freshnessFactor-data publication-lag disclosure
Fama-French factor returns are published monthly, a few weeks after each month ends. A backtest that runs up to today will always reach past the latest published month.
- 5. Distribution shapeMonthly-bar wide-error warning
Monthly data gives about 12 points a year, so any statistic computed from it is much noisier than the same statistic from daily data.
- 6. OverfittingTime-slice stability, and held-out years for goal search
A strategy that worked in one stretch of history and not since may have been fitted to that stretch. The headline figure, which averages over everything, will mislead you.
- 7. LuckCould it be luck? Every test of the idea counts
Try enough variations of an idea and one of them will look good by chance. A Sharpe ratio on its own doesn't know how many tries it took to find it.
- 8. AI-written textFigures an AI model writes are checked against the data
When an AI model writes "the P/E is 45", you can't tell whether that came from the data, from the model's memory, or from a guess.
- 9. Survivorship biasOnly companies that trade today are in the data
A backtest on today's list of companies leaves out every company that was delisted, acquired or went bust during the period. The survivors did better on average, so the backtest looks better than reality.
- 10. Cost realismTotal return, real costs and taxes
Backtests without trading costs, taxes or dividends misstate what you could have earned. Commissions, spreads, slippage, stamp duty and fixed broker fees all eat into returns, and dividends add to them.
Two more rules, outside the result card
One is about a test that cannot run. The other is about research answers.
If the engine can't load real price bars for the requested symbols and window, the run is refused. The agent says the engine had no real price data for the request and shows no result card, so there is no Sharpe ratio or equity curve built on prices that never traded.
When fundamentals or factor scores can't be computed (sparse data, a ticker outside coverage, a metric the company doesn't report), the research agent says so instead of inventing a z-score, a percentile or a peer rank.
What data we actually have
The platform is positioned for serious quant-style work, so coverage limits are stated up front. If your strategy depends on something not in this list, the system will tell you before you spend compute. The full methodology (definitions, assumptions, recompute formulas) lives at /honest-methodology.
How much is covered
The count below is read from the live list of what has daily prices, each time this page loads. It is never typed here.
Counting the shares and funds with daily prices.
US equity OHLCV
US-listed stocks and ETFs (sector, broad-market, international, bond, commodity and volatility funds) with their full daily history. Daily bars only: no intraday bars, no options, no live order book.
Only companies trading today
Only shares and funds that trade today are in the data. A company that failed, was taken over or was delisted is not in it. So a test that picks from a list of shares is flattered, by an amount the test cannot measure. The result card says so on every test of shares. No figure is put on it, because none has been measured for these lists of shares.
UK and European equity OHLCV
FTSE 100 and some FTSE 250 shares, UK ETFs, large Euro Stoxx 50 stocks on Xetra, Paris, Amsterdam, Milan and Madrid, and Swiss majors, priced in their own currency. Backtests and the price-based research answers (returns, risk, correlation, seasonality, dividends, news) cover them; filing-based research answers and paper trading cover US listings only. Each backtest uses one currency.
Returns and costs
Backtests are total return: dividends are reinvested after withholding tax, UK and Irish stamp duty is charged on purchases, and the card itemises what costs and taxes took off. Tell the agent your amount and one of five UK broker presets and the run uses that broker's dealing charges, FX fee and share rules. The buy-and-hold benchmark is bought once on the first day and held.
Fundamentals
Annual figures from 10-K filings via SEC EDGAR XBRL (US filers): revenue, operating income and margin, net income, diluted EPS, free cash flow, ROE, ROA. Fiscal years are named the company's way with exact dates. No quarterly figures; P/E only where a dated closing price is on file. If live SEC data is unavailable the answer says so and uses a dated offline snapshot.
Macro (FRED)
The Research Agent answers from these FRED series: Fed Funds rate, 10-year Treasury yield (monthly average), 2-year Treasury yield (monthly average), 30-year Treasury yield (monthly average), 3-month Treasury yield (monthly average), yield curve (10-year minus 2-year), yield curve (10-year minus 3-month), CPI (headline), PCE price index, PPI (all commodities), unemployment rate, VIX, broad US dollar index, TED spread (ended 21 Jan 2022), M2 money supply, US GDP (nominal), Fed balance sheet (total assets), Fed funds target range (upper limit), Fed funds target rate (to Dec 2008), 5-year breakeven inflation. They are refreshed from FRED every weekday. Every answer states the observation date and links the series.
News, earnings and what isn't covered
News and "why" questions are answered from a cited web search, with every article listed. Earnings history comes from SEC filings with the share-price move around each release; there are no consensus estimates, so no beat or miss, and no analyst ratings: the research agent says so instead of answering. Crypto and intraday data are not supported.
The four surfaces you will use
One paragraph each: what each surface is for, not a feature checklist. Everything else is reachable from these four entry points.
Strategy Agent
Describe a strategy in plain English, combining up to four conditions. The agent runs exactly the rules you asked for through the Rust engine, shows the cost before it runs, and returns the result with every warning attached. A rule it can't test is named and refused, never silently swapped for a different strategy. Give it a goal instead and it searches a fixed set of candidates, judged on years it didn't choose them on.
Open Strategy AgentResearch Agent
Ask about prices, performance and risk, fundamentals, valuation, earnings history, dividends, sectors, macro, comparisons, screens, insider and congress trades, or what a company's 10-K says. Every figure links to its source with a date, and AI-written sentences are checked number by number against the data. When it can't answer, it says why and what it can answer instead, and never makes up a thesis from price rows.
Open Research AgentPaper Trading and Allocator
Follow a saved strategy in paper trading and it holds the engine's exact target weights, cash included, on the strategy's own rebalance days, with the backtest's costs, dividends and interest on idle cash. Allocations run as one netted portfolio. Publish a strategy and its paper record from that moment is locked into an append-only forward track record. US listings only for now.
Open Paper Trading and AllocatorStrategies and Compare
Saved strategies live with their full DSL, the metrics they shipped with, and the warnings they shipped with. The compare surface lines two to four of them up on the dates they share, with drawdowns and their correlation, and warns when they aren't comparable.
Open Strategies and CompareKinds of work the platform supports
Not a list of what to test on day one, but a sketch of the kind of question the platform is built to answer. Each one opens the Strategy Agent with a ready-made prompt you can edit.
Moving-average crossover
Hold SPY while its 50-day average is above its 200-day average; move to cash when it crosses below.
RSI mean reversion
Buy a large-cap stock when its 14-day RSI drops below 30 and sell when it recovers above 50.
Sector momentum rotation
Each month, hold the three sector ETFs with the strongest 12-month momentum.
Asset-class momentum
Each month, hold the two of four funds (stocks, bonds, gold and oil) with the best return over the last 12 months. No month is skipped.
Three ways to start
Read the methodology first if you want to know exactly what every number means before you run anything. Open the workspace if you would rather kick the tyres. Review the plans if you want to see what is metered and what is not.