How it works: methodology & limitations

Every analysis is run by a team of 9 AI agents. Each agent is the same large language model given a different role and different instructions. Stocks and leveraged products (forex and CFDs) use different teams because they are traded differently: stocks are judged on value and trend over months, forex and CFDs on short-term price action, volatility and scheduled events.

1. The app downloads market data for the symbol and computes indicators itself (RSI, MACD, Bollinger Bands, moving averages, ATR, pivots). These numbers are calculated in code, not by the AI.
2. The four analysts run at the same time, each seeing the same data.
3. The bull and bear researchers read the analyst reports and argue opposite sides; the moderator scores the debate.
4. The risk manager and the final decision-maker run last and produce the conclusion.
5. Every agent is told to use only the numbers supplied and to say when something is missing rather than invent it. This reduces, but does not eliminate, made-up details.

If an individual agent fails, its card is marked as failed and it counts as neutral (5/10). If the final decision-maker fails, the whole analysis is marked as failed and no verdict is shown or saved as a result.

For forex and CFDs the stop and target are checked in code: if the AI puts a stop on the wrong side of the entry, the app replaces it with a volatility-based level (ATR) and labels where the levels came from. Stops are never tighter than 10 pips for forex or 0.15% of price for CFDs, and the target is at least 1.67× the stop distance. For stocks, a target or stop on the wrong side of the price is dropped rather than shown.

• Prices and charts: Yahoo Finance public chart data. This is an unofficial, free source; quotes can be delayed (for some exchanges by 15 minutes or more), can have gaps, and are not exchange-grade.
• Live price updates: the app re-requests quotes about every 2 seconds while a page is open. "Live" here means "the latest quote the source has", not a direct exchange feed.
• Chart data is cached for up to 60 seconds on the server.
• Economic calendar: the public ForexFactory weekly calendar (current week only, refreshed every 30 minutes).
• News: Yahoo Finance RSS headlines. Only headlines are read, not full articles.
• Stock fundamentals are limited to what the quote feed provides (for example P/E and market cap). The app does not read company filings.

Each analysis report shows when the analysis was run and when the market data was fetched, so you can tell how old it is. Saved analyses are a snapshot: they are not updated when the market moves.

Conviction (0–100) is the score the final AI decision-maker gives to its own call after reading the team's reports. It is a self-reported confidence level from a language model. It is not a probability of the trade working, it has not been calibrated against real outcomes, and a higher number does not mean a higher win rate. If the model does not return a score, 50 is shown.

The agent ratings (1–10) work the same way: 10 is the most bullish view, 1 the most bearish, 5 neutral.

The stored verdict is always BUY, SELL or HOLD. How to read it depends on what you are looking at:

• Stocks: HOLD means "keep an existing position, but this is not a new entry". If you do not own the stock, HOLD means do nothing.
• Forex & CFDs: HOLD is shown as NO TRADE: there is no new setup worth opening. The report still shows the levels for the better-scoring direction so you know what to watch.
• BUY on forex/CFDs means a long trade; SELL means a short trade.

All trading in the app uses virtual money. Nothing is sent to a broker and no real money can be deposited or withdrawn.

• Balances: a $100,000 virtual stock portfolio, and a separate $100,000 virtual account shared by forex and CFDs.
• Fills: market orders fill at the latest quote from the data source when you click. There is no order book, queue or partial fill.
• Costs: spreads, commissions, overnight financing (swap) and slippage are not modelled. Real trading would cost more, so practice results look better than they would with a broker.
• Leverage and margin (forex & CFDs): forex 30:1, stock indices and gold 20:1, silver, oil and natural gas 10:1, crypto 2:1. Margin = position value ÷ leverage. An order is rejected if you do not have enough free margin.
• Minimum size: forex 1,000 units (a micro lot); each CFD has its own minimum.
• Order checks: before an order is accepted, the stop must be below the price for a long and above it for a short (and the reverse for the target). The ticket shows the loss at the stop in dollars and as a % of equity, the reward, reward:risk, margin and costs before you confirm.
• Stops and targets (forex & CFDs): checked against one-minute price bars whenever you open the forex/CFD pages or your positions are refreshed. A position closes at exactly the stop or target price, with no gap or slippage. If the stop and target are both touched in the same minute, the stop is assumed to have been hit first. Trailing stops move only in your favour.
• Stocks: buys and sells fill at the live quote; there is no automatic stop. The holdings monitor flags positions that cross the AI stop or target, but you decide whether to sell. You cannot sell more shares than you own (no short selling).
• Backtests use a simple fixed rule set (RSI with moving averages) on historical daily data. They show how that rule would have behaved, not how the AI agents would have performed.

• The agents are language models. They can misread data, overstate confidence or produce reasoning that sounds right but is wrong.
• The AI agents' recommendations have not been tested against real historical outcomes, and the app makes no claim that they beat the market.
• Data is free and unofficial; it can be delayed, incomplete or occasionally wrong.
• News coverage is limited to a handful of headlines; important information can be missed.
• The economic calendar only covers the current week.
• Practice results ignore trading costs and assume perfect fills, so they will look better than real trading.
• Nothing in the app is personal financial advice. It does not know your finances, goals or risk tolerance.

Questions? See the FAQ or try the illustrative demo.