We avoid marketing "accuracy claims." Instead, we make forecasting performance measurable. For each model and timeframe we track MAE (the typical absolute miss in price) and sMAPE (the typical percent miss). Forecasts are delivered as probability ranges, not single-point certainties, and you can review confidence and error metrics per symbol/timeframe inside the product. For additional validation, we run a public Quant Lab: the system's decision logic is published via the Decision Feed, and the same logic is evaluated through live-market execution under predefined risk rules—so you can follow outcomes over time and verify accuracy independently. The Trading Arena goes further: different AI models run on identical market data and execution conditions, so you can compare decision quality across models on a live scoreboard.

No. The Copilot doesn't trade for you. It's built to deliver analysis and structured trade plans you can review and apply yourself. You stay in control of every decision. If you want to see how the system's logic performs in real execution, the Trading Arena runs that publicly — live trades, fixed risk rules, open scoreboard. That's where execution happens, and it's fully transparent.

Right now we support a focused set of crypto and U.S. equity symbols, each with multiple timeframes. Crypto: BTC, ETH, SOL, XRP. Stocks: AAPL, ACN, AMZN, GOOGL, META, MSFT, NVDA, TSLA. For every symbol we provide forecasts and analysis across 5m, 1h, 1d, and 1w timeframes. The list is intentionally short because adding a new symbol isn't just a UI toggle—it requires sufficient compute and meeting dataset requirements to train and monitor reliable forecasting models. If you want a symbol added, message us what you trade and which timeframe matters most—we'll use that to prioritize coverage expansion.

General-purpose AI models aren't connected to live market data. When information is incomplete or stale, they fill the gaps with confident-sounding answers that may have no basis in current reality. tradef.ai grounds every response in a fresh market-data pipeline and a controlled context workflow — built by specialized agents before the model ever starts reasoning. You get a data-backed market read, not a convincing guess.

tradef.ai is built for both. If you're newer, the Copilot turns market data into a structured plan (levels, entries/exits, invalidation, risk) and explains it in simple terms. If you're experienced, it speeds up research and provides a systematic cross-check with transparent reasoning—so you can validate your thesis faster.

We retrain based on model drift, not a fixed calendar. After training, we establish a baseline performance window and then continuously monitor forecast quality. When drift is detected—meaning the model's error worsens and its expected behavior no longer matches current market data—we flag the model for retraining.

Stocks data comes from Finance Modeling Prep (FMP), and crypto data comes from Bybit. Data freshness follows the provider's update cadence, so the system's context is tied to the latest available data.

We don't rely on a generic chatbot. The Copilot runs on a controlled context pipeline: agents receive structured market data and curated context within defined scopes, so outputs are grounded in provided inputs rather than invented facts.

The analysis is periodic, not tick-by-tick. The Copilot updates on a 5-minute cadence, using each new 5m interval as the base refresh cycle.

We take privacy seriously. All traffic is protected with bank-grade 256-bit SSL encryption. Authentication is handled via Google sign-in, so we don't store passwords. We only collect what's necessary to run the service and keep access controlled.

You can cancel at any time with no cancellation fees. After cancellation your conversation history is deleted within up to 30 days.