How This Actually Works
Not every workflow works the same way. Here's what's actually happening behind each one.
TradeAnalysis has two kinds of workflows, and they're built differently on purpose. The 4 trading bots run on live market data with a hard, code-level guardrail around risk. The 14 analysis workflows reason through the inputs you give them against institutional-style criteria. This page explains both, plainly.
The Trading Bots
Scalper, Basic Day, Advance Day, and Swing all follow the same pipeline, differing only in which timeframes they read:
- Live price data across three timeframes per bot (for example, Scalper reads 15-minute, 5-minute, and 1-minute candles; Swing reads 4-hour, 1-day, and 1-week), pulled fresh on every request — not backtested or cached demo data.
- Live news sentiment from the last 48 hours for the asset you're asking about, factored in alongside the price action.
- Computed technical indicators — 20 and 50-period EMAs for trend, 14-period RSI for momentum, 14-period ATR for volatility — calculated directly from the real candle data, not estimated by the AI.
The AI's job is narrow by design: given that data, it decides a bias (buy, sell, or hold) and a confidence level, and writes the reasoning behind the call. It does not set the price levels.
The risk-sizing guardrail
Entry, stop-loss, and take-profit are computed by a fixed formula, not by the AI: stop-loss is set 1.5× the asset's current ATR from entry, and take-profit is set 3× ATR from entry — a fixed 1:2 risk-to-reward ratio on every signal, regardless of how confident the model sounds. If the underlying price or volatility data is missing or inconsistent, the bot returns a HOLD with Low confidence and says why, rather than guessing. This is the part of the system that isn't up to interpretation — it runs the same way every time.
The Analysis Suite
The 14 institutional-style workflows — stock screening, DCF valuation, risk assessment, pattern detection, and the rest — work differently. You give the workflow your inputs: risk tolerance, portfolio holdings, a ticker, a sector, a time horizon, whatever that specific workflow asks for. The AI reasons through those inputs against a defined set of institutional-style criteria for that workflow, and returns a structured result — the same fields every time, not free-form text.
Each workflow's exact criteria are described on its own page — see the "How it works" section on any agent page for the specifics of that workflow.
Why It's Built This Way
An AI model that's free to invent a stop-loss or a price target is only as reliable as its last guess. Splitting the job — AI for judgment, code for the numbers that matter to your risk — means the price levels on a trading bot signal are always traceable back to real market data, every time, not something the model made up in the moment.
One More Thing
None of this is financial advice, and none of it guarantees an outcome — see the disclaimer in the footer of every page for the full detail. What this page is for is answering the question we think matters most before you trust any signal: what is this number actually based on?