How our AI-assisted signals work
Transparency matters when you rely on data to research the market. This page explains, in plain language, how TradeHive AI produces the signals and context you see in the workspace.
TradeHive AI is for research and educational purposes only. It does not provide investment, legal, or tax advice. Market data and AI-generated outputs can be inaccurate, delayed, or incomplete.
AI Search factor weights (config 2026.06.1)
Each strategy scores the same factors with its own weights, which always sum to 100%. If a factor has insufficient data, it is excluded and the remaining weights are re-normalized; nothing is filled with a placeholder value.
| Factor | Momentum | Mean reversion | Options flow |
|---|---|---|---|
| Momentum | 30% | — | 20% |
| Trend | 35% | 20% | 25% |
| Volume | 15% | — | — |
| Volatility | 20% | 20% | 15% |
| Reversion | — | 35% | — |
| Oversold | — | 25% | — |
| Options flow | — | — | 40% |
Risk profiles
| Profile | Stop (× ATR) | Account risk per trade | Score needed for a verdict | Minimum confidence |
|---|---|---|---|---|
| Conservative | 1 | 0.5% | ≥ 62 or ≤ 38 | 65 |
| Balanced | 1.5 | 1% | ≥ 58 or ≤ 42 | 45 |
| Aggressive | 2 | 2% | ≥ 56 or ≤ 44 | 30 |
Scores between 45 and 55 are neutral. Probabilities are model estimates, shown between 5% and 95%, and a Bullish or Bearish label also requires the probability, options flow and key-level order to agree. Accuracy is shown once at least 30 past outlooks for a strategy and horizon have been graded.
What we combine
Our research workspace brings together several inputs: market data (intraday and historical), options activity, and commonly used technical indicators. Market data may be delayed depending on your plan and market conditions. AI-assisted analysis is layered on top to summarize context and surface areas that may be worth a closer look.
How signals are generated
- Market and options data is collected and normalized for the tickers you track.
- Technical indicators (such as trend, momentum, and volatility measures) are calculated from that data.
- AI-assisted analysis interprets the combined context and presents it alongside the underlying figures.
- Each signal is shown with supporting context so you can evaluate it yourself.
How an AI Search outlook is scored
- Indicators in code: RSI (14), EMA 21/50/200, MACD, relative volume (5-day vs 20-day) and ATR (14) are computed deterministically from daily bars.
- Score (0 to 100): a weighted blend of momentum, trend, volume and volatility components. The mean-reversion strategy adds a 20-day z-score input; the options-flow strategy adds the call share of options premium. A score of 60 or more reads Bullish, 40 or less Bearish, otherwise Neutral.
- Probability: the chance of closing above the current price at the end of the horizon, estimated from 60-day drift and 20-day volatility of daily returns. Each probability is shown with this definition.
- Levels: target = entry ± 1.5 × ATR × √(horizon days); stop = entry ∓ 1.5 × ATR. The risk level (Low, Medium, High) comes from ATR as a percentage of price.
- Explanation: the AI writes the thesis from these numbers. If its text uses numbers that are not in the data, it is regenerated or replaced with a template summary.
How outlooks are graded
Every outlook is saved with a timestamp and its inputs. When its timeframe ends, it is graded against the daily closing price: a hit means price moved in the predicted direction. Neutral outlooks are not counted. The public track record summary is shown only once at least 30 outlooks have been graded. Historical results do not guarantee future performance.
How the Daily Market Snapshot briefing is created
The "day in 60 seconds" briefing is written by an AI model using only a fixed set of facts our servers collect that morning: ETF stand-in prices for the major indices, Treasury yields, the latest official macro readings, market-calendar events and recent headlines. Every number in the briefing is checked against those facts in code. If the text uses a number that is not in the data, or uses advice or price-target language, it is regenerated once; if it fails again, a plain template summary built directly from the data is shown instead. Each bullet lists its sources, and every briefing is logged with its inputs for quality review. The briefing describes events and context only. It never gives trade instructions.
What signals are — and are not
Signals are research prompts, not recommendations. They are intended to help you decide where to focus your own analysis. They are not predictions of future performance and should never be treated as advice to buy or sell any security.
Limitations
AI-generated outputs and market data can be inaccurate, delayed, or incomplete. Models can misinterpret context, and past behavior does not indicate future results. Always independently verify information before acting on it.