Methodology
How NeoInvest analytical outputs are structured.
Understand the distinction between market data, technical conditions, machine-learning forecasts, portfolio calculations, scenarios and language-model explanations.
The six layers
Outputs that look similar on screen can have very different standing. These are kept distinct throughout the product, because “the model said” and “the arithmetic says” deserve different levels of trust.
Market data
Prices and market state captured from data sources, presented with the currency, the market status and the time of capture. Where data is incomplete or stale, that state is shown rather than hidden.
Technical conditions
Deterministic summaries of observed price and volume behaviour — trend, momentum, volatility and volume. These describe what has happened, not what will happen.
Machine-learning forecasts
Model estimates over a one-trading-day or ten-trading-day horizon, each carrying a direction probability, an estimated range, a confidence measure, the generating model version, a generation time and an expiry.
Portfolio analytics
Deterministic arithmetic over holdings you entered: allocation, concentration, sector and geographic exposure, and currency exposure. Same inputs, same outputs, every time.
Scenario Engine
Applies assumptions you supply to a portfolio and reports the estimated effect. The engine does not judge whether an assumption is plausible.
NeoInvestGPT
A language-model interface that interprets your question, selects approved NeoInvest tools and explains what those tools returned.
Confidence is not certainty
Confidence describes the model output under its current inputs. It does not guarantee a future market result.
In particular, confidence is not the probability that you will make money. A high-confidence estimate of a small move is still an estimate, and can still be wrong.
Supporting and limiting factors
Each forecast lists what contributed to it and what constrains it. These are shown as descriptive contributors — the factor list explains where an estimate came from, and is deliberately not presented as a checklist of reasons to act.
What the language model does not do
Every financial value shown in a NeoInvestGPT answer is produced by a NeoInvest service and passed to the model to describe. The model is not asked to compute, estimate or recall a number.
The model does
- Interprets the question you asked
- Selects approved NeoInvest tools
- Explains structured outputs in plain English
The model never generates
- Market prices
- Forecast numbers
- Portfolio calculations
- Scenario impacts
- Tax figures
Known limitations
- Market conditions change, sometimes faster than any model updates.
- Unexpected events may not be represented in the data the model has seen.
- Data can be incomplete, delayed or revised after publication.
- Historical evaluation does not guarantee future performance.
- Analytical outputs expire, and an expired output should not be treated as current.
Model performance
NeoInvest does not publish forecast-accuracy statistics. Model evaluation is performed on out-of-sample data, and any published performance figures would need an approved evaluation report behind them. Until that exists, no accuracy claim appears anywhere on this site.