LLM, ML & Statistics

FIN uses three different types of intelligence working together: a Large Language Model (LLM) to reason and explain, Machine Learning (ML) to spot patterns over time, and statistics to keep everything grounded in numbers.

Large Language Model (LLM)

The LLM is the "brain" that makes predictions. Every time FIN runs, it feeds a rich set of data to the LLM so it can understand the full picture:

Stock prices — open, high, low, close, and volume
News articles — scored for sentiment (positive/negative) toward the company and the economy
Surprise signals — unexpected events like earnings beats or economic shocks
Psychology scores — 12 cognitive bias measurements (FOMO, herding, loss aversion, etc.)
Technical indicators — RSI, moving averages, volume trends
Competitor data — peer price performance, valuation metrics, and competitor news sentiment
Financial statements — income, balance sheet, and cash flow data
Calendar events — earnings dates, dividend dates, and other market-moving events

All of this is packed into a single prompt and sent to the LLM. The model weighs every signal, considers how they interact, and produces:

A predicted price for 1 day, 2 days, and ~10 days ahead
A confidence band (low–high range where the price is expected to land)
A detailed explanation written in plain English — so you can see why the system thinks what it thinks

What Makes the LLM Approach Unique

Unlike a black-box algorithm, the LLM explains its reasoning. You can read exactly why it gave a certain prediction — which news articles influenced it, which psychology biases were most relevant, and how the technical signals factored in.

Machine Learning (ML)

While the LLM handles individual predictions, ML improves the system as a whole by learning from past results.

Snapshots: How ML Learns

Every time a prediction is made, FIN saves a snapshot — a complete record of everything that went into that prediction:

All the data that was fed to the LLM (prices, news, psychology scores, etc.)
The prediction that was made
The actual outcome once the market closes

These snapshots are stored and accumulate over time. When enough snapshots have been collected, the ML model trains itself to find patterns:

"When sentiment is high but overconfidence is also high, I tend to overestimate the price"
"When RSI is below 30 and news is negative, the price usually bounces harder than predicted"
"When there's a surprise event, the confidence bands should be wider"

Self-Correction

The ML model doesn't just identify patterns — it adjusts future predictions based on what it learned. If the system discovers it consistently predicts too high in certain market conditions, it automatically compensates.

The trained model is user-specific — it learns from your chosen tickers, RSS feeds, LLM settings, and scheduling preferences. No two FIN instances learn the same way.

If performance drops (models get stale as market conditions change), old models are retired automatically and the system falls back to its base prediction method until new ones can be trained.

Statistics

Statistics are the foundation that keeps the LLM and ML honest. They measure, validate, and quantify everything the system does.

Prediction Accuracy

Every prediction is measured against what actually happened:

Metric	What It Tells You
Error %	How far off was the prediction? (absolute percentage difference)
Band Hit	Did the actual price fall within the confidence range? (Yes/No)
MAPE	Mean Absolute Percentage Error — the average error across recent predictions

Gating

If the system's recent MAPE exceeds a configurable threshold, predictions are automatically withheld (gated) until accuracy recovers. This prevents unreliable predictions from reaching the dashboard.

Sentiment Thresholds

News sentiment scores use statistical boundaries:

Below 45 — negative sentiment
45–55 — neutral
Above 55 — positive sentiment

Technical Indicators

Statistics power the technical analysis:

RSI (Relative Strength Index) — a 0–100 score. Above 70 suggests overbought, below 30 suggests oversold
Trend detection — statistical methods determine if the price is moving up, down, or sideways
Volume analysis — whether trading volume is increasing or decreasing relative to historical norms

Psychology Scoring

Each of the 12 cognitive biases is scored on a 0–100 scale using statistical analysis of article text:

Low (0–30) — barely detectable
Medium (30–60) — some influence present
High (60–100) — strongly influencing the market

How They Work Together

   Collected Data
         ↓
  ┌──────────────┐
  │  Statistics  │──→ Measure, validate, set thresholds
  └──────────────┘
         ↓
  ┌──────────────┐
  │     LLM      │──→ Reason, predict, explain
  └──────────────┘
         ↓
  ┌──────────────┐
  │  ML (future) │──→ Learn from outcomes, improve over time
  └──────────────┘
         ↓
  Better predictions

Statistics measure the raw data and validate the outputs. The LLM uses that data to reason about what happens next and explain its thinking. ML uses the outcomes to get smarter with every cycle.