Investment Management

A model to give all investors edge

The Continuum team aims to develop IM-GPT, a large language model application designed to serve as an assistant investment manager.

IM-GPT, or Investment Manager-GPT, represents the team's inaugural endeavour in creating a sophisticated tool tailored for the financial industry.

What does it do?

This application offers users insightful commentary and research guidance across various financial markets, including equities, commodities, currencies, and fixed interest securities.

IM-GPT has the capability to process streaming news flow, discern crucial information within it, and subsequently analyse this content to provide users with valuable research ideas and perspectives.

IM-GPT is not designed to act as a decision maker, but as an idea generator and aid in rapid decision making.

How does it work?

IM-GPT's functionality hinges on its ability to tailor its analysis to the user's specified area of interest or content.

Users have the flexibility to define their focus, whether it be industry sectors, geographic regions, investment themes, or individual securities.

Additionally, users can provide specific information for analysis and instruct IM-GPT on the desired approach to scrutinise the data.

Highlighting insights and anomalies

IM-GPT delivers comprehensive analyses encompassing:

1. A summary of the content

2. Historical context

3. Potential insights challenging existing perceptions

4. Implications for the future

5. Suggestions for further investigation

6. Potential courses of action

How has it been trained?

Historical Datasets

IM-GPT has access to the following datasets, accessed via retrieval augmented generation (RAG) and vector databases:

· Annual Reports

· Company Announcements

· Shareholder Register

· Websites

· Social Media

· News Media

· Academic Papers

· Finance textbooks

Training Dataset

IM-GPT is a fine tuned version of several open sourced large language models. It's fine tuning dataset encompasses the full range of investment management domains

· Behavioural Economics

· Industry Analysis

· Valuation techniques

· Systems Thinking

· Organisational Management

· Marketing and Distribution Theory

· Theory of the Firm

· Portfolio Theory

· Quantitative Techniques

· Balance Sheet

· Cashflow Statements

· ESG

· Remuneration Analysis and Incentives

· Value and Growth Investing

· Risk and Volatility

In Action

The model is able to ingest streaming fundamental data - for example company announcements - and interpret them through the lens of fundamental analysis.

As important, it has access to history - it can compare incoming fundamental data with historical data to detect anomalies. This historical data is not just constrained to the individual company, but its competitors, suppliers, macroeconomic influences and any field related to the incremental news flow.

This enables IM-GPT to highlight to the user discrepancies between prevailing perceptions and reality.

These insights, characterised by differing from consensus views and being accurate, are vital in identifying market inefficiencies.

Psychological Biases

Behavioural economics explores how psychological factors affect economic decision-making.

Within this field, heuristic biases describe mental shortcuts that people often use, sometimes leading to systematic errors or irrational behavior.

IM-GPT has been trained on the vast domain of behavioural finance. It is aware of the many areas of bias and error and seeks to correct them.

For example:

Anchoring Bias: Relying too heavily on the first piece of information encountered (the "anchor") when making decisions.

Availability Heuristic: Making judgments about the likelihood of events based on how easily examples come to mind.

Representativeness Heuristic: Assessing the similarity of objects and organizing them based on the category prototype.

Confirmation Bias: Tendency to search for, interpret, and remember information in a way that confirms one's pre-existing beliefs or values.

Overconfidence Bias: Overestimating one's abilities or the accuracy of one's beliefs, often leading to excessive risk-taking.

Hindsight Bias: The tendency to believe, after an event has occurred, that one would have predicted or expected it, commonly known as the "I-knew-it-all-along" effect.

Sunk Cost Fallacy: Continuing a behavior or endeavour based on previously invested resources (time, money, effort), even when it's not in the best interest.

Endowment Effect: Valuing something more when you own it, leading to potentially irrational decisions like refusing to sell an asset for its market value.

Status Quo Bias: Favouring the current situation and resisting change, even when change might lead to a better outcome.

The product

IM-GPT is designed to give edge to traders and investors. This edge has often been the domain of quantitative based investors and high frequency traders. Fundamental investors now have their own tool to generate market edge.