Can Chat GPT develop a profitable trading algorithm to implement on the financial markets ? And how do I start with implimantation
1 Like
ChatGPT is not capable of generating a trading strategy for the financial markets currently due to its training limitations. It has been trained on data only up to 2021 and does not have access to real-time market information.
From a purely data perspective the stock market increases in value for any given random selection of stock options given sufficient time and a sufficiently wide spread. Trading firms that make a profit over and above the base rise are using additional external data points and perhaps they have some internal secret sauce algorithms, but they are heavily guarded and protected.
There are no get rich quick schemes that you can simply plug an AI into and generate profits without considerable effort being put in, it should also be noted that the GPT models are large language models and are not optimised for numerical analysis directly. I don’t think an LLM is the ideal engine to power a venture of this type.
1 Like
Can it develop a trading algorithm?
Yes. You can give it the kinds of patterns you want to look for, and it can generate Python code or something that might look for those patterns. You can then run that code/algorithm, to do trading.
Can it develop a profitable trading algorithm?
Unlikely. Just because every trading algorithm is unlikely to be profitable long term.
2 Likes
I have a feeling there will be a more profitable strategy that inverses ChatGPT recommendations 
2 Likes
There are several approaches to leverage LLM capabilities to create trading strategies, or at least supplementary indicators.
A very common and often cited approach is sentiment analysis based on press releases, tweets and the like.
Another thing that has been kicked around a lot is the technical analysis approach where the model is fed data and it’s supposed to identify patterns which can then be traded for a profit.
I’m sure there is a lot more but let’s focus on these two common examples. In the first case you still need to set up the infrastructure to get the data and then create the sentiment analysis which needs to be timely in order to trade on hot news. Besides using a LLM as a hobbyist there is zero innovation with this approach and one will discover the large amount of existing competitors who will outsmart any half-baked attempt by miles.
In the second example you will have to fight an uphill battle against the LLMs tendency towards recency biases. In short the model will overweight the most recent data, producing momentum driven approaches which are in turn a statistically overfit and break down in any out-of-sample scenario as the market keeps changing. In this case you could still run the same analysis using already existing tools and get the same results.
So, can you create a trading strategy using an LLM? Yes, but it is not a shortcut in any meaningful way.
1 Like
Yes.
Before ChatGPT, I had already been able to predict, with high confidence, quarterly revenue of some companies, by using a few public datasets, of course.
Now, with ChatGPT, if you could add 20 - 50 more public datasets, it would be very likely to predict the quarterly net income of the companies, even the dividend movement probability. A few days before the quarterly reports are released, that information would give you some advantage. Well, the quarterly performance may not necessarily correlate with the stock prices. But they are strong pointers.
Right now, the difficulty is that the ChatGPT Advanced Data Analysis does not have any API for you to upload the datasets. The manual operations would make the process labor prohibitive. But it is just a matter of time that ChatGPT or its competitors will offer the API. The result is unknown now because nobody has ever tried that, but I think I will find that out pretty quickly as soon as the API is available.
1 Like
I’ve implemented trading strategy with ChatGPT.
Been running 25k fund for 3 months - great results of far. Pure equity options sales.
In 6-8 months it will be more or less on autopilot.
1 Like
Released on March 30th, BloombergGPT represents a groundbreaking development in the financial sector’s application of AI technology. By focusing specifically on finance-related tasks, BloombergGPT aims to enhance existing NLP applications employed by Bloomberg, including sentiment analysis, named entity recognition, news classification, and question answering. Furthermore, this sophisticated model holds the promise of unlocking new possibilities for leveraging the vast amounts of data accessible through the Bloomberg Terminal, thereby empowering the firm’s customers and fully harnessing the potential of AI in the financial domain.
1 Like
Yes, but it likely be as rudimentary as one that you could create on platforms like MetaTrader or Trading View if you just grabbed some common indicators and experimented with their parameters, and some backtesting even if successful (as that doesn’t mean the strategy will be profitable on forward-testing basis).
- As @Foxalabs said, there are no get-rich-quick schemes (that are not scams, I would add).
However, with a complex enough series of prompts and guidance, just as you would with engineering one manually, ChatGPT could assist a developer in coding one in a given programming language, like Python or TradingView’s Pine script, or other languages it knows (but the quality will depend on the amount and scope of work/effort put it).
Here is an example of such a rudimentary one that I built with ChatGPT as an example:
Note: performance is hypothetical based on a backtest and I have not used this with real funds, nor do I plan on using it (was just a use-case example), and beware of people that impersonate me just FYI to everyone here.
I don’t understand what do you exactly mean by public datebases. Do you mean like trading strategies, news, or graphs for successful trades? And how do you find databases in general to make your own gpt (Not specificelly for the stock market) but in general.
Best regards
Billy