[smcn]

We've been experimenting with LLMs to create price predictions and stock analysis for day trading, and I really haven't been impressed with Bard at all. People in the comments are comparing it to gpt-3.5 and, for our use at least, I wouldn't say it's even remotely close. It's a very clear win for OpenAI.

It's not just that the accuracy of the predictions is wrong, it's that the analysis itself is just not in any way usable. We explicitly describe what the response should look like but Bard will do its own thing.

Here is an example of ChatGPT for RIVN:

"Rivian Automotive (RIVN) has shown a strong upward momentum in recent discussions, with the latest figure reaching 105.51. The stock's current price is $25.62, with a significant increase in volume to 1.41M from its 65-day average of 38.68M. The stock has been trading above its 10-day SMA ($22.08), 20-day SMA ($18.28), and 50-day SMA ($15.68), indicating a bullish trend. The RSI values are above 80, suggesting the stock is currently overbought. Considering the positive market sentiment towards tech stocks and Amazon's record sales during its Prime Day event, which could indirectly benefit Rivian as Amazon is a major stakeholder, an entry price around $25.70 seems plausible. The first price target could be set at $26.20, followed by a second target at $26.70. Confidence level: 75%."

Here is Bard with the same prompt:

"Analysis:

Based on the data you have provided, I believe that the stock RIVN is likely to continue to rise in the near future. The current price is slightly below the 20 day SMA, which suggests that the stock is undervalued. Additionally, the momentum data is positive, which indicates that there is strong buying pressure. The latest market news is also positive, as it suggests that the overall market is bullish.

However, it is important to note that there are some risks to consider. The stock is still highly volatile, and it could experience a pullback in the near future. Additionally,"

Yes, it does end after a comma. That's not me trying to make it look bad.

[dudeinhawaii]

I had the same exact experience with Bard/GPT3.5 and I'm using GPT-4 for a similar purpose of generating "analysis reports" from raw data coming out of a trading algorithm system.

GPT 3.5 had similar problems with formatting and following instructions, and I found that I had to move key instructions to the front of the description or else it would wander off the reservation.

GPT-4 operates as if I had dumped a collection of data on a junior analyst's desk and asked them to write up a report for our clients. It can even follow instructions like "insert market related joke or jab at the Fed here" in a completely believable way.