Over the past 21 days, I've learned a lot about when and how to use this AI effectively. While the AI has shown promising results in certain conditions, there are also key limitations to keep in mind. Here’s what I discovered.

Timing Matters: Use Predictions After the US Market Closes

One important lesson I learned is that the model works best when the US stock markets are closed. In my testing, I found that the predictions were more accurate when the markets were less active. The first hour after the US market closes often shows price movements that are harder for the AI to predict, especially because there’s a lot of volatility in the market at that time.

For example, I noticed that if I bought Solana during the first hour after the market close, I often faced losses. However, by waiting a bit longer, the AI predictions started to become more reliable, allowing me to make better decisions.

The AI Works Better on Weekends

Another thing I found is that the AI tends to make more accurate predictions over weekends. This might be because weekend trading volumes are lower and the market movements are less affected by the large institutions that dominate weekday trading. I’ve seen that the predictions during weekends tend to follow a clearer trend, allowing the AI to forecast prices more precisely.

Stability is Key: The AI Performs Better When Bitcoin is Stable

One pattern I noticed is that the AI works better when Bitcoin (BTC) isn’t too volatile. Bitcoin often influences the broader crypto market, and when BTC prices are swinging wildly, it confuses the AI and leads to inaccurate predictions. On days when Bitcoin was relatively stable, the AI predictions for Solana tended to be spot on.

For example, on days when BTC had a sudden spike or drop, the AI was less effective in predicting Solana’s price. But when BTC was steady, the AI showed a higher success rate in forecasting SOL price movements.

Hourly Updates Are Crucial

A big takeaway from my testing is that the AI model requires hourly updates to maintain accuracy. The predictions I got early on were often good, but after the first two hours, the accuracy started to decline. I had to generate a new prediction every hour to keep up with the rapid changes in the market.

If I didn’t update the predictions regularly, I’d find myself relying on outdated data, which led to incorrect forecasts. This also meant that I had to be on top of the market every hour, which can be a lot of work if you’re not prepared for it.

The AI Isn’t Perfect: Mistakes Happen

While the AI has been helpful in many cases, it hasn’t been perfect. I’ve seen the model make some significant mistakes, especially when market conditions were unusual or unpredictable. There were times when the AI’s predictions didn’t align with actual market movements, resulting in losses.

For example, there were occasions when the AI predicted that Solana would rise, but the price ended up falling instead. These mistakes were a reminder that no model is flawless, and it’s important to have a strategy in place for when things don’t go as expected.

Conclusion: What I Learned from 3 Weeks of Testing

After testing this AI model for three weeks, I now know when it’s best to use it and when I should be cautious. The key is to be patient and wait for the right conditions—use the predictions after the US market closes, pay attention to Bitcoin’s stability, and make sure to update the model every hour.

The AI works well on weekends and can give good predictions when market volatility is low. But it also has its flaws, so I make sure to be mindful of the potential for errors and losses.

If you're interested in learning more about how this AI model works or want to hear about my future testing, be sure to like and subscribe.