I'm running a side blog called Markets and Shenanigans as a subdomain to neuralmarkettrends.com. I tend to make subdomains of many of my web properties as scratchpads, tests, and POCs (proof of concepts).
Long-time readers of this blog, if any remain, will know that I tried to use AI and various forms of machine learning and data science to try to predict the markets. I failed and believe that it was a fool's errand to do so.
I wrote about my mea culpa on Medium here:
The simple fact is that long-term investing works well because the market will be higher over time. There's never been a stretch in the market's history longer than 9 years where it was down. The odds are in your favor if you look toward the long term.
At the daily or weekly level, it's all block and tackling. We know that the markets go in cycles and that fear and greed matter. We can be certain that people will become euphoric and greedy and then panic and sell when they're afraid.
I like fear cycles because it means that good quality assets go on sale. I can buy when others are fearful and sell when they're greedy. That strategy has worked out well for me and my family, but I don't need an AI model to tell me that.
So should I abandon using AI for modeling the financial markets? Not quite, there are some models and machine learning applications that are useful. Large language models or using NLP to extract key terms or sentiment is handy for analyzing blogs, articles, and Federal Reserve statements.
They can tell us if, over time, fear or greed is increasing. They can analyze the tone of Jerome Powell's press conference and then can be handy in looking at how far prices have fallen or risen from the mean.
For example, I'm rebuilding my gold and silver price forecast models. I look at the trend and gauge when is a good time to buy or sell physical gold. This isn't new, there are tons of numismatic newsletters (where you pay a hefty yearly subscription) that try to predict if prices are going up or down so small local coin shops (LCS) can add or sell inventory.
Remember, the LCS people make their money from spreads but to hedge for any potential losses they like to buy at low prices and sell at higher prices.
So yeah, these models can be handy as part of the decision support system. They shouldn't be relied solely upon to make a buy or sell decision but be part of the determination equation.
I wouldn't call my price forecasting model AI, I used Python to code it up and autogenerate a post and drop it into markets.neuralmarkettrends.com. Now I'm looking at how successful it's become in helping me decide when to buy or sell my metal, I going to move it out of the "dev" type of account and make it more robust.
I'm debating deprecating the markets.neuralmarkettrends.com subdomain and bringing the price forecasts for gold, silver, bitcoin, and dogecoin here. The models forecast where the prices are heading, up or down, for the next Friday. The models autogenerate a candlestick chart with AI-assisted support and resistance lines and then a trend line.
That's a pretty handy thing because it looks back over time and figures out the key areas where I would expect the price action to become interesting.
Overall, I don't recommend people trade physical silver and gold in general. I don't recommend anybody to do anything with any financial asset (see my disclaimer), but I like using my talents and expertise to make more money.
I don't recommend trading physical gold and silver, but I think it's wise to put a certain percentage of your savings into these hard assets to hedge against inflation. It won't appreciate much but it can keep inflation from eating your savings. Plus it's harder to exchange for cash because of the trouble you have to go through to exchange it.
Additionally, I don't invest in cryptocurrencies anymore but you can't ignore them, so I decided to build these weekly forecast models for Bitcoin and Dogecoin. It's my way of keeping my finger on the pulse of the markets and not being blindsided by things.
So after my long soliloquy, what am I getting at?
- Focus on long-term investing
Gold, Silver, Bitcoin, and Dogecoin weekly price forecast reports will be moved to the root domain (markets.neuralmarkettrends.com)
- I'll be building more python based decision support models.
- I'll be keep score on how well my models do and share them with you