Warning: Useless Ai

Corona Del Mar, CA

Howdy Friend!

With businesses in a mad rush to cash in on Ai, I have a warning for you…

…don't follow any Ai trading “advice” unless they can show you a backtest and an out-of-sample backtest…better yet, actual brokerage statements.

Quick background:

You may have heard the rumor that the commonly used “chart pattern” branch of technical analysis has been disproved. (You know…like head and shoulders, double-tops, etc).

It has. That was us.

Back in 2017-2018, three of us here at team Portfolio Boss spend 18 months on a project to figure out which price patterns work best, and just trade the most profitable.

We took price and volume data of over 6000 stocks (listed and de-listed) since 1986, and looked at every possible pattern.

And I do mean every possible pattern. We didn't just limit it to head and shoulders and all the junk taught in old books before computers. There were about a million possibilities, and then we had the computer search for matches and what price did afterwards. (So 1 million times 1 million = 1 trillion searches each time we ran the software)

Now when you turn something subjective like a chart pattern into something objective, you begin to ask questions like “what exactly makes a top? What makes a bottom?”

Is a top the highest high for 5 days on both sides? Is the top after some sort of average true range pullback?

And then you have to define the pattern in both time and magnitude…is it three peaks and valleys? Four? Five? Too complex, and you won't have many samples. And what about volume?

And what's the margin of error for defining the pattern? Nothing will ever perfectly match. Price data is noisy.

And by the way, this is how we got into cloud computing. A regular PC couldn't crunch these numbers in a month.

So we ended up crunching all the numbers in our cloud setup and then the algo would make a prediction of where price may go in the coming days.

We got that part dialed in nicely.

Here's where everyone is screwing up: The hard part is actually backtesting your newfound pattern prediction algo.

Once we did that…the results were terrible. Price patterns DO NOT WORK.

We even set out to prove they did and accidentally proved they didn't…which is a kinda backwards way to do science, but that's how it worked out.

I won't be getting those 18 months and $300,000 back (which doesn't include my salary).

Yes, millions of traders are wrong.

// Quick aside: I'll also remind you that between 2-10% of Americans believe the earth is flat. Newsflash: It's not. It's very easy to disprove. They even flew flat earthers to Antarctica last year to witness the sun shining 24 hours per day. Their model falls apart because Antarctica is supposed to be some sort of ice wall like the crust on a pizza. Not to mention all the Foucault pendulums that are on live webcams all over the earth. //

Fast forward to today, and I'm seeing even big companies making price prediction AI bots or chart pattern recognition.

And yes, they are committing the cardinal sin of not backtesting their damn prediction bot.

If you don't backtest, all you have is a hypothesis that it works. Never assume!

And even if you do the work of back testing the idea, you need to confirm the strategy works on two sets of unseen data.

You can probably tell I'm pretty sick of all the nonsense peddled as helpful trading tools.

That's another reason why I decided to give away Portfolio Boss Evo, and the training on what actually makes money, how to scale, etc.

I spent a month prepping for all those training videos, and I gave it away for free with PB Evo so you don't blow out three trading accounts like I did decades ago.

I hope this message helps you avoid what could be a minefield for your portfolio.

P.S. We're putting Portfolio Boss Evo back behind a paywall. Download it now before it's priced at $1999. It also comes with an NVDA strategy with rules fully revealed. You're going to kick yourself when you can't get it for free. Download now >>

P.P.S. Just the quotes alone would cost you $480/year at CSI. I cover the cost because a percentage will enthusiastically upgrade to our more advanced tools.

P.P.P.S. Yes, we do have members using PB to manage millions of dollars. Here's a brokerage statement as proof (obviously, results not typical yada yada):