Update: Predicting the Markets Research
I trust today’s email finds you well and set for a fabulous weekend. Fall is in full swing here as the leaves turn brilliant shades of orange, red, and maroon. The temperatures are dropping and snow’s already piling up in the passes which means another one of my favorites – skiing – is just around the corner!
Anyway, that’s enough about that!
My data scientists and I have made a lot of progress in recent weeks when it comes to predicting the markets. So, knowing you’ve got a keen interest in what I’ve been up to in this department, I thought you’d enjoy an update on some of my most recent research.
Without further delay …
One of the crucial measures in any predictive data set is something called “MAPE” which stands for Mean Absolute Percentage Error. It’s a statistical measure used to gauge how accurate forecasting systems are or as a loss function in machine learning.
What I like about MAPE is that it’s easy to interpret because most folks are used to thinking in percentage terms so talking about the absolute error in forecasting models makes sense.
My predictive MAPE is consistently between 1% and 3% — and here’s the cool part – as many as five days in advance. On stocks, on ETFs, on bonds… in fact, on every financial instrument my team and I have tested.
Take a look.
Interestingly, this “run” isn’t as tight as I’d like but, in keeping with my desire to be completely transparent, I’m not holding it back. Unlike many of my peers in this industry, I don’t believe in cherry-picking to make something appear better than it is.
You can see that the predicted line in green reflects a drop over the next five trading days even as the indices inch into yet more record territory.
That’s called a “divergence” in technical trading terms because two important data sets are going in opposite directions. Usually, this is interpreted as a precursor to a major move.
Not always, though.
Remember, we’re talking about projecting the future five days in advance – meaning the actual divergence hasn’t happened yet. But could.
That’s obviously hard to predict but not unreasonable to think about which is why I created a related tool called the “accuracy counter.”
I wanted to know how far “on” or “off” predicted data is from actual in real time because that knowledge will allow me to make potentially very profitable decision based on the “correctness” of projected directional trends in addition to anticipated price movement.
Here’s what that looks like.
[BREAKING] You Better See This Now (Tom Is Telling All)
I designed the calculations to quantify the percentage of total data points in a given analytical period – in this case 240 data points – are correct when price and direction are taken into consideration at the same time.
This is a significant twist to conventional thinking which normally views calculations like this discretely and one with surprising results. As you can see, the projected data is correct 80.83% of the time.
There’s still a lot of work to do (and coffee to drink during my marathon all-nighters, which, thanks to my lovely wife, Noriko, is in ample supply) but the thinking – and my research – is presently focused on how we can use that to trade what happens next before other folks realize it’s even in the cards.
There are loads of people who have tried to do this, but they’ve all got a critical flaw in the math they use. Generally speaking, the calculations they use are based on historic patterns and numerical sequences using probabilistic calculations.
Current options prices models, diversification, portfolio allocation, time-based calendar calculators, price based tools… they’re all based on what’s already happened and the – flawed – assumption that there will be something similar happening in the future.
Contrary to what many people like to think, there’s no predictive value in that whatsoever. Zero, zippo, nada, bupkis, zilch.
However, there is an edge when you have a forward-looking tool like this one at your disposal.
Options premiums are inexpensive when prices are rising which means you can place a series of relatively inexpensive directional bets before things get out of hand. Then, laugh all the way to the proverbial bank when volatility explodes and prices go along for the ride.
I’m still doing a lot of research in this department. Today’s markets trend toward complexity, but the irony is that simplicity is where the real profits hide if you know what to look for.
The most straightforward and potentially profitable way to trade a situation like this is to buy a string of weekly put options at the money, but inverse funds could make a great alternative, too. That’s where gamma – a measure of price sensitivity – is highest and you get the best bang for the buck.
Even the normally taboo triple leveraged funds I encourage you to avoid like the plague could be a great source of profits as would various spread combinations. All of which, I’m still researching.
Speaking of which, I want you to do me a favor.
Post your best questions and share this column with friends who you think may have an interest in the predictive work I’m doing through my sister company, Fitz-Gerald Research Analytics. That’ll help shape my research efforts.
Depending on what we find together, it could even help me help you make a fortune.
Which is, of course, why we’re all here and in this together.
Have a fabulous weekend!
Until next time,