Sign up to gain access to our free alerts service. Set price or indicator alerts for any cryptocurrency Sign up · Login

Understand your Limitations and Trade on What you Know

Tuesday, August 28, 2018 at 9:43pm
It's important to understand what you can predit and what you can't. When I gave up trying to pick individual stocks, I started making serious money.
The “aha” moment arrived in my life when I realized two things:
   1. I understand math,
   2. I can’t pick stocks (and I shouldn’t try).
My personal background is math, and I decided to run with that. 
I'm starting to ignore the investment pages of the Wall Street Journal. I still watch Kramer, but it's for the entertainment value of seeing his head explode. 
The first fundamental question is this... How accurate is our “bitcoin opinion”? Can we trust it? Is our bitcoin opinion better then a monkey and his darts?
My main goal is to model the uncertainty of our predictions. It’s foolish and naive to believe that we’ll be right 100% or even 66% of the the time. My main goal is to be able to create a prediction, and then be able to accurately model the uncertainty of our prediction. Once we can do that, then we can provide information that is actually worth looking at.
The first item I’d like to look at the accuracy of a (modified) 3rd order best fit polynomial approximation of the last 30 days of bitcoin price data. Once we’ve generated the polynomial, we’ll us it to predict the price 7 days from now.
I'm very intersted in this analysis for one reason... I uploaded several bearish candlestick patterens into Excel. I rand the numbers, and my spreadsheets told me to sell. When I loaded bullish candlestick patterns into my software, the software told me to buy. So, I'm very intrigued by the fact that my algorithms are providing recommendations to candlestick patterns. There is a significant amount of agreement between the two. For me, this is exciting. 
Each week, I will make a prediction for the price of bitcoin seven days from now.  Next week, we’ll come back and see how close we are. We’ll do this once a week for the next 20 weeks. Each week, we’ll write down our prediction, and how far off we were.
At the end of 20 weeks, we’ll have good data that shows how accurate this model is. We’ll be able to create a prediction, and also create an experimentally determined cloud of uncertainty that someone (such as yourself) can use to get a sense of the probability density function of all possible bitcoin price movements over the next seven days.
I’m not expecting anyone to blindly trust these graphs. However, it is useful information that someone could use as they go about trading bitcoin.