Why cooking made me a better data analyst and investor
You can ruin a perfectly good steak simply by not letting it rest. When the steak is done, the juices collect in the center of the steak. It’s like when you feel hot. You try to find a cool place to hide. When you let the steak rest, you let it cool and the juices diffuse through the rest of the meat. Cutting the steak too soon releases the juices that have accumulated in the center of the steak; thus ruining a perfectly good steak.
Investing is similar in that if you rush into a popular stock too soon, you may not like the outcome. What expectation does is that it helps us confirm our assumptions while decreasing risk. The trade-off is that you lose the capital gain potential. However, this is better than selling at a real loss.
Accuracy is the closeness of the measurements while precision refers to the closeness of the measurement to the acceptable value. In the kitchen, it is better to be precise than exact. This means that if the recipe calls for 15ml of sauce, one tablespoon should suffice. You don’t need to measure exactly 15ml.
I have seen in data analysis that managers tend to focus a little too much on precision. It is only when they realize that being precise takes too much time that they are satisfied with fairly good results. This means that if you are making an investment decision, there is no need to get the exact numbers. Instead, favor a series of ballpark numbers over a period of time. This way you save time and mental space.
One of my favorite recipes is chicken teriyaki. The ingredients are chicken, sugar, mirin, sake and soy sauce. Then, everything is in the cooking techniques so that the dish is good. In my experience, the simpler the recipe, the more delicious it is.
Good machine learning is not about having a lot of data thrown into a black box algorithm. The best machine learning systems use the minimum number of variables and the right algorithm. For example, a computer vision algorithm for recognizing dogs should have mostly images of dogs as training data with a few false positives, and the data should be fed into a convoluted neural network – not a multilayer perceptron neural network, which may work for recognition but it’s just not as good.
Truffles are delicious, but if you’re making beef stroganoff, regular button mushrooms should suffice. It’s a waste to use truffles in beef stroganoff. Sure, it may take the dish to a different level, but it doesn’t redefine the dish.
Often, investors try to find competitive advantage with exotic data. For insider trading, this has its advantages. For example, I read this newspaper article about this investor who made millions in currency swings by convincing a friend from the national statistics office to give him the numbers ahead of upcoming official announcements. Both were sent to prison because of it.
Other times, investors buy complicated analytical reports with fanciful predictions. They may look nice and give you the impression that you have some kind of advantage, but they tend not to live up to their expectations.
Warren Buffet doesn’t need exotic data. He just looks at the public financial statements that are on the annual reports.
If you’re cooking for someone else, you’d do your best not to serve them burnt food. Likewise, before you finish cooking, don’t you taste the dish to see if it needs a little more salt?
Likewise, if you’re investing, there’s no need to do it all at once if you think you’ve found yourself a ten-bagger. It’s perfectly fine to put money in first and wait a bit to see if the market agrees with you before investing more money. In fact, you can buy the same stock over a period of time. Who said the first purchase is the only purchase?