Member-only story
While researching Machine Learning and algorithmic models at my day job and building models to trade in my personal trading account, I’m thinking that the benchmark to beat is not another complex system, but simple rule-based systems.
Example 1: Simple keyword matching vs. natural language processing, cognitive technology, search algorithms, etc.
Example 2: Simple brain dead rules like passive index investing in tax deferred accounts vs. a complex quant trading platform
Makes me wonder…
Assume you can make $X in a personal trading fund using simple rules. Now consider dedicating massive effort in terms of time, energy, and thinking to make $X + $Y.
The real question is not whether making the additional $Y is possible, since it is with enough research and effort. The real question is the marginal time, energy, and thinking Z to make the marginal $Y. The problem is that a lot of Z can make only a little $Y, and could actually make some “negative” $Y.
Example 1 (continued…): Simple keyword matching with a simple decision tree can be vastly superior to cognitive search algorithms with sophisticated natural language processing.
There’s more to consider though: long tail search, highly expert knowledge, personalities knowledge, etc.
