Overall questions:
What (incl indiv stocks and duration) and When, in what allocation, when rebalance, how enter and exit? (insight 06-20)
“If you over complicate it, the more likely you are to just overfit to past noise, and fail to make money in the future – even if you were right about the general idea.
When I talk about “optimisation being bad” I mean a blind numerical optimisation without viewing the distribution of the data, checking its validity, understanding the trade-offs. You can’t outsource hard decisions about strategy design and configuration to an optimiser. You have to plot, investigate, understand.
I don’t know anyone who has successfully outsourced alpha generation to a neutral network… Nearly all successful professional trading operations are using machine learning techniques to solve large data processing problems, not to generate directional alpha.
Don’t fall into the trap of thinking the trading game is won by those who are the smartest or the best at maths, or coding, or who know the most about statistical learning techniques, or can derive black-scholes with their eyes closed…. that stuff isn’t the stuff that makes the difference. You have to be a pragmatic generalist, capable of grasping the big picture, and effectively implementing the things that matter.
The practical lesson is to always relate things back to the market – and make sure you could explain it to a small child without jargon. You think you need to understand these formal mathematical concepts, but you don’t. You need to get a really good intuition for the dynamics of asset prices. Once you have that, then understanding the maths will be easy. The master plots a bunch of time series charts and return distributions from various subsets of the data and makes a value judgement.” (James)