If I want to get something done, I find procrastination is the optimal strategy. Well, I suppose optimality is dependant on the time horizon. In the short term, it is nearly always optimal. In the long term, it is less clear whether procrastination will always be an optimal strategy. Eventually though you’ll have to do something!
In markets as with anything, we have finite time. We can’t do all the analysis the want to do, and need to somehow focus our energy on the place where we’ll get the most return. Quant space is no different. I would say that in quant it’s just as easy to go down a rabbit hole of spending a lot of time and effort on the wrong technology, wrong approach etc. as it is for discretionary fund managers.
If you are trying to move your discretionary process towards incorporating quant tools and Python, how can you avoid the missteps? What should you focus your initial energy on this move? First, you can try taking small steps. Yes, I know machine learning is a buzzword, but get other things done first. Are focusing on training your team so they can use Python (and yes, I do run Python for finance workshops if you’re interested in hiring me!)? Are you investing in a database to store all the data you need? What data are you capturing? If you haven’t given much thought to the data, then you can’t really go any further in any analysis. Of course, data storage doesn’t sound as cool, as machine learning though!
Once you’ve sorted out all your data feeds and data storage issues and have begun to train your team, you can think about what analysis you can do. Probably the best place to start is on those tasks which are heavily manual. I have spent far too much time in the past updating Excel spreadsheets, which are well in excess of 100 MB – and I really don’t wish that upon anyone else! This might require a bit of brainstorm, to understand what manual processes you have across your team and working out which are in most need of automation. The icing on the cake is to fully automate reports which use these spreadsheets.
Then you can explore doing additional analysis using quant tools, which you may have never been doing before. This might involve machine learning, but in practice, doesn’t always need to be. Also you can start looking at alternative datasets to see what value it can add to your investment process. In all cases, it’s a gradual transition to using quant tools, like Python, and a graduated approach, rather than rushing towards the buzzwords immediately.