“Another big mistake is testing only things you hope are good.” featuretools) or convolutional ones (deep learning). There are very few public tools that help you understand the hidden relationships between these things, unlike relational systems (see e.g. Stock market data (among many other things) consists of both time-series (daily open/close) as well as discrete events (earnings, dividends, news). On the more technical side, I’m working on automated feature extraction in mixed time-series/event data.
In layman’s terms, this means I’m programming my computer to make a lot of small trades in the stock market. Right now I’m working on an automated trading strategy in thinly traded stocks and ETFs. The modern term is “data scientist”, but I like the old school term better. For those who don’t know you, do you mind sharing with our audience a bit about yourself and what you are doing these days?Ĭhris: Broadly speaking, I’m a quantitative developer. Rommil: Hey Chris! I’m really stoked to chat with you today.
We covered a lot of ground including his perspective on the mistakes that Experimenters often make around KPI-design, statistical independence, and the value of just running tests despite your traffic levels. Chris, a self-described “Compulsive gambler”, is the mind behind VWO’s Bayesian testing engine (you’d be right to think he is biased). I recently had a wonderfully in-depth conversation with Chris Stucchio. A Conversion Conversation with Chris StucchioĪ/B testing, while deceptively straight-forward and beneficial, presents many opportunities for misinterpretation and generating unreliable results.
0 kommentar(er)
