Ober: Third Point requires ML stock picks to be explainable to managers
Matt Ober · Fireside Chat: Matt Ober, Chief Data Scientist at Third Point (FirstMark's Data Driven NYC) · Feb 25, 2019 · at 29:47
Matt Ober, Chief Data Scientist at Third Point, explains how his team bridges machine learning output with fundamental investment analysis.
0:00 / 0:30exact quote · 30.1s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“When the model creates a list of companies for us that we should go invest in, we need to be then able to explain to our you know, PMs and analysts internally what's driving that decision, and they really want to get into the deep roots of why it's picking those companies, how they performed over time, and for us, just throwing things that aren't explainable at them, and not being able to really dumb it down into What are the core fundamentals it's picking? Why did it pick that? How did they react with each other? Is kind of what we spend a lot of time thinking about.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Matt Ober
Ober: Fundamental funds need more data per company than quantitative funds
“So it's interesting because when you move from quantitative to fundamental, the amount of data that's relevant is, you know, Enormous on the fundamental side, because you're really diving into these single names.”
Ober: Satellite car-counting data rarely drives hedge fund investment decisions
“I don't think that there's a lot of firms really counting cars in parking lots, and that's what's really making their, ah, decision making.”
Ober: Quant funds require two years of backtestable daily data before purchasing
“If you're going to sell to a quantitative hedge fund, anything less than two years of daily data that's really been stored point in time that you can really back test is really going to be tough to make that ten million dollar sale.”
Ober: Simpler financial models outperform complex machine learning over time
“Going back to the very basics, you kind of find that, like, the simpler it is, the better it actually performs over time, if you're not trading at this higher frequency.”
Ober: Fundamental funds prioritize good communicators over the smartest PhDs
“We may not need the smartest PhD within a fundamental shop because we have to be able to explain and understand what they're doing and really talk their language whereas a quantitative hedge fund, there's more opportunity to maybe work on your own and not have…”
Ober: Fundamental hedge funds began data transformation around 2016–2017
“Sure, so I think kind of a revolution in the hedge fund industry mainly because you have the quantitative hedge funds that have really been leaders in technology and data, and now you have the fundamentals really over the last, call it two, three years, thinki…”