What this covers

Patrick talks with Jeff, whose background runs through blackjack, sports analytics, and technology companies, about how data becomes decisions. The conversation starts at the blackjack table. Basic strategy cuts the house edge from roughly 3 percent of money wagered to about 0.5 percent, yet many people who know it drop it for gut feel. From there it moves through sports betting, team analytics, hiring, and product metrics at Twitter.

The sports material is the most detailed. Jeff says sports data was long poor because it recorded what was easy to capture, not what predicts outcomes, and that player tracking and cameras measuring launch angle and exit velocity now allow predictive models. On betting, he says Rufus Peabody found a golf edge that has since eroded, that perhaps 1 percent of bettors or fewer beat the market, and that the common trait among strong bettors is fortitude rather than brilliance. The contested claim is that blind intuition is unreliable. Jeff qualifies it with Bill Walsh, whose sense of the yardage a first-and-10 play needs matched a model's figure of a little over four yards.

The rest covers how judgment goes wrong and how organizations cope. Omission Bias explains why players decline the correct hit, and judging by results explains why even the Patriots kick field goals the analytics reject. The "beer test" in interviews measures similarity, not ability. Jeff describes Twitter moving its target metric from monthly to daily active users, and engineering analytics tools that only worked once managers mandated them. He also argues for hiring people trained in the scientific method, and says streaming has not yet changed media.

Sharpest takeaway

Jeff Ma argues that analytics only beats human judgment when it rests on good data and is continually re-evaluated, because environments are non-stationary and human biases such as results bias and omission bias distort how decisions are judged.

  • Sports, betting markets, and tech companies change constantly, so static models decay and target metrics need reevaluation
  • Judging decisions by outcomes rather than process, and favoring inaction, lead people to make predictably poor choices
  • Good data must be captured on purpose, and organizations often have to change how people work to produce it

The argument · threads1 threads · 61 claims
0.71

Exploiting biases or uninformed participants is rewarded in platforms and public markets.

2 pointscentrality 4/5
  • Because lies travel faster than the truth and play to existing beliefs, platforms may implicitly reward people who exploit biases to grow reach, concentrating influence among those most willing to manipulate.

    “the whole idea of lies traveling faster than the truth. That if you can play to people's biases or tell them what they believe already, that's one of the fastest ways to grow a reach.”

  • Michael Mauboussin has argued that in public stock markets the dumb money has been pushed out of active trading into index funds, leaving less uninformed participation to exploit.

    “people like Michael Mauboussin have argued in public stock markets that the dumb money, so to speak, has been pushed out”