Sports Analytics Weekly by kubeia.io - 36/2024

In the lead-up to the 2008 election, Nate Silver revolutionized the way we talk about politics, bringing cold, hard, numerical facts to a world that had been dominated by the gut feelings of reporters and opinion columnists. Sixteen years later, he remains a go-to authority on the use of quantitative information in the prediction of political outcomes. But despite his popularity, Silver feels misunderstood. In a 2023 interview with New Yorker columnist Jay Caspian Kang, Silver lamented, “It’s very weird to become very well known for the wrong reasons. People say, ‘Oh you have numbers and therefore a lot of certainty’ and they can’t quite process the fact that you can use numbers to quantify uncertainty as well.” Silver’s readers thought his numbers were telling them what would happen and informing what politicians, particularly Democrats, should do. According to Silver, that’s not at all what he intended, and, he told Kang, he was writing a book to clear up the confusion. At stake was something much larger: not just a question of facts and probabilities, but an entire way of seeing the world, and acting upon it. And the key was not in election forecasting, but another field that was Silver’s true passion: gambling.

Corbin Smith shares some new stories on the unfortunately topical subject of gambling addiction and how it relates to the financing and the sports media. In his article, Smith implicitly makes a strong case that to understand the problem you need to think about interactions between psychology, economics, and politics. The sports, news, and entertainment media are pushing gambling so hard. I guess that in the future we will look back on the present era and laugh/cringe in the same way that we laugh/cringe at the “Mad Men”-style drinking and smoking culture from the 1950s.

Although pass completion rates may still be a mainstay of football analysis on TV coverage, they are no longer considered an accurate representation of player passing skill by many analysts. As we know, not all passes are created equal, and expecting them to be completed at an equal rate is therefore unreasonable. Instead, expected pass (xPass) models have gained popularity due to their ability to consider the execution difficulty of the passes players attempt. StatsBomb released our xPass model in early 2023, but we’ve decided now is the time for an upgrade.

In my first newsletter, I discussed how one of the big goals of this project was to better understand the statistical record of soccer. We now have 15 years of on-ball data, spanning an increasing number of leagues. A typical match has over 2,000 events. In the big five European leagues alone, that means there have been something like 300 million on-ball events logged since 2009-10. It is my contention we should know a lot more about what these data points mean than we do.

So far in our 2024 College Football preview series, THE Coach Matt Edwards has used StatsBomb physical metrics to look at returning players you should know. In this preview, I am going to shift gears and evaluate teams doing interesting things with formations and personnel.On a college football Saturday, if I have no rooting interest during a specific TV viewing window, and there is no big matchup, I usually try and pick a game that will present something odd. It could be an unusual stadium, or uniform, or weather, a weird offensive or defensive style etc. This means I spent a lot of time watching the unique variant on the triple option that Jamie Chadwell and Grayson McCall ran at Coastal Carolina. I always enjoy watching the academies play because the triple option is weird and fun.In this article, I am specifically exploring offenses from 2023 that did weird things worth tuning in to watch. Do not worry, I did not forget about defense, that article will come next.

An AI analyst trained to digest corporate disclosures, industry trends, and macroeconomic indicators surpasses most analysts in stock return predictions. Nevertheless, humans win “Man vs. Machine” when institutional knowledge is crucial, e.g., involving intangible assets and financial distress. AI wins when information is transparent but voluminous. Humans provide significant incremental value in “Man Machine,” which also substantially reduces extreme errors. Analysts catch up with machines after “alternative data” become available if their employers build AI capabilities. Documented synergies between humans and machines inform how humans can leverage their advantage for better adaptation to the growing AI prowess.

Is it really possible to get over 40% annual returns over the past 10 years trading an ML-based mean reversion model?

In this episode I chat with Giuseppe Paleologo – or Gappy as he likes to be called. Currently on garden leave, Gappy has previously worked in Risk & Quantitative Analytics at Citadel, as Head of Enterprise Risk at Millennium, and most recently as Head of Risk Management at HRT.

We begin the conversation with a discussion as to what a quant researcher actually does at a multi-manager hedge fund. As a semi-support role to the fundamental PMs, Gappy explains how portfolio manager coverage, factor hedging, and internal alpha capture can all work together to help maximize firm P&L.

We then discuss the broad field of factor research and portfolio construction, where Gappy shares some of his strongly held views, both on how factors should be constructed as well as how they should be utilized. Topics include returns versus characteristics, mixing versus integrating alpha signals, single- versus multi-period optimization, and linear- versus non-linear models.

Please enjoy my conversation with Giuseppe Paleologo.

Abstract:In this manuscript, we show that any neural network with any activation function can be represented as a decision tree. The representation is equivalence and not an approximation, thus keeping the accuracy of the neural network exactly as is. We believe that this work provides better understanding of neural networks and paves the way to tackle their black-box nature. We share equivalent trees of some neural networks and show that besides providing interpretability, tree representation can also achieve some computational advantages for small networks. The analysis holds both for fully connected and convolutional networks, which may or may not also include skip connections and/or normalizations.

Let’s learn how to perform complex filtering in Pandas.