First of all, thanks to everyone who’s submitted so far to The Barking Crow’s Best of 2022 collections. If you’ve yet to make submissions and you’d like to do that, send them over to allthingsnit@gmail.com, or message us on Twitter or Instagram. Categories remaining are…
- The Barking Crow’s Best Work of 2022
- The Barking Crow’s Silliest Work of 2022
- The Barking Crow’s Right-est Work of 2022
Now. The Barking Crow’s Wrong-est Work of 2022:
You Don’t Have to Be Good at Football
Yes, I’m still bitter. It worked out, but I’m still bitter.
After Thanksgiving weekend, when Michigan pulled away late from Ohio State and USC held Notre Dame at bay, our previously very accurate College Football Playoff rankings model projected Ohio State to comfortably hold onto the fourth-place ranking in the country. Maybe, we hedged, USC would jump Ohio State if they won the Pac-12, but surely the committee wouldn’t stoop to such a level as to put a team ahead of another when every piece of the other’s résumé except one (margin of single defeat) was the better. It turns out, the committee is more a servant to the whims of narrative than we realized. Now, we have to go back and reevaluate a model that gave Ohio State a 99% chance of making the playoff entering Thanksgiving Day.
Trade Markets and Free Agent Markets Aren’t the Same
The thing about being wrong (and these are all me being wrong, that’s why I have to write this, Stu gets to do the thing where it’s unclear what’s serious and Stuart gets to call all his stuff “art”) is that I still think I’m right, and that betting on the take just didn’t pay off. This is probably hubris. This might be why we’re doing this exercise. Maybe I need to learn.
Entering the 2022 MLB season, the narrative in the Chicago Cubs blogosphere was that Willson Contreras was guaranteed to be in his last season with the Cubs. This felt overconfident to me, and in response, I probably became overconfident. In May, I pointed out that if the Cubs hadn’t traded Willson Contreras in the offseason, they weren’t particularly likely to trade him at the deadline, and that meant they might believe the market would come back around to them. In August, after the Cubs didn’t trade Willson Contreras, I took it as a clear sign that they intended to eventually extend him. In October, after the season ended, I said a bunch of the same things I said in August, but I did it again because I didn’t feel people had listened enough the first time.
I do still think the logic holds—I think the Cubs’ front office may have thought in March that the market would come back around—but the piece I misread, and this is a good thing to learn, is how the trade market is not all that efficient. Needs have to match needs, and when you only have thirty franchises as the market participants, that makes it difficult for trades to happen. Were there one thousand teams and massive active rosters and sizable trading desks within each front office, each player might end up in the place that valued them most, but that’s not how it works in Major League Baseball. Prior to 2022, not many teams needed a catcher. At the trade deadline, even fewer needed one. After 2022, plenty did, and the efficiency (to the degree it exists in free agency, which is higher than the trade market but not all that efficient itself) kicked in: Someone who valued Contreras more than the Cubs landed him.
Our Bracketological Struggle
I’m going to tell you guys a secret, and I need you to keep it extremely quiet: Our bracketology has a horrendous track record. Part of this is an easy fix—we’ve realized that the committee doesn’t use the same criteria when selecting the field that it does when seeding the field (the criteria that determine whether a team is seeded 8th or 9th is different from the criteria that determine whether a team makes the tournament in the first place), with a lot of help from the industry. The other part, though, is harder: We often really miss who makes the field. This March, entering Selection Sunday, we named twelve teams who were on the bubble for what eventually became six spots. Our model ultimately only correctly identified which tournament four of the twelve reached. 64 of 68 teams, in bracketology, is not particularly good, and it’s not like we’re blowing the NIT side out of the water.
How are we addressing this, as we prepare to launch this year’s bracketology? Part of it is just getting better, and given our approach has very heavily weighted recent tournaments, that should happen with natural data, but part of it is also putting a broader emphasis on our percentage probabilities, which we need to bring back now that the schedule is more assured again, with Covid cancellations on the way out (please join me in knocking on wood—I do not want this post to be in next year’s wrong-est compilation). Anyway, the point is that we got the bubbles very wrong on Selection Sunday. And we still had far and away the best March, site traffic-wise, in blog history! What an industry.
Don’t Bet on NASCAR
For our final miss, there isn’t one post we can link to. We can merely refer to the totality of our bets, where we dug ourselves a gigantic hole this spring. At the moment, our bets are 87.82 units below water. On the year, our NASCAR bets were 69 units below water, and when you add IndyCar and Formula 1—for which our approach was similar early in the year but managed to recover down the line—the total motorsports deficit on the season was 99.74 units, for a -53% return. This was very, very bad. It was nearly much better—we had Hamilton at Silverstone, we had LaJoie at Atlanta, we had plenty of other very near misses—but it was bad.
Our problems were twofold: First, we trusted ourselves to get our models built as the season went on, so we thought it worthwhile to use very basic models and take shots in the name of content, but because we never finished those models during the relevant seasons, we were somewhat guessing the whole way through. Second, we didn’t understand the markets well enough to take a very mature approach. We solely bet on winners when we could have been taking more reasonable risks on top-ten finishes in NASCAR and podiums in the open-wheel stuff.
There’s good news and there’s more bad news and there’s something peripheral to address with this. The bad part is that we actually lost even more than 99.74 units because of this effort—we would have hedged more aggressively on our World Series bets, capturing a larger profit, were we not tied to the Hail Mary attempt that our MLB futures became once we locked in profitability. The peripheral part is that we also had some rough results on college basketball futures right around the time the NASCAR losses were piling up, and those left us blind to how much we were losing because we kind of Bankman-Fried-style blindly trusted they would turn out well. The good part is that we did probably have some bad luck. Our close losses were frequent. Our close wins were rarer than rare. Even using this same terrible approach again, we could possibly profit by small-sample luck, and we do seem to have a decent grasp on how to construct our models for these three sports. But boy, did we mess up the bets in 2022. We need another election, and we need it soon (but that’s something for tomorrow’s post, in which I get to talk about all the things I got right).