Kirk's Hammer: Closing Line Value, Political Betting and Knowable Events

On Tuesday night, I was waiting to board a plane that was a few hours delayed—the ideal setting to start a long-winded argument over text, fueled by boredom. I was messaging a bettor I very much respect, and we were discussing the Zohran vs. Cuomo election that had happened earlier that day. We were talking politics betting in general—how sports bettors fare in political markets and whether they have an edge. His hypothesis was that the market had gotten so much more liquid from 2020 to 2024 that similar strategies in 2020 probably wouldn’t work in 2024. I somewhat agreed, but then I said, “I’m not certain that Kamala was a bad bet in ‘24.” He responded, “Kamala bet the day of the election was undoubtedly bad, because really no variance to play out.” I disagreed.

We weren’t really discussing closing line value here—more so whether a bet is automatically bad if the results are knowable and you get it totally wrong. The next day, I tweeted a poll with a slightly different hypothetical: “If you bet Kamala Harris to win the presidency at 37% (closed around 42%), that bet was: Good or bad?” This tweet was very interesting because something that rarely happens on Gambling Twitter occurred—the smart people disagreed. Not only was there disagreement, but very confident assertions on both sides that it was a good bet and that it was a bad bet—both from people I respect.

Before I get into why I think the bet may have been good or bad, my main answer is: I’m not really sure. I rarely bet politics, and it’s far from my expertise. That being said, I think this scenario goes far beyond just political bets. The real crux of the question is: how should you evaluate bets past the close, and how do you differentiate that analysis for a “knowable” event like an election versus a more random event like an NBA game?

If you are a bettor, the close can be your north star forever, and you’ll be fine. Realistically, if you consistently beat the close in pretty much any market, you’ll win money—and if you can do it for a long time, you’ll win a lot of money. But once you bet enough, you realize that the close means a lot—but certainly not everything. However, once you start questioning the close, you allow a lot more human bias to creep into your analysis. Like everything in life, it’s a balance.

OK, back to the topic at hand. I’m going to post pictures of what I thought were the most interesting responses and interactions from that poll and write my thoughts on what I think said people got right or wrong.

--

First off, we’ll start with fellow Canadian Peanut Bettor. His argument is essentially that the bet was bad because a “good bet” is one where the probability of what you’re betting on is greater than the number you’re betting it at. And he’s arguing—rightly—that when you placed the bet on Kamala, the true odds were certainly less than 37%, so the bet was bad.

However, I very much disagree with Peanut here on his way of evaluating bets. I think a much better guiding principle is: was the bet good with all knowable information at the time?

Yesterday, I bet on Indiana Fever over 171 against the LA Sparks. Later on, Caitlin Clark was ruled out due to injury. The line had gone from 171 to 174.5, and then dropped to 167.5. At the time I clicked over 171, it was definitely under the 52.5% I needed for the bet to be +EV, because Caitlin very likely had already been determined to be out—but that info was not known yet. To me, that is still a good bet, despite being bad according to the rules Peanut posted.

I think it’s a very different story if I clicked over 171 and it immediately got jammed back at me, and I took more—then someone likely had the info, and it was a bad bet. But the line moving to 174.5 likely means no one in the market had that info, and 171 was a good bet despite technically being negative EV when I clicked it.

This is a very important distinction for the Trump–Kamala market as well, and why I think 37% could have been good. If someone—or a group of people—in the market knew Trump was very likely to win, they would have kept betting and betting and driven the price. But you can only compete against the information that is in the market, and then adjust somewhat based on the information you get after the event occurs. Just because the true probability at the time was under 37%, doesn’t necessarily mean it was a bad bet.

This is especially true because the man now known as the “French Whale,” who was seemingly Trump’s biggest backer, didn’t appear to have wildly superior info. If you’re not familiar with him, he was betting tens of millions on Trump on Polymarket, consistently moving the price toward Trump. After the election, he did a few interviews about his bets, and based on what he said, he seemed to really rely on “neighbour polling” — where the poll asks who you think your neighbour would vote for, instead of who you're planning on voting for.

It’s possible his intuition was right and that this type of polling ended up being more accurate than traditional methods. But it’s not like he had some crazy information edge that no one else had. I’m skeptical that was even the case — and it’s also very plausible he just got lucky. Either way, it doesn’t seem like he had any sort of proprietary insight, which to me leans even more toward the possibility that Kamala bets were good.

--

Then there was this interaction with Canzhi, Gerry, and Kostya. I’ll be honest—I had no idea what the difference between aleatoric and epistemic uncertainty was before Canzhi posted this.

I believe what he is saying is that in a basketball game, there is aleatoric uncertainty—pure randomness—versus an election, where if you had complete data (perfect polling), you would know exactly who is going to win with essentially zero uncertainty. And since the election was essentially determined at the time of the bet, the 37% on Kamala was bad.

Gerry and Kostya both disagree—but in somewhat different ways. Gerry argues that since the information in the election isn’t known, it might as well be the same as the randomness of a basketball game. I agree. I’m not sure how much it matters that the information is hypothetically knowable if no one in the market actually knows it. There are over 300 million people in the U.S., and most bets are just people trying to interpret incomplete polling data. The uncertainty of how people will vote essentially mimics the randomness of a basketball game—it doesn’t matter that the info is technically knowable if it’s not known.

Kostya argues that if you had perfect data and perfect modeling capability (we don’t), you could model out the bouncing of a ball in a basketball game or anything else we currently treat as “random.” In a later tweet, he says that with enough data and modeling, any aleatoric (random) uncertainty becomes epistemic (lack of information). This is a bit abstract, but the gist is: randomness is just our inability to see the full picture.

Again, I mostly agree. The results of an election feel much more knowable than the results of a basketball game, and in some ways they are—but a lot of what we call randomness in sports is just our lack of perfect info.

However, a key point against Kostya is that if the election were held again the next day, Trump would be a 99%+ favorite. In contrast, if two basketball teams played again the next day, the previous result would matter, but the market wouldn’t move nearly as much.

--

Then there was this debate between Shipper and Pads. Shipper argues that you shouldn’t stop evaluating a bet at the close—the actual result should inform the quality of your bet. I totally agree. But as I mentioned earlier, this introduces a lot of bias that just using the close avoids.

A good example: let’s say you bet a player over 35.5 PRA because the market projected them to play 25 minutes, but you’re convinced they’ll play 32. The line never moves and closes at 35.5. If the rotation goes exactly how you thought, and the player gets 32 minutes, and the next game the line opens at 37.5—that bet was likely good, even though it didn’t close.

But if the player plays only 25 minutes, gets a massive usage bump you weren’t expecting, goes way over, and the next game opens at 37.5? That bet was probably bad. The prediction was wrong—you just got lucky.

This is mostly where I land on the Kamala Harris 37% bet too. As annoying as it is, the only real answer is: more context is needed.

Let’s say you have a unique perspective on Mexican Americans, and polling shows them 50/50 for Trump and Kamala, but you’re extremely confident that Kamala will win that group 65/35. You know that shift would significantly boost her odds overall. If you bet Kamala at 37% and that prediction was spot-on—but the rest of the electorate swung to Trump—you probably still made a good bet.

But if you bet Kamala at 37% just based on general polling and thought you had an edge, and Trump won comfortably despite the bet closing at 42%—I think it’s much more likely that was a bad bet.

Just like Shipper said, evaluating bets based on the close or other heuristics is useful because it helps you assess how accurate your predictions are. If your prediction was right while assuming the rest of the market was too, but the rest of the market just happened to be wrong, your bet was likely good. If the bet moved your way but the premise was totally off—your bet was much more likely bad. Of course, how much the bet closes also matters. If you got Kamala at 37% but she closed 90%, no matter what happened in the bet I would say that is a good bet. That being said, I don’t think that necessarily means that 37% is a good bet if it closed at 42%, the degree of close of course is very important here as well.

At the end of the day, figuring out whether a bet was good or bad is hard. And that’s exactly why the close is so valuable — it’s a clean, consistent benchmark that helps you avoid tricking yourself. But if you want to go further, and actually evaluate bets past the close, it gets messy fast. You need to ask what you believed that the market didn’t, whether that belief ended up being right, and if the outcome tells you anything useful going forward. It’s not about whether you won or lost — it’s about whether your predictions and assumptions held up when they were tested.

--

Sign up to Underdog and deposit using code CIRCLES to receive up to $1K in bonus cash + a gimme pick or simply click this link: http://play.underdogfantasy.com/circles

Get yourself an Edge Boost Visa Debit to eliminate the headaches of deposits and for cash back on your betting: https://www.edgeboost.bet/guide/?oid=196&affid=117