On Monday Jason Mercier won his 4th World Series of Poker (WSOP) bracelet in Event #16: $10,000 2-7 Lowball Draw Championship, and I wrote about what it meant for his side bets. Now, in the blink of an eye, he’s looking like he has a strong chance of making it five. Only eight players remain in Event #20: $10,000 Razz Championship, and Mercier has a substantial lead on the pack with nearly a third of the chips in play.

This is almost unbelievably terrible news for Vanessa Selbst, who stands to lose an eye-watering $1.8 million if Mercier manages to pick up three bracelets over the course of the summer. That’s not an unheard-of feat, but it is a rare one. Only two people have managed it in the modern, post-Hold’em boom WSOP with its large field sizes: Jeff Lisandro in 2009, and George Danzer in 2014. Selbst accepted a $10,000 bet from Mercier at 180-1 odds, just two days after she’d made a similar bet at 200-1 odds with Dzmitry Urbanovich.

The bet with Urbanovich was widely publicized because it was made in the presence of members of the media, but although the second bet with Mercier was known to members of the pro community, it had mostly flown beneath the public’s radar until Mercier won his first of the summer. Selbst says alcohol was a factor in both bets and that she began regretting both soon after they’d been made.

She’s made no attempt to get out of her bet with Urbanovich, however, yet has made two offers to Mercier to annul theirs. The morning after the bet was made, she offered him $1000 (plus his $10,000 stake) to back out, which he refused. After his win in Event #16, she offered him a $100,000 buy out, which he likewise refused.

Fans of Mercier, including his girlfriend and fellow poker player Natasha Barbour, have been critical of Selbst for creating “drama” about the bet only once it started to look more likely that Mercier could win it. Selbst has explained that it was known that she regretted her bet with Urbanovich, and that a sober Mercier took advantage of her in an “almost blackout drunk state” to rope her into a similar bet with him. She feels that the circumstances of the bet are such that Mercier should have accepted her buyout offers out of compassion and friendship.

To be fair to Selbst, the only public reference she initially made to the bet was a joke about therapeutically favoriting Mercier’s bustout tweets. It was Mercier who first went public about the bet, talking about it with Remko Rinkema on his podcast for PokerNews. Selbst’s disclosure of the circumstances of and her feelings about the bet came only once she found herself being trolled and criticized for it on social media. As for why she has pressured Mercier for a buyout but not Urbanovich, she says that in Urbanovich’s case, he was as drunk as she was, and is not a close friend like Mercier.

Indeed, to anyone who is not a professional poker player themselves, the idea of talking a drunk friend into a bet they wouldn’t make sober, then refusing to let them out of it in the morning does seem quite distasteful. But professional gambling is inherently predatory, and Selbst is a predator in that regard herself. If the situation were reversed, would she have let Mercier off the hook easy, or held him to the bet? Hard to know either way, but I definitely wouldn’t feel confident in saying she would. Drunk or not, as a professional gambler, I’m sure Selbst knows well the company she keeps, and isn’t in the habit of making bets with the expectation that she’ll be allowed to change her mind later. Letting people out of huge bets is certainly not the norm in the gambling world, even among friends.

So, although the “nice” thing for Mercier to do would have been to accept Selbst’s immediate $1000 buy-out offer, he’s within his rights to refuse it, and Selbst really should have known better. “Don’t make a bet with a 7-figure downside without at least sleeping on it,” is good general life advice even when you’re not under the influence, and assuming she really wanted out in the morning, there likely would have been some price at which he’d let her buy out.

Although no one would relish being in Selbst’s situation now that Mercier has one bracelet in the bag and another in reach, one interesting aspect of this story is that even fairly knowledgeable people are having a hard time deciding whether it was actually a bad bet to begin with. Evaluating the probabilities involved in long-odds bets is always tricky, as a 1% chance and a 0.1% chance feel similar at the level of human intuition, but give radically different expectation values for a bet like this. Matt Glantz, a friend of Mercier’s, thinks the correct odds would be 50-1, while FlushDraw journalist Haley Hintze’s impression was that Selbst had the better end of things at 180-1.

One easy place to go wrong in looking at the bet from Mercier’s side is in assuming that bracelet-winning odds multiply straightforwardly. Your odds of getting dealt a pocket pair in Hold’em are approximately 6%, so your odds of getting two in a row are 6% x 6%, or about 0.35%. It’s tempting then, to say that if you estimate that Mercier is 25% to win a bracelet in a given year, that he’s 25% x 25% x 25%, or about 1.5% to win three, which would give you 60-1, close to the 50-1 odds suggested by Glantz. I don’t know if that’s how Glantz did his calculation, or if he just thinks Mercier is much better than 25% to win the first, but if it’s the former, he’s wrong.

Statistically, the first bracelet won by Mercier should have come about halfway through the series on average (though in reality it came much earlier). That only leaves him half the series, on average, to win the second. So if he’s 25% to win one when he has the whole series ahead of him, he’s only 12.5% to win a second, given that he’s won the first. With two bracelets won, the second will on average be 66.7% of the way through the series, so the third is only 25% x 33.3% = 8.3% to happen once the first two are in the bag.

So, if we estimate that he is in fact 25% to win one (which may or may not be a good guess, but is probably within 10%), his total odds of scoring three are only 25% x 12.5% x 8.3% = 0.25%, or 400-1. If that’s the case, then Selbst’s bet looks very good indeed. However, it’s important to note that these odds are extremely sensitive to small changes in that initial estimate; if you instead set the line at 30% for Mercier to win one, then he’s 30% x 15% x 10% = 0.45%, or about 220-1 to win three, and it’s much closer. The break-even point for Selbst’s bet, using this math, is when we postulate a Mercier who wins one bracelet just over 32% of the time. For Glantz’s 50-1, we would need him to win nearly 50% of the time, which seems unlikely.

In Daniel Negreanu’s famous 2014 prop bet, he wagered that he and Phil Ivey were better than even money to win at least one bracelet between the two of them. Assuming he considered each of them equally likely, that implies odds of about 30% for either. So, unless we believe Mercier is a significantly better than Negreanu and Ivey, Selbst’s bet may have been good on paper.

There are a few problems with this way of thinking, however. Firstly, in estimating a given player’s odds of winning a bracelet, it’s natural to look at their past performance, but in reality, a player with substantial bracelet bets on the line should in fact win bracelets at a higher rate than their usual expectation. This is because the bets change their effective payouts and therefore their playing strategy, as I discussed in yesterday’s article; a player with additional money staked on a first-place finish should win somewhat more often, but show lower in-the-money (ITM) and return-on-investment (ROI) stats, because chip accumulation is more important to them than survival.

Secondly, when the odds are as long as they are, and one party’s upside is as large as it is in the Mercier-Selbst bet, the possibility of collusion looms large. If Mercier wins the Razz bracelet tonight, he’ll only need one more to book a $1.8 million score, larger than any payout in almost any event left on the schedule. I don’t believe Mercier would straight up offer to pay someone to let him win, but it’s not hard to imagine that a friend of his, or a friend of Selbst’s, could consciously or unconsciously alter their play to make it easier or harder for him to win. As extreme examples, imagine Barbour and Mercier at a final table together, after he has two bracelets in the bag, or conversely Mercier and Selbst herself. The potential for scandal and hard feelings is huge.

Finally, when it comes to extreme cases like this one, we have to consider the emotional factors involved. Theoretically, professional poker players are supposed to be risk-neutral, but no one is truly risk neutral. For players as successful as Mercier and Selbst, $10,000 is essentially “fun stakes,” but $1.8 million most definitely is not. Selbst says she felt extreme anxiety in the aftermath of making the bet with Urbanovich and that her bet with Mercier was due to having convinced herself that making a second similar bet would somehow lessen rather than increase the anxiety.

That’s about the most degenerate thing I’ve ever heard, but it illustrates the problem with looking at these bets the same way as ones with more typical odds. Flipping a coin multiple times or playing lots of tournaments does indeed reduce your variance, it’s true… but when you’re praying to dodge a bullet, trying to dodge two at once is not an improvement. For her sake, I hope this is not a lesson Selbst has to learn the hard way.

Alex Weldon (@benefactumgames) is a freelance writer, game designer and semipro poker player from Montreal, Quebec, Canada.