Week In Review #27 September 22nd-September 28th
I go long on MLB Playoff Odds (Let's Go Blue Jays)
Like many people, I begin my days with a dual dose of caffeine and dopamine. Ideally in the summer, my caffeine intake begins with a large iced green tea I’ve brewed the day before, but it often begins with a stale Diet Coke that my wife opened but did not finish last night. Don’t shame me. I’m tired, I don’t have time to wait for water to boil. My dopamine hit comes from, of course, my phone, but the first thing I look at varies seasonally. When Learned League is in session, I look at trivia questions there first. Sometimes I get into another game, like crosswords, Connections, or Immaculate Grid, but this year, I have usually gone to another source: FanGraphs MLB Playoff Odds, to see how likely it is that the Toronto Blue Jays make the playoffs and win the AL East.
For those who don’t follow or care about baseball, the Blue Jays have clinched a playoff spot and are, as of Friday September 26th, 60.7% to win the division.1 That might sound good, but on September 16th, they were 98.26% to win the division and 100% to make the playoffs (though technically they had not clinched the playoffs, it was a mere formality). Dropping from 98.26% to 60.7% is bad, but if you’d like to be optimistic about it, the same model had them as 2.8% to win the division on May 27th, when they were 8 games back of the New York Yankees. Those swings are large, but they are not even the most extreme swings in baseball this year— The Detroit Tigers odds of winning the AL Central peaked at 99.9% are now 33.8%2, as the hottest team in baseball, the Cleveland Guardians, have overtaken them and currently lead the AL Central. The New York Mets were 96.2% to make the playoffs in June and fell all the way to 50.0% before recently rebounding to 78.3%.3 There are only 30 teams in MLB, and three of them were in such a safe position that they, in poker terms, needed a one-outer to lose their lead. If I were in a poker game where there were three one-outers hit in one night, I would never play in that game again. So what’s going on in baseball this year?
The first caveat I need to get out of the way, is none of these leads have been blown— yet. As I sit here on Friday, the Jays and the Mets are favourites to win their division and make the playoffs, respectively. If you got all-in with 22 on A28 rainbow vs. AK and the turn were an ace, your opponent’s odds increased from 2% to 16%, but this hand would only become notable if the river were an ace, king, or eight. Even the Tigers, who are now underdogs to win the division and might miss the playoffs entirely, have a 1/3 chance to win the division. Secondly, FanGraphs playoffs odds account for team quality. Before a game was played this season, the Yankees were 30.6% to win the division and the Blue Jays were 14.8%; if their team quality projections were off, their model was inaccurate from day one.4 Understanding shifting probabilities and Bayesian updating are important skills for a poker player and are increasingly becoming key life skills. Between elections, sports seasons, sports games, and a whole host of oddities on prediction markets like Polymarket and Kalshi; I am seeing more attempts to model the live probabilities of things. I think those models are often sound, but I think the way most people interact with the data presented in them are not.
The next time you play poker, the first card you will be dealt will be the Jack of spades. Thousands of people read this post, so this will happen to some of you (if it does, please post in the replies). It’s only a 1/52 chance, and cards dealt all around the world in different venues to all my different subscribers are more or less independent events. The odds of someone getting one-outed in a pot is rarer; you need an unlikely setup where someone plays a pot where their opponent has exactly one out with one card to come. However, if you play enough hands, you will see it happen (hopefully in your favour). Things like MLB Playoff odds or election odds are not independent events in the way getting dealt cards are. The day to day odds are correlated— being in first place on June 1st means your June 2nd odds will be high— these probabilities are subject to external shocks that poker hands are not subject to. In baseball, Fangraphs doesn’t account for future trades that might occur, even though it’s known good teams will try to improve via the trade market. In politics, this can mean a host of things—scandals, economic shifts— but one example I can use is last year’s Canadian election, where Justin Trudeau’s Liberal Party were, to use a poker term, drawing dead to beat Pierre Poilievre’s Progressive Conservative party, but Trudeau was replaced as a leader by Mark Carney, who handedly defeated the PCs. A model based on polling data for a Trudeau vs. Poilievre election that did not account for the fact that Trudeau might be ousted underrated the Liberals’ chance of winning. Within a poker hand, there is no equivalent to this; the rules or your cards never change mid hand.
However, I do not think the chaos in baseball this season was underestimated because FanGraphs didn’t have certain inputs in their model, nor do I think the chaos this season represents a one-in-a-million baseball season and we are witnessing something rarer than Halley’s Comet appearing on the same day as a solar eclipse. Last year, the Tigers were 1% to make the playoffs on July 30th and they ended up making it. In 2021, the Cardinals were 1.8% to make the playoffs and made it; the Atlanta Braves were 7% to make the playoffs and ended up not just making the playoffs, but winning the World Series. Therefore, when conceptualizing how unlikely a comeback, the question that one is trying to answer is, of the twelve playoff teams— which one’s odds will have the lowest nadir before eventually making it? In a model that did not account for team quality every year a team that started at 3.3% would end up winning the World Series and when gathering hundreds of datapoints in a given season, there will always be unusual results. Even teams that have relatively smooth paths to division titles hit low points. The LA Dodgers began the season at 83.8% to win the division; their season was not very dramatic— they reached 98.4% on July 2nd— but even they tumbled all the way down to 57.4% in late August. When you monitor these charts regularly, there will always be ups and downs that seem improbable as they’re happening, but at the end of the day are just noise. The New York Yankees were 91.8% to win the division on May 8th; that means by the end of this weekend, one of the Yankees or the Jays will have blown a 90%+ chance to win the division. These daily forecasts have been an entertaining way to begin my day and to measure the current probabilistic landscape of a specific event, but when you look backwards to cherry-pick peaks and valleys, you will almost always find some shocking outputs. Do not let monitoring probabilities trick you into thinking that an unlikely thing you witness is miraculous.5
To bring this back to the ostensible subject of this newsletter, poker tournaments, I see this same bias all the time when discussing any individual on a heater. In a five month span, I was twice down to one big blind at a final table and unbelievably won the whole tournament, but it or something like it was bound to happen to someone. The probability of a specific individual winning two WSOP bracelets in one summer or back-to-back tournaments or final tabling the Main Event twice are miniscule. The probability of anyone doing it are rather high. When you see someone on a heater that seems like a million to one shot, it’s good to remind that someone going on a run like this is bound to happen. There will always be individuals currently chosen by the gambling gods to be the luckiest over a certain period of time and their results will always border on being impossible— someone reading this newsletter will likely have a run like that, but pr not me or you.
Thank you for indulging me and letting my write about the minutia of MLB playoff odds. If you’d like more writing like this from me or just good old fashioned hand history analysis please consider becoming a paid or unpaid subscriber. All the subscription tiers are explained here and you can subscribe or upgrade your subscription by clicking the button below.
Additional Sims For Premium Subscribers
Premium subscribers get the raw files of sims I used to write my POTDs, sims that are more accurate and appropriate than equivalent sims in the big public libraries, videos of me walking through the sims, and a text summary of how I ran the sims. This week I uploaded
Rocket Solver flop and turn sims and a PIO river sim for POTD #131
Two PIO turns sims that forced me to use my actual size for POTD #132
Two PIO flop sims one that forced me to check for POTD #134
Two HRC sims that looked at the actual sizes used and a size test for POTD #135
Additional Analysis for Premium Subscribers
Everyday Premium Subscribers get an extra bit of analysis not included on Substack. Today I’ll share #onemorething from POTD #135 where I wrote about the A5 metagame in MTTs.
POTD #135 onemorething
Let’s talk about A5, the poker hand, not the highest grade of wagyu beef. The secret is out; you get to play aggressively with A5 preflop. Wheel aces always have okay equity against hands that aren’t AA— vs. KK you just need to make top pair or a wheel, vs. AK you just need to counterfeit, hit a five, or make a straight. A5 is the best wheel ace, because with the highest kicker it gets counterfeited less than any other one. In ring NLHE, A5suited is a common four-bet bluff-shove. Occasionally you get to play aggressively with the offsuit version— for instance, BU vs. SB ~30 deep, it’s a common four-bet bluff-shove. Sometimes you play it aggressively from the blinds and sometimes you don’t; it depends on the situation. Sometimes A5s three-bets a lot in ring, other times it doesn’t; it also depends on the situation.
However, what I want to talk about in this hand is why I think A5 is a pretty bad hand for me to be aggressive with here. I think most humans three-bet A5o far too often from the BB. It’s rare that humans are finding the exact mix of three betting A2o-A7o with some frequency, and they often shortcut to three-betting A5o too often. This means that if I am four-bet bluffing, one of the worst cards for me to have in my hand is a 5. I block my opponent’s most common three-bet bluff. This is especially true with A5 offsuit; literally any suited ace has better equity when all-in and called than A5o, and if I have a side card that is not a 5, it will have better blocker qualities because it won’t block the A5 three-bet bluffs from the big blind.
When I first started randomizing in live poker, I considered randomizing via suit (“hearts is the aggro suit today”), which worked so long as no one caught on and saw me playing crazy with hearts and adjusted. Well, in modern poker, a lot of people are picking a 5 instead of randomizing their kickers, which means the meta adjustment should be to never four-bet bluff with a five. If both of you are only bluffing with suited hands preflop, it should not be too big a concern, but once there is preflop aggression coming from offsuit hands, be wary.
Housekeeping
If you’d like to sign up for Octopi, Run It Once or GTO Lab you can get a discount using the following codes.
Run it Once use code: POTD for 10% off
Octopi Poker use code: PUNT for 50% off 1st month for monthly subs and 2 free months for annual subs
GTOLab use code: POTD for $25 off any product. It can be used multiple times.
There is a chat for subscribers here
Media
My child took the TV remote and hid it somewhere, which has made watching TV difficult (yes, I can still cast from my phone and watch stuff, don’t cry for me). I am probably going to watch One Battle After Another this upcoming week.
So I will recommend two things I haven’t seen and one thing I have seen. Hall of Fame Twitter poster Brooks Otterlake wrote a movie called Balestra that was released last year and is now available on streaming. He promoted the release of Balestra on streaming by sharing a video he made where he edited clips of Mr. Bean over the score of Koyaanisqatsi you can watch it here. I recommend, in descending order: Koyaanisqatsi, one of my all-time favourite documentaries; the Phillip Glass score for Koyaanisqatsi, which I listened to while writing this post; and then the works of Mr. Otterlake (the Mr. Bean edit, which he titled SUMĀṢAKA6, and Balestra).
As always I can be reached
Substack
Instagram
Twitter
Bluesky
64.8% as of the morning of 9/27
31.1% as of the morning of 9/27
and falling again to 44.6%
This can happen in poker. Isaac Haxton’s PLO skill might make him more likely to convert a chiplead into a Triton tournament win (more on this Monday), but it does not make him more likely to win an all-in vs. Nacho Barbero or Jesse Lonis.
To be fair to Cleveland sports fans this Guardians run is bordering on being miraculous.
https://sanskritdictionary.com/sum%C4%81%E1%B9%A3aka/266940/1
Amusingly, the first hand I looked at (I tend to read these while playing with a very light table-count) after reading your prediction was indeed JsTd with Js being the first card.