Week in Review #9 May 18th-May 24th

One of the key tenets of POTD is that solvers are tools and a tool is only useful if it is used well. A hammer is a tool, you wouldn’t use it to caulk a bathtub. Artificial intelligence, artificial general intelligence and large language models are a hot topic of conversation nowadays: I was at a baby shower last week and someone jokingly(?) commented that the baby would be born in the last cohort born before AGI. Those in poker have been aware of the power of AI for quite some time, but solvers like PIO are not LLMs— poker solvers, solve a poker hand through self play until it reaches a mathematically unexploitable equilibrium. LLMs predict a likely response based on its training data, poker solvers spit out a complete answer for a given set of inputs. During EPT Barcelona 2024 there was chatter at the table about if ChatGPT would know all-in equities; my hypothesis, which I tested, was: if you asked an LLM about a preflop all-in, it would give you probabilities of the most frequently discussed preflop all-in QQ vs. AK. I expected no matter what hands I asked about, an LLM would answer 55%/45%. I asked what the preflop equities of J2s would be all-in vs 66 and I got the wrong answer that I predicted I would get, 55% vs. 45%. I soul read the computer, unfortunately, I didn’t soul read the players at my table and bet on the outcome of my prompt, whoops.

One notable subscriber of POTD, Nate Silver, who I first knew as Nate Tha Great on TwoPlusTwo, then as the Baseball Prospectus guy, then the election forecaster and now all those things and a very popular and entertaining Substacker wrote a long and interesting post about what happens when you try to get an LLM to play poker. You should check it out (and not just because he name checks POTD). LLMs are powerful tools, but for now, don’t use them to play poker; they don’t even know the rules.¹

Shortly after posting Week in Review #8 this Substack crossed an important milestone and now has over 1,000 subscribers. I’d like to thank you all for the support and if you are one of the hundreds of unpaid subscribers, I’d suggest you upgrade and become a paid subscriber. The subscription tiers and all the ways to contact me are listed here.

Testimonials

A lot of very glowing things were said about me on Twitter this week. I am blushing. Thank you for the kind words.

Dr. Kamikaze

This is possibly the best hand history analysis I've ever read, both for its informational value, but also because

@SamGreenwoodRIO's writing is so witty and concise. I laughed, I learned things, and I hope you will too!

OG 2+2er Jonas Entin riverboatking compared POTD to the works GOAT poker content maker and head honcho of Run It Once, Phil Galfond here

your POTD are absolutely fantastic you have a very @PhilGalfond like skill of explaining complex concepts in a very digestible way n also make it fun n engaging.

Wesley Nysm recommend that people do something crazy and give me money

One of the best things I have [come] across in the poker education world - is @SamGreenwoodRIO ‘s substack. So much value, and absolute GEMS every post. Guarantee you on a [ROI] standpoint you will make back that $20/month. Thank me later.

Additional Sims For Premium Subscribers

This week at the request of premium subscribers we have added a new feature and I’m happy to announce that POTD is now a multi-media operation (multi-media empire? tbd). In addition to posting the files of sims that I am recording short videos so that people without the software I am using or people who consume poker content mostly on mobile or people who just prefer video analysis to text analysis can have me walk them through some sims. Below I’ve posted a sample from Tuesday’s free post: POTD #42. I also shared:

• A Rocket Solver sim for the flop and a PIO sim solved from the turn

• Four different HRC sims to show when Kristen can fold after putting in most of her stack

• A PIO test to see how many combos of JTo the SB has before they can x/f

• PIO postflop ICM sims for the hands posted Thursday and Friday

Additional Analysis for Premium Subscribers

Everyday Premium Subscribers get additional analysis, here is a sample from POTD #41

POTD #41 ⁠onemorething In my blog I write about how there isn't a big difference between betting 1bb and 2bb on the river. However a common mistake I write about is putting too much money in the pot. This hand presents an interesting framework about how to think about putting too much money in the pot. When you bet 2bb instead of 1bb you are putting in twice as much money as you’re supposed to, however it can never lose as much EV as incorrectly betting 200% pot all-in instead of 100%. When you bet 1bb instead of 2bb in the absolute worst case scenario, the extra 1bb is immediately lit on fire; you are losing 1bb in a 15bb pot. Even if you lose 10% of that bet it’s only a 0.1 bb mistake. If I shoved 30bb into 15 and lost 10% of my bet it’s a 3bb mistake.

Media

I was on Connor Richards Life Outside Poker podcast. Despite the title we talked a lot about poker and had a fun and wide ranging conversation. You should watch or listen on your podcast app of choice.

I also did commentary alongside Nick Walsh for the SCOOP Event #62-H. You can watch the archived Twitch stream here

The only media I have to recommend is The Studio, which I initially found very annoying, but slowly won me over over the course of the season. I still find many of the inside baseball elements of the show to be annoying, but it is funny. Lots of things irritate me, but few things make me consistently laugh. So I will recommend it here.

Substack
Instagram
Twitter
Bluesky

1. This is also a reason I have not used AI to copy edit my blogs. Basic spell check hates that I write “check folded” instead of “checked and folded”. Everyday I am reminded to write “the top pair” or “two pairs”. I suspect LLMs would make similar copy editing suggestions and the LLMs would not save me time.