> What's interesting about this is the minimum defense frequency is based on what the strongest hands you can possibly have in that situation are and the opponents possible hands do not even factor into it.
This is actually not quite true. MDF is purely a formula based on pot size and the bet size (pot size / pot size + bet size). The fact that it doesn't consider various ranges is why it's not really useful - it was a simplified formula used to try to understand the game before solvers existed.
There are situations where your opponent can bet any two cards profitably and you do have to fold - imagine they bet the size of the pot, but have the better hand 99% of the time, you're simply forced to let them bluff the 1% of the time they're bluffing. MDF is a pre-solver concept and not an especially useful concept in the modern game.
I'm pretty sure mdf applies to rivers when you are last to act. I'd be interested in being proven wrong however if you have solver output that shows it. I remember studying solver output and seeing it in action.
I know that before the river there are range advantages that make defending mdf a losing play.
What's true is that equilibrium strategies typically converge to solutions where the better makes the caller indifferent between calling and folding. In the toy example I've given where the betters range is so strong, the caller should always fold, the better now has an incentive to add more bluffs to the range to take advantage of the folds. Then the caller will want to call more. This might converge to the MDF which might be what you're suggesting, assuming we started with ranges that could have enough bluffs given the runouts.
If you open up the solver, and give one player only Ace-Ace as their starting range, and the other player a pair of twos, and the board Ace-Ace-Three-Three-Three, then the pair of twos will fold 100% on river and will not call at MDF.
I think another way to say this is that MDF works only if you're in a spot where you have hands that are strong enough to call. If you play every hand, and you see every river in that 100into100 situation, you shouldn't call with 50% of your hands because your hand range is too wide for that to be profitable.
So you can't make a ton of mistakes say "MDF" and call off, you have to have done the right things in previous streets to end up with a range that can call at MDF. That range (and those street actions) require an understanding of GTO (and the adjustments needed when someone isn't playing GTO).
Within solvers, you can do something called "node locking", which means you "lock" a tree in the game node to play a fixed strategy. You would typically lock it to play as you suspect your opponent plays. This lets the solver calculate the optimal exploitative solution against your specific oppoents.
Piosolver, the first public solver and the one mentioned in the article, has this feature.
However, what often happens is if you lock one node, then several other nodes in the game tree over-adjust in drastic ways, forcing you to lock all of the, which may be infeasiable. As a result, Piosolver recently introduced "incentives", which gives a player in the game an additional incentive to take a certain action . For example, you may suspect your opponent calls too much and doesn't raise enough, so you can just set that incentive and it will include that in its math equations and give you something similar to an exploitative solution with a much simpler UX.
This feature was literally just introduced a few months ago so it's still very much an active area of research, both for game theory nerds, and people trying to use the game theory nerd research to make money !
I want to see strong AI used in video games, especially strategy games. People often retort that strong AI is not fun; it's too challenging and that's not what players want. But once we have a strong AI we can adjust its goal function in fun ways. What you're describing is effectively the same, and it's the first time I've seen this used in a strong AI.
I do personally dislike that GTO became the nomenclature , as I prefer "theory-based", since it causes this confusion, but trying to fight it at this point is hopeless because GTO is the search term people are using. And when people say they "play GTO" they usually mean "equilibrium" rather than "optimal against my specific opponents" which is "exploitative".
If you actually watch what the top players advocate for, everyone suggests you want to play exploitatively. However, there's one equilibrium solution and effectively infinite exploitative solutions, so equilibrium is a reasonble starting point to develop a baseline understanding of the mechanics of the game. It's tough to know how much "too much" bluffing is unless you know a baseline.
Furthermore, if you "exploit" people by definition you are opening yourself up to being exploited so you need to be very careful your assumptions are true.
Also, with solvers like piosolver, you can "node lock" (tell a node in the game tree to play like your opponent, rather than an equilibrium way plays), but there's many pitfalls, such as the solver adjusting in very unnatural ways on other nodes to adjust, and it being impractical to "lock" a strategy every node in the tree. There's new ideas called "incentives" which gives the solver an "incentive" to play more like a human would (e.g. calling too much) but these are new ideas still being actively explored.
Rock paper scissors is frequently used to explain GTO but it's not the best example because equilibrium in rock paper scissors will break even against all opponents, but equilibrium poker strategy will actually beat most human poker players, albeit not as much as a maximally exploitative one.
There's two other huge pieces this article glosses over:
1) It's as impossible for a human to play like a computer in poker as in chess - in fact far more impossible, because in poker you need to implement mixed strategies. In chess there's usually a best move, but in poker the optimal solution often involves doing something 30% of the time and something else 70% of the time. The problem is that, not only are there too many situations to memorize all the solutions, but actually implementing the correct frequencies is impossible for a human. Some players like to use "randomizers" like dice at the table, or looking at a clock, but I find that somewhat silly since it still so unlikely you are anywhere near equilbrium.
2) Reading someone's "tells" live is still a thing. While solvers have led to online poker to decline due to widespread "real time assistance", live poker is booming (the 2024 World Series of Poker Main Event just broke the record yet again) , and in person in live poker, people still give off various information about their hand via body language. From the 70s to the early 2000s, people were somewhat obsessed with "tells" as a way to win at poker. Since computers have advanced so much, it's fallen out of favor, but the truth is, both are useful. It's totally mistaken to think that advancement in poker AI , GTO , and solvers have rendered live reads obsolete. In fact, in 2023, Tom Dwan won the biggest pot in televised poker history (3.1 million) and credited a live read to his decision, in a spot where the solver would randomize between a call and a fold.
Yes indeed glad you noticed! Been too addicted to that godforsaken game at points so figured borrow some of its qualities for my studying apps...in general I'm interested in gamification + studying.
The technique of journaling as you work is sometimes called “interstitial journaling” and I became a big fan of it as a way to help focus as well as keeping track of what I was working on.
I made a tool to associate those notes with a color coded project and timestamp:
It ended up being unintentionally similar to an invoicing time tracking tool a freelancer might use but the use case Im interested in is more personal productivity.
Can’t say the project has generated much interest outside my own personal use but I find it very nice to track notes as I go and then easily see how much time I spent on a given project. You can also add a #hashtag in the notes and then filter by that hashtag in the calendar view.
Completely free in case anyone else finds it helpful!
People underestimate how much cultural baggage influences things.
I'll give a very simple example. I did a few SWE interviews in 2020, and several companies did the initial screen over the phone, and the on-site over Zoom.
In both cases it was a remote interview. There was no reason not to do both over Zoom. The only reason was that the previous process was a phone interview and then an in-person onsite, and they realized they had to replace the in-person on-site with Zoom, but they didn't think to replace the phone screen. If you started from scratch it makes no sense though.
In this case, the whole origin of the Leetcode interview is "we're going to hire the smartest people in the world.". You can dispute whether that was true back in 2009 but it was certainly part of Google / Facebook's messaging. Now, in 2024, I think it has morphed much closer to a standardized test, and even if people might begrudgingly admit that, there's still the cultural baggage remaining. If a company used a third-party service, they'd be admitted they're hiring standardized candidates rather than the smartest people in the world. Which might be an "unknown known" - things that everybody knows but nobody is allowed to admit.
I definitely agree that this industry, for all of its self-proclaimed freethinking and innovation, is rife with cultural baggage. Allowing for an independent standardized interview step would defy the not invented here syndrome that many leading corporations ascribe to, that their process is best. Not to mention reducing friction for applicants (by don't repeating your Leetcode stage) is inimical to employee retention incentives, that is preventing them from shopping around for new employers. So me saying that we oughta have a standardized test to save everybody's time is more wishful thinking than anything.
This is definitely a factor. "You don't understand, we have a really high bar and we only hire the best people" is a bit of a meme in recruiting circles because you will never ever ever ever not hear it on a call.
I don't think we found it a barrier to getting adoption from companies though - perhaps because "we're a really advanced company using this state of the art YC-backed assessment" satisfies that psychological need? Unclear.
> but it was certainly part of Google / Facebook's messaging.
It entered the online cultural zeitgeist before that, with Microsoft talking about their interview processes, and indeed early interview books were written targeting the MSFT hiring process that many other companies copied afterwards.
I graduated college in 2006 and some companies still did traditional interviews for software engineers (all soft skills, and personality tests, no real focus on technology, except maybe some buzzword questions), and then you had the insane all day interview loops from MSFT and Amazon.
Back then, Google famously only hired PhDs and people from Ivy Leagues, so us plebs didn't even bother to apply. In comparison, Microsoft would give almost everyone who applied from a CS program at least a call back and a phone screen.
What’s ironic is that Michael Seibel has discussed many times on the YC podcast that you should avoid building whatever’s hot for VCs because their attention tends to change every year but you’ll be stuck building for a decade.
2020 was remote work, 2021 was web3, now we have the big LLM boom.
Honestly it seems there’s a lot of advantages to “riding a wave” and a lot of advantages to being contrarian. But if raising money is your priority I do think you should ride the wave. Being contrarian sounds romantic, but don’t expect funding from people who disagree with you.
The most charitable thing I’d say about YCs AI focus is it’s hard to think of a startup idea that couldn’t benefit from AI in some way.
Honestly, I think any person or organization that bought in to the blockchain hype should be barred from making any financial decisions of consequence. If they bought into a scam as obvious as the blockchain crap, they're clearly not capable of holding any real responsibilities.
Tbf I feel remote work really improved significantly in the past years, though I don't know if the contribution from those startups matters or not. Web3 and blockchain is a moot, there's little to no practical reason to have them.
AI though, will be very useful, at least a good one. Theoretically, AI can swim in a good ocean of company documentations and save time searching. They can help doctors diagnose a ct scan faster (if not already).
So it's funny because I _set out_ to build productivity software for lots of people, I just completely failed at that goal, probably in part because I did build what I wanted, which is not what the mass market wants. Plus, design is hard and very important in the space. But I completely succeeded at building software that made my life better, at least.
I know HN sometimes dislikes relenteless self-promotion which this comment will be so I wanted to give a shoutout to two productivity apps I have no affiliation with - Inbox When Ready which lets you search Gmail without being distracted by new emails, and Unhook which lets you search Youtube without seeing recommendations, both of which I find useful for avoiding distractions.
So I made an ultimate habit tracker personalized for my desires. But I feature creeped it to death with features I want and nobody else wants. So one of the "habits" I optimize for is getting enough focused work done on any given day, which I use a time tracker for. And I like to take notes on my work as I do it so I can focus. This is called "interstitial journaling". So I refactored just the time tracking and journaling app into a new app called Interstitch
This app _completely_ flopped on Product Hunt and Reddit as well, despite being free, for a few reasons but one is that people think "interstitial journaling" is more about journaling when its' really more time tracking.
However, it was a double-edged sword because the Reddit post I got 2 upvotes on is now the #2 Google Search result for "interstitial journaling app" which gets me about 5 new users every week. (useful lesson to be learned that sometimes the best SEO is not on your own website).
Most time tracking apps really focus on the invoicing or employee time market, understandably since it's more B2B, but I prefer my app because it's more focused on personal productivity. If you want to lose weight, track what you eat, if you want to focus more on work, track how much time you spend focusing, it keeps you honest and unveils trends. I'm also a big believer it's a helpful tool for ADHD people because "time boxing" / "calendar blocking" is very prescriptive, whereas journaling what you actually did lets you be more flexible with your specific task but still keeps you honest about your overall productivity.
Specifically, my gripe with habit trackers was:
1) They are usually optimized for tracking just a few things. But I have a _lot_ of habits I want to track - health habits, work habits, hell, giving my dog a monthly bath habit. It's not even about perfectly doing everything, it's about gathering the data so that you can view insights and understand yourself. This is also called "life logging" or "the quantified life".
2) They were too prescriptive instead of descriptive. So in my habit tracker, you can just set a goal to say "exercise every day", then if I track that I specifically lifted weights, it "bubbles up" to say that I exercised, since weight lifting is a child of exercise in a big DAG. In this big DAG, I have a bunch of "views" of different sub-trees, so my dog has his own sub-tree that reminds me to get him baths and bring him to the vet, but that's not in my daily core view.
3) I want something with both a good web and mobile experience, though there's a few other apps that do this such as Everyday, none of them also do 1) and 2). My real dream is also super slick Apple Watch integration which I started but it was too much work to build SwiftUI for mobile and Typescript for web and integrate them nicely.
So the app is called Navigoals ( https://navigoals.com ) . For reasons I still can't explain, YC actually gave me an interview to pitch it but the pitch went about as terribly as it possibly could have. Because of aforementioned feature creep, I closed signups and if I build on anything it would just be Interstitch which is simpler and more polished but Navigoals does have a Youtube demo on its landing page.
I want to say that people love to dunk on programmers building another TODO-list apps or habit tracker. I do think it's going to be a bad business strategy the vast majority of the time unless you have top notch design and marketing skills. But, I think it's a fantastic way to do things like learn a new tech stack or design stack and the end result is a tool that's optimized for your own use case. Productivity tends to be a very personalized thing without good "one size fits all" solutions , even things like Notion which try to be the kitchen sink don't have basic things like real features you need for a great habit tracker, so I highly encourage people to ignore the naysayers and build the best tool for themselves.
I can say personally that Navigoals and Interstitch both forced me to be a lot more honest about some bad habits I had and how I spent my time. For example, I was in huge denial about the impact of marijuana use on my productivity but seeing the days I used it and the clear decline in other health and focus metrics in the week I used it made me realize I had to quit partaking in it if I wanted to achieve my other goals.
That’s a fair perspective but consider that spaced-repetition can be an “MVP” of the broader concept of adaptive / personalized learning. You track some state about your knowledge of the world, in this case your performance on a flash card, and optimize the next piece of educational content around it.
It feels obvious to me that, in theory, we could do a lot more to leverage your current world knowledge to recommend the next piece of educational content, to optimize not just your understanding but also other things like how engaged you are. For example, a system could throw you an easier question if it helps you focus longer.
When I see the state of current interview prep it’s mostly “here’s a big list of questions” , perhaps tagged by difficulty and grouped under related topics such as “graph problems” or “tree problems” but I’m personally convinced a more sophisticated system could serve you the best possible next problem to stretch your brain in the right way.
Spaced rep is simply the proven starting point. The rest of adaptive learning has a bit of a troubled past because it was trendy to VCs , but pushed on educators and presented more as an alternative to a teacher rather than something to augment a teacher. I worked for a company that raised 100M+ to work on it but the CEO was great at terms sheets but uninterested in actually building a great education project.
But the reason that I joined the company was the high level idea still resonates with me. Surely many HN users have a big list of things they want to learn- perhaps about LLMs. But a typical course will have zero knowledge on where you’re starting as a student . It might bore you with stuff you know already or take for granted key prerequisites and skip them.
Anyway, I’m currently building my own adaptive learning platform but focused on helping professional poker players learn game theory . My idea being that it’s an easier ed tech app to bootstrap as the knowledge very directly translates to money. And really the same criticism applies because you can’t truly memorize a game tree , it’s more important that you build a high level conceptual understanding. But , as much as rote memorization deserves to be maligned when done in isolation, it’s not so bad when done as part of a broader learning strategy. For example, you can’t memorize vocab to learn Spanish but certainly knowing 5000 words in Spanish is a very nice starting point compared to not knowing any words. And tools like spaced rep have been proven by research to help with that goal and I view as a pathway to more broad adaptive learning strategies.
Ironic you post this because I just rediscovered a 20yo blog post that I remembered reading, re-read it again and found it very motivating and inspiring and submitted it.
reply