Episode 111: Nate and Andrew Talk About Some Hands

16 thoughts on “Episode 111: Nate and Andrew Talk About Some Hands”

1. Im sorry for your loss Andrew.

   • Thanks.

2. It’s pretty cool that you were able to get your grandma on the podcast which will live on forever. I imagine there will be a fresh crop of fish who will find this show next year like I did and get to meet her just like the rest of us.

3. My condolences Andrew to you and your family.

   • Thanks much.

4. I ran my CFRM code on hand 1 using these ranges

   SB: 22+, AJo+, ATs+, KJo+, KTs+, QJo, QJs-65s, A5s-A2s

   CO: 55-JJ, AQs-ATs, AQo-AJo, KQs-KTs, KQo-KJo, QJo, QJs-76s

   I’d be happy to try again with different ranges. Personally, I prefer to play 3bet or fold from SB, but I’m a fish, so whatever.

   I allowed 60% pot sized bets only, which leads to a problem that I’ll come to below.

   SB has 52% equity on the 8s7s7c board, and an EV of 9.7bb compared to CO’s 9.3bb (yes that’s 2bb to rake!).

   SB checks 44% and bets 56% of his range. Most of the strategies are mixed. Hands that tend to bet include T9s, 65s (no club or diamond) JTs, QJs (only with a club or diamond), QdJx, overcards with a diamond, Acxc, ATs (no diamond), some nut flush draws, 44-66 (with or without diamonds, a bit surprisingly), 88-KK and AA (no diamond).Checking is mainly to check/fold or check/call, with a small check/jamming range (mainly flush draw + overcards and some overpairs).

   AcQc is check/call 71%, bet 29%.

   Here’s the problem. My game tree is set up so that if a player has less than the smallest allowed bet behind after he bets and gets called, he just jams.

   If SB bets, CO calls 60% and jams 6%. Let’s say he calls.

   On a Qh turn, SB checks 92%, jams 8%. AcQc is checked 100%.

   After a check, CO checks back 86% and jams 14%. This is the problem with the bet sizes I’ve allowed. After the flop goes bet/call, CO doesn’t have enough behind to do anything other than jam, according to the way I’ve set things up, so I can’t answer questions about SB’s check/raising range.

   If CO jams, SB calls with 29% of his range, top pair or better, including AcQc, and some flush draws, mainly nut ones, which are also ahead of some of CO’s bluffs.

   I’m going to redo the calculation with half pot bet sizes, which allows for bet/call, check/bet/jam. I’ll probably post next week.

   • Obviously meant 8d7d7c flop. Sorry for the typo.

   • Very interesting, thanks for taking the time to write all this up. One of the most interesting things that I picked up from PokerSnowie, and that I see here as well, is the importance of accounting for backdoor flush draws. It makes sense, especially with deeper money, that your ranges need to be resilient to a club-club runout on a board like this. Do I understand correctly that SB is mostly betting 88 on the flop? What about 77/87s/76s?

     • Bets 88 78%, 77 7%, 87s 87%, 76s 97%, check calls the rest.

       Seems reasonable mainly to slowplay quads.

   • I’ll add my thanks as well.

     I’m really inexperienced at building ranges. What sticks out to me is the notion a particular hand isn’t simply in one’s betting or calling range. It is often in both, in some ratio.

     These approximations John provides are very helpful.

     • When computers are constructing ranges, that’s true. Of course they are much better at managing precise details like that. There are times when mixing up how you play certain hands is truly important, but I suspect that in many cases the EV difference between betting AQ with the Ac 87% of the time vs 100% of the time is extremely small, as the likely result is just that you are exploitable for a small fraction of the pot by large bluffs when the turn and river run out clubs and you have slightly fewer nut combos than you theoretically should.

       • That’s true, but there are plenty of combos for which the equilibrium strategy is close to 50/50, and also where it’s 0/100 or 100/0. For example, in this hand, an interesting one is 65s, for which 6h5h and 6s5s are 100% bet/fold, 6d5d is 100% check, mainly to check/call, and 6c5c is 64% check/call and 36% bet/fold. Another one is KdTd, which is 45% check/raise and 55% bet/call, or KQo with one diamond, which is 35% check/fold, 35% check/call and 30% bet/fold.

         Working out what is the useful information that you can extract from these solutions to use in real play is very tricky. Alex Sutherland has a very interesting video about this at http://blog.gtorangebuilder.com/2014/11/improving-your-turn-play-with-gto-gto.html . His solver is much more flexible than mine so he can fiddle about with the decision tree and see what is and isn’t important, although he’s only doing this from the Turn for now.

         What always amazes me when I look at these GTO solutions is, at the risk of stating the obvious, however the board runs out, booom, both players get there with balanced bluffing and calling ranges. That’s 1176 possible runouts. I can’t imagine any human brain could manage that, so anything that let’s you get even close is probably good.

         I’m also always reminded of Harrington on Cash Games, which I didn’t much like, and Andrew’s review of which I recently read. Harrington expounds his ‘looking at your watch, seeing where the second hand is, thinking, I want to check/call 50% of the time, watch says check’ approach, which always made me laugh but, well, that’s what the GTO solution does…. :^}

   • I reran my code with half pot sized bets allowed. There are some differences in the solution.

     1) The EV of each player changes a little. SB’s EV goes down to about 9.45bb and CO’s increases to 9.4bb, with a little more going to rake.

     2) SB bets less often on the flop, down to 46%. There are some small changes to the details of the hands that bet, but not huge changes, with some exceptions….

     3) AcQc now becomes a half pot bet 84%, check/call 16%. I have no idea why this hand in particular should be so affected by the change in bet size. If the flop goes bet/call and the turn is Qh, AcQc is now 45% bet/call, 27.5% check/call and 27.5% check/jam (CO calls with top pair or better). If SB checks, CO checks behind with 55% of his range, including some fraction of full houses, but no trips, some top pairs, some second pairs and some flush draws.

     After the turn goes check/check…

     If the river is 3d, AcQc is mainly bet/call, which seems surprising, although he only has 23% of his AQ combos left by this point. The best hand he folds is QJ.

     If the river is 3s, AcQc is 2/3 bet/call and 1/3 jam (CO calls all the way down to half his remaining combos of TT). Clearly AQ is the nuts now. :^}

     If the river is Ks, AcQc is a bet/call 1/3, bet/fold 2/3, in other words, it’s on the borderline. SB calls with top pair.

     I don’t claim to understand what’s going on here. I’ve uploaded the new solution file if you want to have a look for yourself. It seems to me there’s enough information in this solution to keep oneself occupied for a significant amount of time, possibly without ever improving one’s game at all.

     I’m toying with the idea of running it one last time with half pot and full pot bets allowed so that you can get some insight into how many streets you should use to get the money in (i.e. pot/pot or half/half/jam).

     • I now have a solution with both half pot and pot sized bets allowed and have uploaded it so you can view it if you want to. My gui is particularly awful when there are many betting options, so good luck with that.

       I won’t go into much detail on the overall strategy (EVs don’t change much from the half pot only solution), but just say what the strategy is with AcQc.

       On the flop, it becomes 41% check/call a half or pot sized bet, although CO never makes a pot sized bet in the GTO strategy, and half pot bet/call the rest of the time. (Interestingly, SB only pots it with 8% of his range, including some combos of overcards with a diamond, flush draws and overpairs).

       If the flop goes bet/ call and the turn is Qh, AcQc is 20% check/call and 34% check/jam after CO’s half pot raise, and 46% half pot bet/call, which isn’t much different to the half pot only solution.

       The option of betting pot doesn’t seem to make much difference to the solution in this situation.

       That’s all I have for now. :^)

   • @John, regarding the ranges, you might want to get rid of some of the smallest pairs and “offsuit trash”, and add more suited aces. The post-flop playability and the “blocker value” of the ace in Axs tends to give those hands more EV both as light 3-bets and calls.
     As noted above, backdoor flush draws have an interesting impact on equity and barreling frequencies, but it’s hard to have a worthwhile backdoor flush draw with 22 or QJo. 😉

     You also wrote “at the risk of stating the obvious, however the board runs out, booom, both players get there with balanced bluffing and calling ranges” which reminds me that when I’ve examined the equities of ranges in Snowie, it’s pretty amazing how often the bettor and the caller have close to 50% each post-flop on most board textures. There are occasional spots where one player “accidentally” gets a clear range advantage due to board texture and ‘weird’ runouts, but most of the time a solid pre-flop range will flop enough equity to bet or call at a decent frequency on every street. This leads to the slightly unnerving (albeit slightly incorrect) observation that “GTO bots are calling stations”. Their ranges are just so strong that they *can* call down more effectively than unbalanced players.

     • You’re probably right about the suited Aces. It was hard to figure out from the information in the podcast what range SB was using, so it’s kind of an unsatisfactory average between what I thought he was doing and what I’d do.

       The GTO solution seems to get coverage on every board by calling with almost everything with any equity at least some of the time, which appears to allow it to call a lot more often than a real player would, although I have no experience of how these things go at high stakes.

       Of course, none of this tells us anything obvious about how to construct preflop ranges.

Comments are closed.