I did not run the hand between Berkey and Adelstein through pio, but i think what Paul missed in his amazing analysis is the effective stack size. I would imagine if this were 100bb effective instead of close to 250bb effective, the SPR will subsequently be very different and hence the the tree. As we all probably know, if we get deeper, one pair hands go way down in value, and the potential to make a good hand goes way up in value like PLO, which explains the merge betting and the OOP choosing to call hands that can improve on rivers rather than hands currently ahead in absolute hand strength. If this were 100BB effective, I would imagine many OOP overpairs to check shove turn for equity denial, which leads to IP betting way fewer combos of hands like Q10 which really wants to realize its equity; since the SPR is still relatively large in the actual situation, these hands are not worried about OOP check shoving and losing their equity.
Killer comment. A recent debut video by Patrick Cronin is highly relevant to what we're talking about here, since Patrick goes over multiple 3bet spots where he basically demonstrates the difference between being 100bb and being deeper and how that effects optimal strategies. Check it out
With Garrett raising to only 2.5x preflop and no rake I'd imagine that Robl would be getting in there with this 76o combos. He'd likely be check raising these combos much less often which would then give him ton of straights on the turn even though the suited 76 is heavily discounted.
Pretty sure what’s going on with KJ and AJ checking back turn in Berkey/gman hand has to do with shared cards and the fact AK is floating 100% on flop but folding turn. KJ and AJ can’t bet three streets and are incentivized to check not only for bluff catching rivers but because when our opponent rivers Top pair, we make two pair. Q10 also bets at this stack depth at high frequency on turn because it will only ever get to bet large on river on a 10 or 9, so its more incentivized to merge bet for protection since it struggles on almost all rivers. Plus by folding out oop’s AK on turn, it will be able to more comfortably bet Q rivers.
I like this video. I like to know what kind of caliber poker player i'm watching, but i've never heard of most these guys. They all live cash pros?
Also I am left a bit confused at the 2463 board, where you said it makes sense for Young to lead his entire range for small on the turn. Why does it not make sense to have a bigger betting range where we can put more money in with sets and flushes?
I think it's because going small with his entire range allows him to make thinner protection value bets before polarizing his range otr. Similarly, IP is still uncapped after the flop check with a well-played mixed strategy. This tends to lend itself to smaller bets, even though this is a great card for OOP overall.
I really like the idea of these kinds of videos and I enjoyed the content.
However, you need to allow for multiple bet sizes, especially smaller sizes for the OOP 3better. Otherwise you risk the good old, garbage in = garbage out.
The first hand between Berkey and Garret for example, you say that you were suprised to see that Pio wants to check almost the entire range. However, Pio actually wants to use an overall mixed strategy of close to 40% cbetting and 60% checking, when given the option of betting 25%, 33% and 50% on the flop (and Pio unsurprisingly prefers the 25% bet size).
These results are based on the ranges that you show in your video and allowing for multiple betsizes across all streets for both players, including overbetting the turn and river (I used 50%, 67%, 100%, 150% and 200% plus all in for turn and river).
So just be wary that you risk drawing false conclusions, when you limit the game tree too much and don't allow for the most natural cbetting size.
I hope this doesn't come off too negatively - Keep it up :)
Hey ST, nice comment. And nw, definitely didn't come off negatively.
There's no argument against the fact that more bet sizings on each street will create an output that's closer to "perfect poker". I'm curious though: are you giving OOP multiple flop sizings because you're hoping to implement multiple flop sizings yourself? Or are you looking to see which sizing is most commonly used? If it's the latter, and you're presumably looking to play a strategy with two options (bet/check) at this node, then you'd need to re-sim it with just that sizing anyway for a better output.
There are some people who want to play multiple sizings on every street. Personally, I mostly disagree with this approach, at least on the flop. My goal, generally, is to play as simple of a strategy as possible while retaining a high % of EV compared to a more-complicated equilibrium. There's no way I'm capable of playing a perfect strategy with all its complexities. Similarly, I've yet to see a flop spot where having more than 2 strategic options at the cb node increases EV by substantial amount -- but I guess that's subjective.
I don't think you were advocating for a complex strategy as mentioned in the above paragraph, but that was just a side tangent for anyone interested. I think I gave OOP the option to bet 33% or check. If the perfect single sizing here is 25%, I'm not making a big mistake.
With all that being said, you're right and I should have included more sizings for turn+river. Multiple flop sizings make the sim too big and results too convoluted imo. As long as the flop sizing I gave is reasonable and the ranges are okay (which is the biggest factor imo), I'm fine with using the sim to analyze the hand in a broad strokes way.
ST, how big was EV difference between those 2 strategies (yours and Paul's), obviously we can improve our EV results by running simulation after simulation, tweaking it and getting better results. But does EV difference really worth it?
Personally, I I give Pio the option of multiple bet sizes in order to figure out which one is the prefered one/the one with the highest EV.
Like you, I'm not really trying to employ a stratigy with multiple bet sizes myself but rather simplify and going with one size, obv depening on the preflop action and board texture.
You are right that in some situations it would then make sense to do a resim with just that one bet size. However, often/some times it is not really neessary as Pio will have the other bettting options approximate 0% usage and thus those parts of the game tree will be almost empty anyway. Obviously, this depends on how long you run the sim for.
In the hand between Berkey and Garret, Pio rather quickly had the 33% and 50% bet size options that I gave it, drop below 1%.
In the sim for the hand that you show in the video, you had only given the option of a half pot C-bet, which I thought seemed off and would produce vastly different results than when/when also given the option of a small bet size, which I mostly see and use in this situation.
Now you could argue that I'm also limiting the game tree by not giving Pio even more bet sizes to choose from, and that would be correct. In fact, it is very likely that Pio would find an even smaller bet size to be higher EV in this spot.
In generel though, I try not to assume to know what the right betsize(s) is(are) and thus give as many options as my 64 gigs of RAM allow for.
Jeff_
I didn't run the sim for for that long. I had it run until exploiatable for 0.11% of the pot though. This produced an EV for OOP of $6,122 whereas Paul's sim produced an EV of $6,079.
So you are right that the difference in this case is negligible, in which case it's definetily easier to employ a pure strategy.
Ah, for some reason I thought I gave OOP the option to b33, not b50. I can see how that would affect outputs more than I thought initially in terms of betting frequency. Luckily, the EV differences are small. Thanks for all your analysis, and I think we're mostly on the same page philosophically.
The first hand is interesting, especially the turn spot where IP is supposed to merge quite heavily with a big size on the turn. My guess would be that it's really sweet to get protection from AK/AQ/KQ, and by betting 33%, we're letting villain realise too much equity. Also, IP has so many good hands in his range that maybe it's simply too costly to construct a range that bets with a small size, despite the fact that maybe some not so clear valuebets would prefer that in a vacuum but can't do it because then you'd also have to bet small with some really strong hands in order to be properly balanced? Here I'm assuming that very strong hands are much better off betting big, which would make sense?
Cool stuff, cheers. I liked the video despite having 0 hands with Hollywood Haxton.
Don’t fall into the idea that bets as part of an equilibrium strategy are always classifiable as value or bluffs or anything else. Some combos can e.g. call , fold or raise for equal EV in a particular configuration. Action boards lead to such complex equilibrium with sometimes difficult to classify actions. Often they can be seen to serve a purpose later in the game tree. In any case EV for the range is the optimized parameter. All plays serve just that purpose.
Yeah, you're totally right. It's difficult to interpret/implement strategies when we can't classify bets into a category to create some intuitive understanding, hence the attempt to do so.
really gross spot here for Garrett and a wicked raise by Robl here on the turn, Andrews actually got a ton of outs but has to feel hes behind, Garrett the old top pair top kicker has to be somewhat cautious with calling here as there are some scary rivers, interesting hand!
Loading 27 Comments...
I did not run the hand between Berkey and Adelstein through pio, but i think what Paul missed in his amazing analysis is the effective stack size. I would imagine if this were 100bb effective instead of close to 250bb effective, the SPR will subsequently be very different and hence the the tree. As we all probably know, if we get deeper, one pair hands go way down in value, and the potential to make a good hand goes way up in value like PLO, which explains the merge betting and the OOP choosing to call hands that can improve on rivers rather than hands currently ahead in absolute hand strength. If this were 100BB effective, I would imagine many OOP overpairs to check shove turn for equity denial, which leads to IP betting way fewer combos of hands like Q10 which really wants to realize its equity; since the SPR is still relatively large in the actual situation, these hands are not worried about OOP check shoving and losing their equity.
Makes sense! Nice analysis
Killer comment. A recent debut video by Patrick Cronin is highly relevant to what we're talking about here, since Patrick goes over multiple 3bet spots where he basically demonstrates the difference between being 100bb and being deeper and how that effects optimal strategies.
Check it out
Very nice and interesting video. I like this kind of content
With Garrett raising to only 2.5x preflop and no rake I'd imagine that Robl would be getting in there with this 76o combos. He'd likely be check raising these combos much less often which would then give him ton of straights on the turn even though the suited 76 is heavily discounted.
Pretty sure what’s going on with KJ and AJ checking back turn in Berkey/gman hand has to do with shared cards and the fact AK is floating 100% on flop but folding turn. KJ and AJ can’t bet three streets and are incentivized to check not only for bluff catching rivers but because when our opponent rivers Top pair, we make two pair. Q10 also bets at this stack depth at high frequency on turn because it will only ever get to bet large on river on a 10 or 9, so its more incentivized to merge bet for protection since it struggles on almost all rivers. Plus by folding out oop’s AK on turn, it will be able to more comfortably bet Q rivers.
I like this video. I like to know what kind of caliber poker player i'm watching, but i've never heard of most these guys. They all live cash pros?
Also I am left a bit confused at the 2463 board, where you said it makes sense for Young to lead his entire range for small on the turn. Why does it not make sense to have a bigger betting range where we can put more money in with sets and flushes?
Hey therapist,
I think it's because going small with his entire range allows him to make thinner protection value bets before polarizing his range otr. Similarly, IP is still uncapped after the flop check with a well-played mixed strategy. This tends to lend itself to smaller bets, even though this is a great card for OOP overall.
Cool stuff love this kind of video's!
Curious about jamming 99 in the 1st hand...
I really like the idea of these kinds of videos and I enjoyed the content.
However, you need to allow for multiple bet sizes, especially smaller sizes for the OOP 3better. Otherwise you risk the good old, garbage in = garbage out.
The first hand between Berkey and Garret for example, you say that you were suprised to see that Pio wants to check almost the entire range. However, Pio actually wants to use an overall mixed strategy of close to 40% cbetting and 60% checking, when given the option of betting 25%, 33% and 50% on the flop (and Pio unsurprisingly prefers the 25% bet size).
These results are based on the ranges that you show in your video and allowing for multiple betsizes across all streets for both players, including overbetting the turn and river (I used 50%, 67%, 100%, 150% and 200% plus all in for turn and river).
So just be wary that you risk drawing false conclusions, when you limit the game tree too much and don't allow for the most natural cbetting size.
I hope this doesn't come off too negatively - Keep it up :)
Hey ST, nice comment. And nw, definitely didn't come off negatively.
There's no argument against the fact that more bet sizings on each street will create an output that's closer to "perfect poker". I'm curious though: are you giving OOP multiple flop sizings because you're hoping to implement multiple flop sizings yourself? Or are you looking to see which sizing is most commonly used? If it's the latter, and you're presumably looking to play a strategy with two options (bet/check) at this node, then you'd need to re-sim it with just that sizing anyway for a better output.
There are some people who want to play multiple sizings on every street. Personally, I mostly disagree with this approach, at least on the flop. My goal, generally, is to play as simple of a strategy as possible while retaining a high % of EV compared to a more-complicated equilibrium. There's no way I'm capable of playing a perfect strategy with all its complexities. Similarly, I've yet to see a flop spot where having more than 2 strategic options at the cb node increases EV by substantial amount -- but I guess that's subjective.
I don't think you were advocating for a complex strategy as mentioned in the above paragraph, but that was just a side tangent for anyone interested. I think I gave OOP the option to bet 33% or check. If the perfect single sizing here is 25%, I'm not making a big mistake.
With all that being said, you're right and I should have included more sizings for turn+river. Multiple flop sizings make the sim too big and results too convoluted imo. As long as the flop sizing I gave is reasonable and the ranges are okay (which is the biggest factor imo), I'm fine with using the sim to analyze the hand in a broad strokes way.
ST, how big was EV difference between those 2 strategies (yours and Paul's), obviously we can improve our EV results by running simulation after simulation, tweaking it and getting better results. But does EV difference really worth it?
Hi Paul, thank you for your reponds.
Personally, I I give Pio the option of multiple bet sizes in order to figure out which one is the prefered one/the one with the highest EV.
Like you, I'm not really trying to employ a stratigy with multiple bet sizes myself but rather simplify and going with one size, obv depening on the preflop action and board texture.
You are right that in some situations it would then make sense to do a resim with just that one bet size. However, often/some times it is not really neessary as Pio will have the other bettting options approximate 0% usage and thus those parts of the game tree will be almost empty anyway. Obviously, this depends on how long you run the sim for.
In the hand between Berkey and Garret, Pio rather quickly had the 33% and 50% bet size options that I gave it, drop below 1%.
In the sim for the hand that you show in the video, you had only given the option of a half pot C-bet, which I thought seemed off and would produce vastly different results than when/when also given the option of a small bet size, which I mostly see and use in this situation.
Now you could argue that I'm also limiting the game tree by not giving Pio even more bet sizes to choose from, and that would be correct. In fact, it is very likely that Pio would find an even smaller bet size to be higher EV in this spot.
In generel though, I try not to assume to know what the right betsize(s) is(are) and thus give as many options as my 64 gigs of RAM allow for.
Jeff_
I didn't run the sim for for that long. I had it run until exploiatable for 0.11% of the pot though. This produced an EV for OOP of $6,122 whereas Paul's sim produced an EV of $6,079.
So you are right that the difference in this case is negligible, in which case it's definetily easier to employ a pure strategy.
Ah, for some reason I thought I gave OOP the option to b33, not b50. I can see how that would affect outputs more than I thought initially in terms of betting frequency. Luckily, the EV differences are small. Thanks for all your analysis, and I think we're mostly on the same page philosophically.
The first hand is interesting, especially the turn spot where IP is supposed to merge quite heavily with a big size on the turn. My guess would be that it's really sweet to get protection from AK/AQ/KQ, and by betting 33%, we're letting villain realise too much equity. Also, IP has so many good hands in his range that maybe it's simply too costly to construct a range that bets with a small size, despite the fact that maybe some not so clear valuebets would prefer that in a vacuum but can't do it because then you'd also have to bet small with some really strong hands in order to be properly balanced? Here I'm assuming that very strong hands are much better off betting big, which would make sense?
Cool stuff, cheers. I liked the video despite having 0 hands with Hollywood Haxton.
That could be it. I'm really enjoying reading everyone's responses here and the analysis has been really cool. Thanks guys.
In typical Hollywood fashion, we might need a sequel ;)
Don’t fall into the idea that bets as part of an equilibrium strategy are always classifiable as value or bluffs or anything else. Some combos can e.g. call , fold or raise for equal EV in a particular configuration. Action boards lead to such complex equilibrium with sometimes difficult to classify actions. Often they can be seen to serve a purpose later in the game tree. In any case EV for the range is the optimized parameter. All plays serve just that purpose.
I enjoy your work, thank you.
Yeah, you're totally right. It's difficult to interpret/implement strategies when we can't classify bets into a category to create some intuitive understanding, hence the attempt to do so.
Nice stuff, sub tree configuration definitely best part because didn't know that feature oO
nice video. was a good format that was somewhat chill/relaxing to watch yet also informative.
The format makes for a very visually entertaining video. Would love to see more of these
32:00.
PIO tends to use lower size when the turn card complets 2str8 draw.
This is not Always the case but often the case
Adelstein actually raises hijack and Berkey 3bets from the SB, not the BB.
Interesting analysis anyway!
some of my favourite players to watch play live! thanks for doing the hand breakdowns here very informative!!!
really gross spot here for Garrett and a wicked raise by Robl here on the turn, Andrews actually got a ton of outs but has to feel hes behind, Garrett the old top pair top kicker has to be somewhat cautious with calling here as there are some scary rivers, interesting hand!
Be the first to add a comment
You must upgrade your account to leave a comment.