oh, it looks like you beat me to it, just saw your comment ♦
Notifications
Clear all
Originally posted by jules97
oh, it looks like you beat me to it, just saw your comment ♦
Yep. I'm a big Thinking Poker Podcast fan. We mention them in the book, and I've had a couple of emails read out.
Originally posted by jules97
The biggest thing to me is still the mixed strategies. Which as I've said before aren't the most common. I'm going to have to go away and have a good think about coming up with some simple, solid rules to implement some more mixed strategies in my game.
One thing to bear in mind is that there are obvious examples on the river where a GTO mixed strategy is clearly nonunique but the best strategy to play in practice is obvious. For example, if Villain has nuts or air in a GTO strategy and you have to call a bet with 50% of your bluffcatchers, which might be, say 99-QQ, an algorithm that finds the GTO strategy, like CFRM, may converge to a solution that calls a total of 50% of these hands, but some random mix of them, rather than the obvious JJ & QQ, which is clearly better. Calling with any 50% of these hands is GTO, and all have the same EV. Calling with JJ & QQ and folding 99 & TT is actually using exploitative reasoning to choose the best practical solution. I've implemented this in my code for river jam calling ranges, because the solutions suffered from this obvious problem.
I've mentioned this for two reasons.
Firstly the same may apply on earlier streets. Have a careful look at, say, a turn raising range, and you may find that the strategy is mixed, but may be co-optimal with a purer strategy. Not necessarily, but it's possible, if you get the overall frequencies right. I may be able to get the algorithm to sort this out automatically. Allowing Villain to play every strategy, even the crappy ones, with a small frequency, should do the trick, and once I have the epsilon exploitability coded up, I should be able to assess the effect of this more quantitatively. I tried it before, and it looked like a small change like this was having a nontrivial effect on some of the frequencies, so I stopped doing it, but, with hindsight, this may just have been the sort of shifting effect I was hoping for.
Secondly, the CFRM GTO strategy has plenty of decision points where it takes some actions with zero frequency, e.g. jamming 100b on the flop. This means that it has not calculated a GTO response to this action, which may in turn mean that, as I've coded it up, the solution may be exploitable by, for example, a strategy that involves jamming some flops. I think I should exclude these from my epsilon exploitability calculation, but it feels wrong. Again, allowing Villain all strategies with a small frequency should, in principle, if the algorithm runs for enough iterations, allow it to develop the best response to, for example, flop jams.
Enough already.
Anybody else want to play tomorrow night?
Club ID number: 895115
Invitation Code: jbpatzer123
Originally posted by roopopper
Originally posted by UPAY4DINNER
8:15I'm in
Awesome, I like fish with my chips
Originally posted by UPAY4DINNER
Originally posted by roopopper
Originally posted by UPAY4DINNER
8:15I'm in
Awesome, I like fish with my chips
Just got home from work Bricklayers on site took me for a few beers after work....and I have beer in fridge...just going to order a Chinese then it's Zimmers for me
Told you I would make it
Best comment of the night....... "because you left it to the last minute" 
One of us is seriously disturbed, jb
When I was a kid, facing taunts from those both bigger and faster than I, I would simply imagine their heads exploding. Worked wonders. It never actually happened unfortunately...
As an adult, I found that this was common.
Peace,
VS
"Spin and Go. The most perfect recreational iPad poker game ever invented. When those numbers spin and I get to see what we're playing for, I get all excited! Maybe I should get out more.... "
+1 
Originally posted by deralte71
"Spin and Go. The most perfect recreational iPad poker game ever invented. When those numbers spin and I get to see what we're playing for, I get all excited! Maybe I should get out more.... "+1
I think I've got it out of my system now. I made about $100 playing $3 Spin and Gos, but there's nothing like a solid run of $6 prize pools, where you need to win 50% in a three handed game to break even, to put you off them. Still the perfect format for toilet poker though.
Originally posted by jbpatzer
Still the perfect format for toilet poker though.
Toilet poker?
I understand the "go" part, but "spin and go" doesn't work with the toilets we have in Canada.Peace,
Originally posted by jbpatzer
:rage:
with this stuff
can it be used as a basis to make gigantic arse exploits?
The GTROB blog post on it, does some piddly little exploits. Which is great and all. But bigger ones are more fun & profitable, even if riskier. I remember Alex said he was planning to eventually build some into GTORB. That would be pretty awesome to play with.
My very advanced exploiting strategy is, do what I think I can get away with. It would be nice to have a more systematic and more efficient approach. Without doing any maths of course...
Got to try the turn stuff on GTROB recently, that was fun. Handles some pretty big ranges in less than a minute.
I just updated my hud yesterday. Got some new fancy heatmaps and in the process needed to start a new database and export/import a whole heap of hands. No sooner had I done that, then pokertracker updates and says... 'oh yeah, by the way, reindex your cache- 3034938434 days remaining' so now repeating the process... 
Originally posted by jules97
Originally posted by jbpatzer
with this stuff
can it be used as a basis to make gigantic arse exploits?Must resist the temptation to post a gigantic arse picture. Must.....resist......
Spoiler
Phew...that was close.
The basis of calculating epsilon exploitability is to allow one of the players to be maximally exploitative, which is always to play a pure strategy, i.e. if player 1 bluffs 0.00001% too much he calls 100%. You can't get an arse much more gigantic than that.
My current attempt at this gives the exploiting player more EV than I would think is reasonable in a very small game, and when I ran it on a 3bet pot range (took 18 hours because I haven't translated it into C++ yet) it gave the exploiting player less EV than he has at equilibrium, so I've clearly been, to continue the theme, a gigantic arse. Struggling to motivate myself to work on it at the moment.
On the bright side, Thursday night is Zimmers night! Home game anybody?
This appeared in our local paper today.
What do you think?
(sorry -- this is the only place I could think of where someone might know)
Thanks,
VS
