I usually use Pio pro version and its "load script" feature to create multiple cfr files and use them for statistical analysis. Then, I'm considering using Amazon's EC2 for more solving speed. Anybody who has similar experience?
My questions are;
- If I upgrade pro to edge version, can I activate my license on EC2?
- If I stop or delete the instance of EC2, is the license available again?
- I heard that there is no upper limit on the number of cores available for edge, but is there a clear speed difference between 8 cores and 16 cores, or 16 and 32?
- If you know some recommended instance settings, please let me know:)
Dec. 25, 2019 | 9:12 a.m.
I suspect that most high stakes hold'em players use preflop solutions such as Monker. Whatever range an opponent has, we can hold fixed EV and learn easier postflop GTO strategy by imitating Sims.
However, preflop GTO is not completely solved. The main reason is that the game tree of hold'em is too huge, but from the perspective of solving Nash equilibrium, I think the bigger problem is multi-way possibility. Theoretically, we cannot solve any Nash equilibrium in multi-way situation.
For example, when you 3-bet from SB against BTN, there still remains a player(BB) behind you. So if you want to do some GTO stuff here, you have to assume that BB will fold to 3-bet 100% of the time. Then heads up with BTN and SB, now you can solve the Nash equilibrium. But, of course, this premise is unrealistic. Because nobody folds AA, KK, AK, or maybe KQs from BB against SB's 3-bet.
So my questions are;
Do Sims really have reasonable performance even in multi-way situations?
How are Sims made actually?
What do you guys think about this?
Sept. 30, 2019 | 11:35 p.m.
Great. Good to hear that. Thanks.
The reason why I stick to flop OOP is postflop approximate GTO. For example, I run script of UTG vs BTN single raised pot situation with multi bet sizing, and I get to know the frequency of CB is relatively low. So I assume UTG's approximate flop GTO strategy is just check with all his range. Then I have to confirm that to run script with OOP force to check with all range and see EV loss.
Sept. 19, 2019 | 6:59 p.m.
I usually use nodelocking to force OOP to check especially on the flop. But as far as I know, you can't run a script at the same time you use nodelocking. What do you do when you run script and don't want OOP to donk bet? I guess the only way to do this is leaving flop bet size column of OOP blank. Yes, there is the "force OOP check/IP bet" checkbox , but this is not perfect because I don't want IP to bet 100% of the time. Leaving the column blank looks still working OK, but I’m worried about the effects on turn or river strategies.
So, my questions are:
1) When you usually run a script, how do you force OOP to check on the flop 100% of the time without using "force OOP check/IP bet" checkbox?
2)What do you think about the effects of a blank column on strategy or EV?
Sept. 16, 2019 | 8:29 p.m.
Thank you for your lucid instruction. Yeah, you are definitely right. I kind of knew that and now I'm sure.
So, how do I study multiple bet sizing? I think there is no single approximate strategy especially on SRP. At least 3 or 4 strategies are required, so post-turn aggregation report does not work well.
We have to study SRP post-turn GTO tendency with other way? Any idea?
July 13, 2019 | 2:49 p.m.
I'm sorry that my English isn't very good. Let me explain that more carefully.
For example, I run a script with 487subset. This is SB vs BTN 3bet-pot situation. Then, I get the aggregation report and make it understandable by statistical processing . So I finally know that SB should bet with high frequencies on pair boards and bet sizing is relatively big, but on the other hand, SB should check with high frequencies on 8 high boards and bet sizing is very small.
But this is just a flop GTO tendency. So I have to know that on turn or river to use it on real game. I start to do the turn stuff in the same way, but I think I can't because the flop strategy uses multiple bet sizing.
When I try to get aggregation reports on turn, I click one specific node on Pio(eg. OOP x, IP 30%pot bet, IP call, turn card). So with this configuration, turn aggregation report shows me only "OOPx, IP 30%pot bet, IP call" node over every flop subset. I think this is a problem. Because some of flops never use such line or bet sizing. So the turn report becomes meaningless.
Maybe am I wrong? I have lost my confidence in myself...
July 13, 2019 | 10:18 a.m.
It is generally known that GTO strategy has a larger EV over the range with multiple bet sizing. Aggregation report allows us to determine which board is appropriate for each size. For example, LJ vs BB SRP: LJ can make a small size bet with whole range on K high boards, while LJ can make a bigger size bet or check on low pair boards. It's fine, imo.
So my question is, how do I get some aggregation reports of the post-turn strategy when using multiple bet-sizing on flop?
Aggregation reports of turn require specific one specific node of flop. For example, OOP checks, IP 30% pot size bets, OOP calls and then we see a turn card. But when we use multi-bet sizing on the flop, the nodes we select on the flop are different for each board. So we can not create the turn report properly over multiple boards which we use on the flop.
I was exhausted by looking into this issue, so I simplified my strategy in 3betPot flop by using one bet size. However, in the case of SRP, the EV loss is quite large when using one betsizing with whole range.
What do you guys think about this issue? I need some advice on this.