Hi, I just bought the standalone and have some comments/suggestions & questions. I'm a gtorb user, so some or most of my comments will be about things I can do in gtorb but not in Simple Postflop (at the moment).

1) When calculating the flop, you basically iterate over all turns & rivers, using the betsizes you provided at the start.
But let's say I want to add a very specific line on the turn, for example an overbet lead of 2x pot for OOP when IP checks back on the flop. I don't want the 2x pot to be available in every node, only in the case of a flop check back.

This can be done by adding a node on the turn, but it doesn't alter the flop strategy that's already calculated (correct?). This seems flawed, since our flop strategy should take into account the possibility for a turn overbet lead.

Wouldn't it make sense to be able to edit the whole tree (all 3 streets) before calculating the flop (instead of only being able to edit flop)?

2)
When adding/editing a raise node, it uses 'raise by', please change this to 'raise to'.





3) When making a flop 3bet, seems like i can't 'click it back'. Bet 9, raise by 15 to 24, minraise should be by 15 to 39, but seems like simple postflop only allows 2x 24 as a minimum raise.





4) When manually entering effective stacks, I can only 2 digits before the comma (no problem for 'main pot'). I can however use the arrows to go above 99, so it should be possible. Pls fix :P





5) It would be nice if frequencies of how often a combo is still represented in a certain node of the tree could be visible in the 'strategy show' view.

Simple Postflop:


Gtorb:


I like how gtorb does it.

6) I have tried some examples from gtorb in simple postflop and sometimes got slightly different results. I assume this is because of the semi-indifferent hands which can take different lines and all be GTO for the given epsilon?

7) When I want to change all the betsizes used in a tree, I have to click 'clear situation'. But this deletes the starting ranges, which is not what I want.

That's it for now, but definitely looks good

Issue 4) is gone atm, maybe I did something stupid, maybe it happens only occasionally..

EDIT: and it's back, don't see the logic..

Request/remark nr 8):
I'd like to be able to edit the flop without having to 'clear situation' (which makes me lose my preflop ranges again). I'll probably want to change the flop a couple of times and do the same calcs for the same betsizes & ranges.

Does it do flop calcs?

Issue 4, effective stack are capped at 200,00.
this isn't really a problem since you can just scale any stack down (so 1.400,00 stack and 200,00 pot becomes 140,00 and 20,00) - it was confusing for me at first as well.

I agree with you on #8, having ranges deleted because you needed to 'clear situation' to edit something is clearly a mistake.

yes.

Ah, tx. Yeah, it's not really a problem, but it's still an inconvenience that could be easily solved imo.

Additional remarks:

9)When forcing an OOP check, you can delete the 'donkbet node' & recalculate. However, instead of displaying a 100% check, it still shows the previous % (when leading was still an option). I do think it's just a matter of displaying the correct %, as internally, there seems to be a 100% check.



10) With an epsilon definitely <<< 0.2 (don't remember exactly), why would we ever be indifferent between all 3 options as shown here:



(It's a 100bb pot, facing a bet or raise of 10)

11) In a very rare spot (the raise to 28.88 was taken 0.2% of the time or something similar), I encountered some NaNs for Evs:

12) The save & load function doesn't seem to be working? I did some trees and saved them, now unable to load any of them:

I had been saving with the 'Save Tree' above 'Clear situation'.

Talked to teelXp on skype:
"Save Tree use for save tree with betsizing and ranges but without board and strategies, with help save on menu you keep strategies and board"

So if you want to save a tree for later use, use the File menu on the top left and then click save. This did the trick for me and it seems to be working now.

13) Why I have the standalone version and do local calcs, why does it start with 'sending' ?


At first I thought it was just cosmetic, but I had actual timeouts which seem weird with local calcs.

14) It would be nice to be able to define the epsilon instead of having to choose between low, middle & high.

The trees would imo be much easier to read if "Player 1" and "Player 2" were renamed "OOP" and "IP" (or custom names allowed, so people could name them "BB" and "BTN" or w/e).

What does "Around bets" and "Treshold for all %" exactly mean ?

Hello mynameiskarl!
Sorry for the delay in responding.

So, we have released a hot fix which has resolved most of your issues. I also want to answer your questions below in order:
1) As you say, it is possible to adjust all bet sizings in the trees manually but you need to edit turns for every endpoint of flop. We'll think further on how to improve the bet sizings menu to build more custom trees.
2) Fixed in update
3) Fixed in update
4) The idea was to evaluate stakes and pot in BB. So I guess, you don't need stacks more than 200 bb. As for now, this is an interface restriction. We’ll add this to our future list of improvements.
5) We'll add this to our future list of improvements
6) We use different algorithms with different convergence to gtorb.I assume that there would be significant differences in our results so it’s actually quite interesting that many of the solutions are close to identical. Both results are measureable compared to GTO.
7) There is an option to save the tree by clicking “save tree” which saves the bet sizings and ranges. You can load this later by clicking the drop down menu next to “apply situation”
If you want to save the whole tree, ranges and board, you save the information by clicking “file” then “save” .
8) Same issue as 7 I guess
9) Interface bug, fixed in update
10) It depends on Nash distance and accuracy mode. If this ran for more iterations and the calc became more accurate, there would be less of these issues.
11) I’m guessing one of the players has zero range in this branch – let me know if this isn’t the case
12) You've already got the answer
13) There is a process which allocates memory and for the complex trees it sometimes fails by timeout. This has been fixed in the update.
14) We'll give this some thought.

Hi!

1)Thanks for the proposal we'll fix this in the future - it's imposible to correct this through the hot fix as these labels used are used in many places in the program
2)We've made the labels more intuitive in the hotfix update

Just to be sure: to get the hotfix, we simply go to the website and do the whole "download" process again, correct ?

No, you just need to reboot the program and you'll see the pop-up window which will recommend you to get the update.

I want to add:
For action frequencies with specific combos in general GTORB and Simple Postflop should match very closely, especially in spots where both programs Nash Distance is low. For aggregate frequencies (eg the % of the time that a player checks) GTORB displays the frequency that would be observed where the two strategies to play against each other, incorporating the effects of blockers from the opponent range. Simple Postflop instead shows the aggregate checking frequency ignoring any blocker effects due to the opponent range composition so these aggregate numbers will often be somewhat different even if the strategies are identical because the same data is being displayed in a different way.

IME every time I open SP it will automatically prompt me to download if there is a new update.

Mynameiskarl,
Re: #10
"<< 0.2" isn't enough information.

The distance is given in terms of the same units (chips, bbs) of your scenario.
a 0.2 distance in a scenario with stacks of 100 and a pot of 2 would be a much different distance than if it was .2 with stacks of 200 and pot of 20 (the latter one would be much better solution since .2 is a smaller fraction of the 20 pot than it is of the 2 pot)
To normalize things you can take the distance and divide it by the pot or by the effective stack. So for pot it would be [nash distance]/[starting pot] = %-pot-nash-distance.

Np & tx for the response

So you would say both you and gtorb provide adequate & correct solutions for a given scenario (ranges, board, betsizes) for a certain epsilon?

Can you elaborate on how your algorithm differs from gtorb, or what that be giving away too much?

I very often encounter spots where the EV of bet & check is extremely close, so I can imagine slight shifts in strategy can occur that still is GTO for a certain epsilon.


Had to think about that, but I think I get it

Ok thanks, I finally get it now But I do think this system is extremely confusing at first. Mostly because it's also 'save tree' while it doesn't save the whole tree.

Thanks for the quick fixes

I figured the EVs would also be expressed in the same units... So if we have an epsilon of < 0.2, there can be no doubt between an action of +0.5 EV and -0.5, since the Ev difference is almost 1, which > 0.2.

Don't have that tree anymore and don't remember the exact spot. But I'll keep my eyes open.

GTORB and Simple Postflop will both be accurate to the given epsilon distance and in low nash distance situations they should produce identical or near identical strategy solutions.

However, because of the difference in how aggregate action frequencies are displayed (as I mentioned before) it might appear that solution strategies are different when they are in fact the same. If you find spots where you believe there is a difference, make sure that you are looking at the frequencies with which actions are taken with a specific hand combo and not at something like the overall betting frequency.

At the specific hand combo level, it is possible for action frequencies to differ by a few percent and to still be well within the given nash distance so you may see small differences in actions with specific hand combos when comparing GTORB and Simple Postflop but usually these differences should be very small.

We'd rather not divulge any information on our algorithm.

Np, tx for explaining.

I encountered some new issues:

1)Initialization error: had this a couple of times, can't pinpoint the steps I took causing it. Mostly happens when a run a scenario, clear the situation, change betsizes, reload some saved ranges... and then try to run the altered scenario.

Clearing situation and reloading a previously saved 'tree' for 'apply situation' solved the problem.



2) While manually adding/editing a river node, the bet sliders lowest point seemd to be the previous betsize. In this example: 4.12 was the last bet on the turn and the lowest point of editing a river node. However, this 4.12 didn't count as 4.12, slider at 4.12 seemed to be a bet of 0 or -0.01...



This caused random behavior and got me a bet of -0.01, which I was unable to delete with 'right click -> delete node'. Had to rerun the whole solution to get rid of it.

In general, it might be a good idea to add a feature that enables logging.

I've recorded a review/tutorial for Simple Postflop.
I didn't get to cover everything I would have liked to, but I believe I covered most of the basics of setting up a scenario and navigating through the data of the results. I also give my impression of the software and near-nash solvers in general.

Disclaimer: I'm not an elite level player, and I can't speak to the absolute accuracy of the results (although everything I have tested has looked in-line). I'm just a poker player with a passion for poker, game theory, and teaching - I was excited to test out this software and I felt like making these review/tutorials for other players who were interesting in buying or had bought but needed some help understanding it.

If you're watching on your computer I'd recommend changing the play-back speed to 1.25x (in the youtube player click the gear icon >> speed >> 1.25x)


I think Simple Postflop is a very powerful tool and I think eps'-equilibrium tools like this can be very useful for learning new concepts and getting a deeper understanding for how the game works - as long as you are very careful about what you try to extrapolate from the results and how you go about finding it; not just trying to mimic their play - and with multiple bet-sizes, flop+ calculations, and the ability to lock/edit a players strategy and calc a new nash, SP is, IMO, easily going to be worth it's cost for any grinder who wants to put in the time to work with it properly.


I don't work for SP, I'm not affiliated with them and I don't have any financial stake in their company. I don't know anything about the team and this product is the first I've heard of them - I'm simply reviewing the product I've used.

1) This error is caused by the large size of the tree which couldn't be solved by your computer.
If you reduce the number of bet-sizings and increase your sizings, you could make your tree less complex.
Btw we'll try to reproduce the bug so we can provide a more useful error message.
2) Yes, that is an interface bug. We are going to fix it soon.

Thanks for the suggestion.
We will continue to improve our pop up messages and error tracking processes

Thanks a lot! This is a really nice job. I am sure that your video will cover most of our follower's questions.