[GTO] GTO+/CardRunnersEV?
06-25-2008
, 10:19 PM
06-26-2008
, 03:23 PM
06-26-2008
, 04:08 PM
Never thought you were sarcastic. 
06-28-2008
, 02:02 AM
06-28-2008
, 07:34 PM
Hi, I just downloaded this and I'm trying to do a simulation for preflop 5-bet-shoving. My goal is to figure out what the optimal value for var1 is to shove against villain's 4betting range in the given situation. However, the graph produced is very spazzy so I figure I likely did something wrong.
Setup:
Graph produced:
Also, the graph that is produced looks very different every time I create it.
EDIT: I figured that the "top var1%" might not be the right thing since it's not "top x% of my range at that point", it's "top x% of all starting hands" which isn't what I want. I changed it to "var 1% weighted - all hands" and the graph now looks like this:
Look good?
EDIT 2: I suspect that "var 1% weighted - all hands" is just "x% of the time, we shove random hands" - is there a way to do what I want (shove the top x% of my range at this point)?
Setup:
Graph produced:
Also, the graph that is produced looks very different every time I create it.
EDIT: I figured that the "top var1%" might not be the right thing since it's not "top x% of my range at that point", it's "top x% of all starting hands" which isn't what I want. I changed it to "var 1% weighted - all hands" and the graph now looks like this:
Look good?
EDIT 2: I suspect that "var 1% weighted - all hands" is just "x% of the time, we shove random hands" - is there a way to do what I want (shove the top x% of my range at this point)?
Last edited by goofyballer; 06-28-2008 at 07:50 PM.
06-29-2008
, 04:39 PM
Hi Goofyballer,
You definitely need the first method and are correct with how you approach it (the second approach doesn't work).
The results are most likely spazy because most of your runs go through the part of the tree where the button folds instead of raising to 54.
The easiest way to get around this would be to make his original raising range the same as his "raise to 54" range.
Now all runs go through the part of the tree where the sb raises to 200 or folds.
Alternatively you could of course increase the number of runs. The first method is faster though.
Cheers,
Scylla
You definitely need the first method and are correct with how you approach it (the second approach doesn't work).
The results are most likely spazy because most of your runs go through the part of the tree where the button folds instead of raising to 54.
The easiest way to get around this would be to make his original raising range the same as his "raise to 54" range.
Now all runs go through the part of the tree where the sb raises to 200 or folds.
Alternatively you could of course increase the number of runs. The first method is faster though.
Cheers,
Scylla
Quote:
Hi, I just downloaded this and I'm trying to do a simulation for preflop 5-bet-shoving. My goal is to figure out what the optimal value for var1 is to shove against villain's 4betting range in the given situation. However, the graph produced is very spazzy so I figure I likely did something wrong.
Setup:
Graph produced:
Also, the graph that is produced looks very different every time I create it.
EDIT: I figured that the "top var1%" might not be the right thing since it's not "top x% of my range at that point", it's "top x% of all starting hands" which isn't what I want. I changed it to "var 1% weighted - all hands" and the graph now looks like this:
Look good?
EDIT 2: I suspect that "var 1% weighted - all hands" is just "x% of the time, we shove random hands" - is there a way to do what I want (shove the top x% of my range at this point)?
Setup:
Graph produced:
Also, the graph that is produced looks very different every time I create it.
EDIT: I figured that the "top var1%" might not be the right thing since it's not "top x% of my range at that point", it's "top x% of all starting hands" which isn't what I want. I changed it to "var 1% weighted - all hands" and the graph now looks like this:
Look good?
EDIT 2: I suspect that "var 1% weighted - all hands" is just "x% of the time, we shove random hands" - is there a way to do what I want (shove the top x% of my range at this point)?
06-30-2008
, 12:41 PM
Hi guys,
I hereby release version 2.0 of the Stoxpoker EV calculator.
A new math-based engine
I've spent the past month writing a math-based engine for heads up play.
The engine is used by StoxEv if
a) The board is unknown and all the action is preflop
b) There is postflop action and at least the flop is known
The results will be mathematically accurate in case the ranges aren't too wide, otherwise I use an approximation which is extremely close to being mathematically accurate. The only exception being if you start shoving with deep stacks and wide ranges (which of course makes little sense).
Here's a comparison of a graph with the old and the new engine:
How do I turn it on?
You can turn the new engine on by selecting extra->mathematical solution.
For postflop play, if you also select extra->forced enumeration then the results will be entirely mathematically accurate. However, in case only the flop is known this may take some time (maximum I've ever had while testing was 45 seconds), depending on how wide the ranges are and how many lines there are to calculate.
Preflop play is always mathematically accurate.
Time
Preflop calculations take less than a second (I could make it faster but have decided not to bother for now).
In case of postflop play or in case the turn and/or river are known the results are instant.
In case only the flop is known and "forced enumeration" is turned off, the results typically take less than 4 seconds.
I recommend always leaving "forced enumeration" turned off, however, I've included it for now for testing purposes.
Results
The mathematical engine works exactly the same as the Monte Carlo based one and even includes weighing (not a trivial accomplishment). You don't need to know anything special about it to work with it.
The engine will calculate the equities and EV for
- all decisions
- all actions
- all conditions
- all individual holecards.
I've decided not to include the preflop play, chance of best hand and hand analysis features in order to achieve more speed.
Testing
I haven't found any differences between the Monte Carlo engine and the Math engine, but I would greatly appreciate if everybody compared the results of both engines and inform me of any discrepancies.
Other changes:
Checkdown mode
I've included a checkdown mode to the possible actions. The pot will be divided among the active players according to their equity.
In case for instance 3 players are in the hand, player 1 has bet 35, player 2 has bet 120 and player 3 has raised to 1000 after which player 3 goes to checkdown the following will happen:
Players 1 2 and 3 will get their equity in the main pot after which player 2 and 3 get their equity in the sidepot.
I will add another mode in a while where you can select how much of the pot you feel you'd win. This will greatly reduce the need to build big trees and allow you to achieve comparable results with small ones.
Selecting custom ranges within a preflop condition
You can now use another range for a preflop condition than the general one. Select 1, 2 or 3. If you select nothing then the general range is used.
Bugfixes
Some minor bugfixes. Also Prima imports should work again.
Also, when you now add a new action, an "all hands" condition is automatically added to it. Works much more pleasant.
Go here for the files or if that doesn't work go here: http://www.megaupload.com/?d=76EZLA7X
Cheers,
Scylla
I hereby release version 2.0 of the Stoxpoker EV calculator.
A new math-based engine
I've spent the past month writing a math-based engine for heads up play.
The engine is used by StoxEv if
a) The board is unknown and all the action is preflop
b) There is postflop action and at least the flop is known
The results will be mathematically accurate in case the ranges aren't too wide, otherwise I use an approximation which is extremely close to being mathematically accurate. The only exception being if you start shoving with deep stacks and wide ranges (which of course makes little sense).
Here's a comparison of a graph with the old and the new engine:
How do I turn it on?
You can turn the new engine on by selecting extra->mathematical solution.
For postflop play, if you also select extra->forced enumeration then the results will be entirely mathematically accurate. However, in case only the flop is known this may take some time (maximum I've ever had while testing was 45 seconds), depending on how wide the ranges are and how many lines there are to calculate.
Preflop play is always mathematically accurate.
Time
Preflop calculations take less than a second (I could make it faster but have decided not to bother for now).
In case of postflop play or in case the turn and/or river are known the results are instant.
In case only the flop is known and "forced enumeration" is turned off, the results typically take less than 4 seconds.
I recommend always leaving "forced enumeration" turned off, however, I've included it for now for testing purposes.
Results
The mathematical engine works exactly the same as the Monte Carlo based one and even includes weighing (not a trivial accomplishment). You don't need to know anything special about it to work with it.
The engine will calculate the equities and EV for
- all decisions
- all actions
- all conditions
- all individual holecards.
I've decided not to include the preflop play, chance of best hand and hand analysis features in order to achieve more speed.
Testing
I haven't found any differences between the Monte Carlo engine and the Math engine, but I would greatly appreciate if everybody compared the results of both engines and inform me of any discrepancies.
Other changes:
Checkdown mode
I've included a checkdown mode to the possible actions. The pot will be divided among the active players according to their equity.
In case for instance 3 players are in the hand, player 1 has bet 35, player 2 has bet 120 and player 3 has raised to 1000 after which player 3 goes to checkdown the following will happen:
Players 1 2 and 3 will get their equity in the main pot after which player 2 and 3 get their equity in the sidepot.
I will add another mode in a while where you can select how much of the pot you feel you'd win. This will greatly reduce the need to build big trees and allow you to achieve comparable results with small ones.
Selecting custom ranges within a preflop condition
You can now use another range for a preflop condition than the general one. Select 1, 2 or 3. If you select nothing then the general range is used.
Bugfixes
Some minor bugfixes. Also Prima imports should work again.
Also, when you now add a new action, an "all hands" condition is automatically added to it. Works much more pleasant.
Go here for the files or if that doesn't work go here: http://www.megaupload.com/?d=76EZLA7X
Cheers,
Scylla
Last edited by scylla; 06-30-2008 at 12:47 PM.
07-01-2008
, 12:41 PM
07-08-2008
, 08:46 AM
Hi guys,
I've just released stoxev 2.0.1.
Go here or here: http://www.megaupload.com/?d=LBJANV0F.
New features:
Math engine
- The results for the Monte Carlo engine and the new Math engine can now be merged so that the Monte Carlo engine supplements everything that the Math engine does not calculate (hand distribution and such). To turn this feature on/off go to "Math engine->merge monte carlo and math".
- Progress bar for the Math engine for flop calculations
Improvements to "add action menu"
The edit boxes behind call and raise still work the same way as before. However, I've added slider bars. If you turn on "odds" or "pot" then using the slider bar will vary the contents of the edit box expressed in odds/pot.
Improvements to checkdown
The previous checkdown let everyone win their equity in the pot. In 2.0.1 it is now also possible to let one of the players win more or less than their equity in the pot. The standard value is 100%, meaning the selected player will win 100% of his/her equity in the pot. This is the standard checkdown. However, you can now select for instance that the small blind wins 120% of his equity in the pot.
If a player wins more/less then his equity then the discrepancy will be deducted/added from the other active players.
The stop action has been removed
I've removed the stop action because it is a simplified version of a checkdown. If you want to for instance let the small blind win nothing, then just let him check down with a 0% share of his equity. The new checkdown mode is superior to the stop action because no chips are lost.
Warning: If you have files that use a stop action, the stop actions have now been turned into checkdowns. However, you still need to tell the program which player wins 0% of his equity.
Cheers,
Scylla
I've just released stoxev 2.0.1.
Go here or here: http://www.megaupload.com/?d=LBJANV0F.
New features:
Math engine
- The results for the Monte Carlo engine and the new Math engine can now be merged so that the Monte Carlo engine supplements everything that the Math engine does not calculate (hand distribution and such). To turn this feature on/off go to "Math engine->merge monte carlo and math".
- Progress bar for the Math engine for flop calculations
Improvements to "add action menu"
The edit boxes behind call and raise still work the same way as before. However, I've added slider bars. If you turn on "odds" or "pot" then using the slider bar will vary the contents of the edit box expressed in odds/pot.
Improvements to checkdown
The previous checkdown let everyone win their equity in the pot. In 2.0.1 it is now also possible to let one of the players win more or less than their equity in the pot. The standard value is 100%, meaning the selected player will win 100% of his/her equity in the pot. This is the standard checkdown. However, you can now select for instance that the small blind wins 120% of his equity in the pot.
If a player wins more/less then his equity then the discrepancy will be deducted/added from the other active players.
The stop action has been removed
I've removed the stop action because it is a simplified version of a checkdown. If you want to for instance let the small blind win nothing, then just let him check down with a 0% share of his equity. The new checkdown mode is superior to the stop action because no chips are lost.
Warning: If you have files that use a stop action, the stop actions have now been turned into checkdowns. However, you still need to tell the program which player wins 0% of his equity.
Cheers,
Scylla
07-09-2008
, 07:11 AM
Hey Scylla, so I'm still a little confused about this. Here's my model again (slightly updated):
I defined my 3bet range as "top 30%" using the slider.
When I graph my decision to shove or fold by EV, I get this:
Is this seriously telling me that in this model, I should be shoving 100% of my 3bet range over a 4bet? I guess I can see the possibility of that being true if he's folding too often, but it still seems like a very surprising conclusion.
Just wanted to double check that I'm doing everything ok.
I defined my 3bet range as "top 30%" using the slider.
When I graph my decision to shove or fold by EV, I get this:
Is this seriously telling me that in this model, I should be shoving 100% of my 3bet range over a 4bet? I guess I can see the possibility of that being true if he's folding too often, but it still seems like a very surprising conclusion.
Just wanted to double check that I'm doing everything ok.
07-09-2008
, 07:16 AM
07-09-2008
, 08:40 AM
07-11-2008
, 02:05 AM
07-11-2008
, 06:31 PM
07-11-2008
, 06:40 PM
07-12-2008
, 04:50 PM
