Hi, I'm trying to code a program that calculates starting hand equities depending on the number of players around the table. I wanted to check if my program is working by comparing it to existing data. People suggested me Equilab and propokertools.com for that purpose so I checked both of them and they give different result. to each others and to my program. The difference is too big to be caused by the monte carlo approximation.
more worrying: on a 2 player set up, where calculations are deterministic (enumeration of all possibilities, no monte carlo), the 2 tools still give different results.
I don't know which one to trust. Does anyone here have some insight on how trustable both of those are ? Do you know a way to contact people who wrote the code in order to figure out what they did to have different results ?
Hi pasmaintenantsvp,
Welcome to PokerStrategy.com!
There are different methods of ranking hands
I'm not familiar enough with either to know if that is what the reason is.
Can you give us some examples?
Thanks
VS
Hi, thank you for replying. Here are 2 examples using a simple 2 player set up showing that the programs give different results even with calculations that are supposed to be deterministic:
Hand1:7s2h
Hand2:unknown (entered as "random" on Equilab and "XY" on propokertools)
propokertools gives : 34.68 % (exhaustive)
Equilab gives: 34,58% (enumerate all) and 34,48% (monte carlo)
Hand1:AsAh
Hand2:unknown
propokertools gives : 86.08 % (exhaustive)
Equilab gives: 85.20% (enumerate all) and 85.23% (monte carlo)
What do you mean different methods of ranking hands ? I thought the rules of hold'em were pretty clear and unified...
Here is a wikipedia article that has a look at a couple of different ways of ranking starting hands.
Again, I don't know if this is the reason for the difference, but it does show that there are different ways to view the same set of cards
https://en.wikipedia.org/wiki/Texas_hold_%27em_starting_hands
Best Regards,
VS
That should not impact the code at all. Thoses are categories based on the equities that I'm precisely trying to calculate.
Some part of the page is also about statistics over real games. ("Statistics based on real online play" section with expected values)
My simulation tries to figure out who would win if nobody folds to know how good is your hand preflop. And I know that Equilab and propokertools are trying to do the same thing as me because:
- they're using the monte carlo approximation
- all our results are somehow close, but not close enough
I was thinking that maybe the difference was lying in the way tie games were calculated but the maths don't add up.
I tried Equilab and propokertools. Gave one player AdAh and for the other a random hand in equilab and in propokertools I just put X there. The results were the same. When I changed the X in propokertools to VR (which means vs random I quess) then the results were different. So I guess the difference comes from how the software randomizes the unknown range or something like that?
It was quite weird that the results in propokertools are different if you input X or VR.
It gives a third answer using XY to set up the random.
On AdAh with 2 players, the second player gets:
with X: 85.20% (like Equilab)
with XY: 86.08%
with VR: 85,49%
And the number of exhaustive trials changes. So I suppose one of them is using the random function on the 52 cards of the deck while another one is using it on the 50 cards (whole deck - the 2 cards that are already selected).
That's definitely a problem since the first one is wrong. And I have no idea how they get the 3rd result.
Anyway that would settle that the one using only X is the right one since it's equal to the Equilab one, but I tried it with more players, when it's not deterministic anymore (using Monte Carlo) and the result between Equilab and Propokertools are still significantly different (With X, XY or VR)
Do you know of an alternative software/source of data that might be more reliable ? (I'm thinking both of them are spaguetti-coded right now)
I have a Mac but no windows-based PC. Is there a compatible version of equilab available for the Mac?
Hi RebelPirate,
I'm not aware of Equilab for the Mac.
Does the Mac still have the facility to run Windows programs?
If so, it should work in that.
All the best,
VS
you're doing it wrong... unfortunately I don't have it PPT installed here to report detailed but:
Originally posted by la55i
It was quite weird that the results in propokertools are different if you input X or VR.
€dit
X is used for suits. V and R are used for a rank.
Originally posted by pasmaintenantsvp
It gives a third answer using XY to set up the random.
On AdAh with 2 players, the second player gets:
with X: 85.20% (like Equilab)
with XY: 86.08%
with VR: 85,49%
(...)
Anyway that would settle that the one using only X is the right one since it's equal to the Equilab one, but I tried it with more players, when it's not deterministic anymore (using Monte Carlo) and the result between Equilab and Propokertools are still significantly different (With X, XY or VR)
You should use * for random.
€dit:
X is a suit
XY is off-suit (two different suits)
VR are two different ranks
-> three different ranges
You can also use the range explorer to dig deeper.
I don't have anything installed either. I used this web calculator: http://propokertools.com/simulations
If X is just a suit, how I can give player 1 a hand like AdAh and player 2 I only input X? What hand range it would then give player 2 and why the result is the same than Equilab using random?
Sorry did an oopsie.
Originally posted by la55i
When I changed the X in propokertools to VR (which means vs random I quess) then the results were different. So I guess the difference comes from how the software randomizes the unknown range or something like that?
VR is without pocket pairs.
* a random hand
X Any of some suit x and a random card
In this case it's the same but it's read different. One notation you take a random hand; on the other you have to check first if it is the case that this hand has some suit.