TripleDraw is a command line utility for studying card probabilities in the games of 2-7 Triple Draw And Single Draw (poker)
The main goal of this project was to build a Python app that uses some advanced Python programming techniques such as multiprocessing, decorators, lambdas and matplotlib charting while also learning more about poker and card probabilities. It is free to use, copy and modify, however no warranties are provided as per the standard open source licence at the end of this document.
TripleDraw lets you evaluate the odds of hitting a given target hand, say 8 low, from pre-flop stage to the last draw stage.
TripleDraw implements a Monte Carlo-style approach to estimate the probabilities of a given scenario. It does that by simulating large amounts of random draws, apply card playing rules and average out the results.
As such it is important to note that TripleDraw does not calculate exact probabilities but rather approximations. The section on Understanding TripleDraw's Results illustrate the difference with an example.
From your terminal window:
git clone git@github.com:jule64/TripleDraw.git
cd TripleDraw
pip install --editable .
The last step will install dependencies and create a tripledraw script which you can call from anywhere in your command line.
The below line will evaluate the odds of being dealt a Jack low (or better) pre-flop:
tripledraw --starting-cards '' --draws 0 --target J
This will print the results as follow
starting cards | nb draws | target hand | simulations | odds (%)
---------------|----------|-------------|-------------|----------
any | 0 | J low | 100,000 | 9.692667%
You can also study more advanced scenarios, such adding one, two or three draws and specify starting cards. In fact setting the number of draws to 1 allows you to study the single draw variant of the game!
The help menu shows you how to run simulations in more details:
tripledraw --help
The --plot option displays a chart of the simulation once the simulation completes. I works by taking snapshots of the simulation at various stages of the computation and renders these values in a time series chart.
A typical simulation chart will show the workers simulations plotted in black while the consolidated results of those simulations are shown in red. The final simulation result is the last data point on the red line and is displayed on the chart as a number.
Charts are nice way to visualize how the simulation converges to its final result.
Odds of drawing to 9 Low after 3 draws
This is a typical example of how a simulation chart looks like:
Odds of drawing to 7 Low pre flop:
This is the strongest hand in Lowball poker, and getting it pre-flop is a rather rare causing the simulation to take this interesting 'saw'-shaped form.
Big simulation (just for fun!)
And finally we plot an 800k simulation using 50 workers, just because it looks cool :)
UPDATE: I have recently ported the code to Python 3 but have broken compatibility to with Python2 and Pypy in the process, I am working on making the code backward compatible and will make it available shortly. Apologies for the inconvenience.
TripleDraw's performance will highly depend on the number of cores on your machine as well as the complexity of the scenarios being evaluated. For example, 100,000 simulations should run in under 0.4 seconds in parallel mode (on 8 cores macbook). 1 million simulation will run in about 4 seconds under the same scenario. However, in a three draws scenario, these performances drop to 1 seconds and about 8 seconds respectively. Note You can set the number of worker/cores to use by setting the --procs option.
Additionally, you might want to try to run TripleDraw with PyPy (http://pypy.org/). PyPy is an optimised Python interpreter that can significantly speed up your program without changing any of your code. In our case, a large simulation will run over 4 times faster with PyPy than with regular CPython, see for yourself, the difference is impressive:
1 million simulations using PyPy vs CPython:
TripleDraw should work on any version of OSX and linux/ubuntu with python 2 or 3 installed.
If you are on Windows the installation step might be a little different but it should still work once installed.
pip uninstall tripledraw
The big advantage of using simulations is that it allows to evaluate complex scenarios without using complex maths, unlike you would using probabilistic calculations. The flip side is that simulations contain some degree of imprecision compared to the actual probabilities of the evaluated scenarios. Let's look at an example to illustrate this point and see how to minimize this problem with TripleDraw.
Suppose you wanted to know the odds of being dealt a Jack low (or better) pre-flop. The mathematical approach gives this scenario a probability of 9.654%(*). TripleDraw by contrast gave values of 9.804%, 9.66% and 9.57% in three separate runs using the default 50,000 simulations, roughly a 0.1% precision around the true value. Since there is no real advantage in using odds beyond one decimal place in poker, the way one could read these results is to say 'the odds of being dealt a Jack low pre-flop are "about" 10%'.
If however you feel those results are too "wide", you can fix that by simply increasing the number of simulations. To take our example again we re-ran the same scenario but this time using 1 million simulations. The results came back as: 9.653%, virtually the same as the actual odds.
Precision however comes at the cost of longer running simulations, and it is up to you to balance out this constraint with your own requirements. Note to change the default number of simulations use the -n option, for example -n 500000 to run five hundred thousand simulations.
(*)By the way if you want to know how this number is calculated have a look in the study folder where I have added all the calculations step by step. Needless to say it's a lot of equations, and that's only for a simple pre-flop probability. By contrast the simulations-style approach of TripleDraw means no complex maths are required :)
If you have any questions or suggestions please feel free to raise an issue or contact me at jule64(at)gmail(dot)com.
Contributions are welcome!
Copyright (c) 2016 by Julien Monnier and contributors.
THIS SOFTWARE AND DOCUMENTATION IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE AND DOCUMENTATION, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.