Recently, I just imported PT4 database into SPSS and ran some correlational analysis. It seems fun. Basically in many spots, we have a tiny sample of a specific action (e.g. bet flop then c/f turn). But we may have a very large sample of other stats (e.g. WTSD, PFR, etc). We can check if Specific Small Sample Stat A is correlated with General Large Sample Stat B. If A is positively correlated with B, we can use B to infer and guesstimate A.

Anyone also use SPSS/Python/Matlab or whatever programs to do this sort of stuffs?

I am very new to programming. Apart from correlational analysis, what analysis can be ran (which is useful for poker)? What programs are best for doing this sort of work?

Not quite tried that. But I might at some point. I've been using Python recently to do a bit poker-economy simulation for interest / programming practice. I'd recommend Python or R for the sort of statistical analysis you're after. For Python, use Anaconda (Google it) and Jupyter Notebooks. For R, use RStudio. Check out Datacamp.com for training on both.

I've used R and Python/numpy a bit for various academic research. I'd go with Python just because you're learning a real-life programming language that's useful for a lot more than just high-powered stats. R feels like it was written by academics, which is was, so its funky syntax tilts me. I did discover one really useful R function that wasn't present in a Python library, I forget which, but in general Python is fun and intuitive.

(R is valuable in the job market because it's mostly data science folks using it; Python is valuable because it's far far far more common in industry jobs.)

+1 for Anaconda -- comes with numpy, scipy, pandas, all libraries you'll want to be using. The Jupyter Notebook is nice for sharing quick and dirty calculations in a systematic format.

For serious debugging, I highly recommend PyCharm (which is free for students).

R syntax confirmed horrible

Thanks a lot for the advices!

What kind of academic research?

I may do poker research in the future, from judgement and decision making perspective, psych-oriented but may tough a bit on behavioral economics (not the hardcore maths stuffs tho). Maybe sth related to risk but I'm not sure yet.

Both R and MATLAB toolboxes provide great options for neuronal networking, psychophysics/behavioural analytics research and genomic and bioinformatics work.

If you would like to pursue decision making research with a focus on computational neuroscience/psychology R is perfect for such analysis. Use stack overflow, GitHub for information hacking.

If you would like to design experimental paradigm MATLAB (psych toolbox) and
PsychoPy are initial platforms to use.

Of course PsychoPy and R are open source with great developer communities so I personally prefer using them over MATLAB. However in academic research (I work on cognitive neuroscience) avoiding MATLAB completely is not possible.

ps: I have used SPSS extensively and would suggest using R over SPSS any day.

I actually like R with regards to these kind of exploratory analyses. Scikit-learn in Python, for example, is much more about "create model, save model for future" than for generating insight into the data (and I hate the statsmodels package.

It'd be fun to take standard spots leading up to a street and try to calculate the exploitative optimal play against generic (and specific) opponents for sure. Obviously when I say generic exploit, I mean like "how does the average NL2 guy play?"

How does one define generic play as in the play by the 'average NL2 guy'? Unless one does a ton of data mining to define the generic model? Also may be calculate exploitative optimal play between specific opponent and one that plays the mathematically optimal game (obv. the latter is a hypothetical)? calculating how much a specific player deviates from optimal play (rather than the average NL2 play) might give more stable information.

Thank you. I'm more into Kahneman & Tversky framework. I may look into stuffs like N-armed bandit though. So mainly experimental psych. Yes I have tried psychopy b4 and will explore further.

Do u work on any poker related research? Think signal detection theory is quite related to poker. I only hv brief knowledge of cognitive computational neuroscience though.

The software mentioned such as R and SPSS are great for data analysis and more.

To expand the subject matter somewhat, if you are interested in modeling then, as an old-timer, I rely almost exclusively on Excel. Along with its VBA programming language you can do just about anything other modeling programs can do perhaps not as efficiently but I find good enough. I have written dozens of Excel poker analysis programs, some of which includes fairly complex simulations. These range from finding critical equity values to insure +EV or value to bluff ratios for indifference, to determining the bankroll that avoids ROR to simulating various hold’em situations.

Since so many of us use Excel for other purposes and many have had some background in BASIC, the reduced learning requirement may be an edge this program has.

I might start with R, because it is HUGE and it is a decent IDE for C++ (using Linux) though I haven't used the C++/R integration they propose yet. I have frames between the file system, code, and terminal. The comfort of being able to use C++ safely and soundly though is comforting. It seems like there are MANY paths to do the same kinds of tasks including R, MatLab, TensorFlow, ... C/CUDA

The biggest problem with this is that C++ uses the (much more "cache friendly") "row-major" ordering for matrices, whereas R uses the (more familiar to mathematicians) "column-major" ordering for matrices.

This means you end up writing lots of confusing/ugly/inefficient code constantly transposing matrices in/out of the different forms...

Juk