My Flop Analysis Spreadsheet and a Slight Research Focus Pivot

This post was originally published on April 24, 2020, on my personal website, Lukich.io. I have since consolidated all of my poker-related content by reposting it onto Solver School. Please note that the analysis spreadsheet discussed in this article has been retired and is no longer publicly available. If you are interested in discussing custom analysis or ways to replicate for your own work, email me at [email protected].

I haven’t written much over the past week. This isn’t because I haven’t been working. Instead, I’ve been spending a ton of time in Excel developing an analysis tool to help inform my flop strategy. I realize that working in spreadsheets full of data might sound scary to some, but I feel right at home when doing so.

In my previous posts, I have referenced the large data set I built consisting of solver data from a 184-flop subset over 32 formations. This data forms the foundation of this spreadsheet, ultimately helping me look at equilibrium flop frequencies in many different ways (by formation, by flop type, etc.).

Since I started this website, I’ve been asked by a few people about how to study data effectively to glean any insights. I’m sure many will have different approaches, but building an organizational structure to look at metrics efficiently is critical. At least, it is for me. To take this formulaic approach, I plan to utilize this workbook.

Working in thousands or millions of rows of data can be daunting unless you know how to aggregate it effectively. In my day job analyzing marketing data, a common tactic to do this is segmentation or a way to group things (most often customers) together. Understanding the data broken down by different customer segments helps to inform business strategy.

Even though poker data is very different from marketing data, some tactics we can analyze are similar. I can learn a lot from this data. I must understand how to group it best to find key insights. When doing so, I tend to start at the highest macro level and move on to more granularity. Sometimes, that can lead me down different rabbit holes, feeling a bit like Alice in Wonderland. However, the structure I develop upfront will keep me on track as I progress.

As I mentioned, this workbook will serve as the basis for my analysis and ultimately guide the development of my flop strategy. It can also be a reference or a jumping-off point for future work.

Today, I thought I’d share this part of the process. Instead of doing it through screenshots, I created a video showing how I organized the data. Enjoy!

On another note, I have had somewhat of a realization the past couple of days. The data I’m using for this analysis is based on solves geared towards the live environment. I have developed 9-handed ranges and am exploring flop play when 200BB deep. While I am, first and foremost, a live player, I’m also a realist and don’t expect to step into a card room much over the next year or so.

Just because I don’t prefer playing online doesn’t mean I won’t choose to do so — especially given that it’s my only option. I do realize the games are a bit different from one another. All of the cash tables I play online are 6-handed with 100BB buy-ins. While the same game concepts apply to both live and online, there’s a ton of nuance between the two formats.

To maximize the impact of my study time on my results, I’ve decided to shift my current focus to reflect the online game better. That means updating some of the data in the spreadsheet in the video above to be more online-specific. Given the current circumstances, I think this will help me to realize the greatest current benefit from my ongoing work.

My study process will be consistent, and my plan for exploration in the upcoming months remains unchanged. The workbook I demonstrated above will come in very handy in doing so. What will change is the underlying data.

Fortunately, I’ve built this workbook with flexibility in mind. I can input any PioSolver-generated data set and analyze it similarly since the front-end UI only references the source data. I’ve started solving the 184 flops again using a variety of formations to rebuild an online-specific data set. When that’s done, I can input it into this workbook, and the entire front end will refresh.

I started the process of running solves this week with a few modifications to the solver configurations:

• I used 100 BB starting stacks to model an online game

• I decreased the preflop raise size to 2.5 big blinds to represent the smaller opening sizes

• I adjusted ranges for both me and the opponent

The first two changes are simple adjustments to some of the solver inputs. The last one is a bit more complex and requires me to consider different formation interactions. There are some key differences in these formations versus those I built in my live data set.

In my original analysis, I grouped some positions, simplifying them into early, mid, and late position. With the reduction in table size of 9 to 6, I didn’t think that made much sense in an online analysis. As a result, I chose to include all positions separately for single-raised pot situations.

I also decided to study a data set more closely modeling equilibrium (or at least an estimate of what we believe it to be). Instead of using any of my or others’ ranges, I have used a 6-max preflop solution for 100BB. I understand that preflop solutions do not perfectly represent equilibrium. I also understand that this won’t model the online game accurately. That has never been my goal. Instead, I believe this approach will help me understand things conceptually and develop the right baseline from which I should anchor myself when I choose to deviate based on the game or situation.

The good news is that when using these ranges, I can cut the number of solves my computer has to run in half. Because I’m utilizing a single equilibrium set of ranges, I can look at each solution from either player’s perspective, helping me complete two formations with each script. For example, when I solve for EP vs BB single-raised pot formations, I am also creating solve data for BB defense strategies against the EP.

In the meantime, I plan to advance using my existing data set. My current focus explores various dimensionality and how the data looks when aggregated at those levels. I plan to post on that topic within the next week or two. But shortly, I’ll shift over to using the online database as the foundation for my work.

If you have any comments or thoughts, please feel free to leave any comments below. You can also contact me at [email protected] or on Twitter or YouTube through the links in the footer below.

Thanks for reading.

-Lukich