Saturday, February 7, 2015

A Critique of the GPI Ranking System

The Global Poker Index, or GPI, is a leaderboard system that ranks the top live tournament poker players in the world according to their performance. CEO Alexander Dreyfus has stated that his goal with GPI is to define who the star players are in order to increase exposure, creative compelling narratives, and grow the industry as a whole. To this end, his efforts have already paid dividends. The rankings have not only been readily accepted in the culture and the poker media, but also in major mainstream outlets such as USA Today. However, the system as it stands today has flaws that threaten to undermine its credibility as comprehensive measure of tournament poker skill.

Recently, GPI has started to draw some criticism from players who say that the way it scores performance doesn't make sense. As a tournament player myself, I've shared this attitude, and I've done a bit of analysis on the issue.




Consider the following examples, where in two very different scenarios the points awarded by the GPI formula are almost exactly the same:

EXAMPLE #1
A) 1st place finish in $570 event with a field size of 5,000 (103.79)
B) 1st place finish in $1,090 event with a field size of 24 (103.56)

EXAMPLE #2
A) 1st place finish in $1,090 event with a field size of 1,200 (231.03)
B) 120th place finish in $10,500 event with a field size of 1,200 (231.50)

EXAMPLE #3
A) 25th place finish in $2,200 event with a field size of 1,250 (209.89)
B) 25th place finish in $10,500 event with a field size of 200 (209.07)

In each example, scenario A is arguably a more impressive, profitable, and skillful accomplishment than scenario B. So why does GPI reward both scenarios equally? Answer: it appears to place a much larger emphasis on buyin relative to other factors.

Not all ranking systems are like this. The Bluff rankings, run by Bluff magazine, takes into account the same three components as GPI: finishing position, buyin, and field size. The difference is how it uses them.

Consider the examples from before, this time using the Bluff formula:

EXAMPLE #1
A) 1st place finish in $570 event with a field size of 5,000 = 145 points
B) 1st place finish in $1,090 event with a field size of 24* = 0 points

*minimum field size for this buyin not met

EXAMPLE #2
A) 1st place finish in $1,090 event with a field size of 1,200 = 200 points
B) 120th place finish in $10,500 event with a field size of 1,200 = 20 points

EXAMPLE #3
A) 25th place finish in $2,200 event with a field size of 1,250 = 36 points
B) 25th place finish in $10,500 event with a field size of 200 = 11 points

These results are more in line with what any poker player intuitively recognizes–that in each example scenario A is a more skillful display poker than scenario B, and consequently more deserving of ranking points.

While poker rankings do need to account for a greater level of skill required for success in higher buyin tournaments, GPI rewards cashes in these events disproportionately. So much so that for the higher ranked players, otherwise remarkable results in lower buyin tournaments (especially <$1,500) are virtually guaranteed to have zero impact on that player's overall score. Not the case in the Bluff system.

The following graphs further illustrate the difference between the two systems (click for enlarged version):










Raw data for graphs:
Link

The value in a ranking system lies in its ability to appropriately reward performance as a reflection of skill level. While the GPI system has made solid strides towards providing the community with good tournament rankings, it does, in my opinion, have something to learn from its more balanced competitor, Bluff. Adjusting the GPI formula to decrease the influence of buyin relative to overall performance should yield a more robust and practical ranking system.