Welcome to the PGstats open beta!

We'd like to use cookies to help analyze and create a better experience for you.

You can find out more in our Privacy Policy and Terms of Service.

pgstats

Introducing SPR and UF

By Jack "Jackie Peanuts" Moore | 12/16/20

Today, we would like to introduce two new stats launching with the website: Seed Performance Rating (SPR) and Upset Factor (UF). These statistics leverage bracket math to put bracket runs and upsets on a simple and easy to understand scale. Both operate on the same principle. You can always describe somebody's placement in a double elimination bracket in terms of the number of rounds they were away from victory, like so:

As obvious examples, the winner of the tournament is always zero rounds from victory, and second place is always one round from victory. As more illustrative examples, 7th place is always 5 rounds from victory and 49th place is always 11 rounds from victory. That is to say, if you finished 49th, you would have needed to defeat 11 more players to win the tournament. 

SPR: Seed Performance Rating

This statistic measures a player's performance in a bracket relative to their seed. Using a couple of examples that we have already covered, consider a player who is seeded 49th, but makes some big upsets and runs it all the way to 7th place. Their seed, 49th, means they were expected to finish 11 rounds from victory. In reality, they finished only five rounds from victory. From here, SPR is a simple calculation: take the expected rounds from victory (11) minus the rounds from victory in reality (5) and you get the resulting SPR of +6. If somebody underperforms, for example placing ninth (six rounds from victory) when they were seeded to finish fourth (three rounds from victory), the SPR will become negative: expectation (3) minus reality (6) results in an SPR of -3.
This allows us to put players of wildly different seeds on the same scale. In the above example, the player 49th seed with the +6 SPR outperformed his seed by 42 places. Somebody who finished 97th but was seeded 139th, would also have outperformed their seed by 42 places. These runs should not be considered equal, though -- the 139th seed likely made just one upset to reach 97th, but the 49th seed may have made as many as five upsets to outplace his seed by the same amount. This is reflected in the difference in SPR: The 139th seed (expected 14 rounds from victory) who finishes 97th (actually 13 rounds from victory) has a mere +1 SPR. As such, using SPR and its reliance on rounds from victory rather than raw seeds gives us a better tool to compare players’ performances against expectations across a wider spectrum of skill levels.

UF: Upset Factor

The issues that can come up when using raw seed values to analyze a bracket become even more apparent when analyzing upsets. If I told you the 380th seed had just upset the 260th seed, the difference in raw seed numbers might suggest it was a huge deal. But in reality, both players were actually seeded to finish in the exact same round of bracket. The tie for 257th place extends all the way to 384th place. So even if this 120-seed jump sounds big enough to indicate a significant gap in player skill, both players were actually seeded on the same line. As such, it should be considered a low-grade upset at best, if it's considered one at all.
Consider a pair of real-life upsets that can help illustrate the value of UF. First, take Wizzrobe vs Tweek at Smash ‘N Splash 5. Wizzrobe made this massive upset as the 97th seed at the event, taking it over top-seeded Tweek. The 97th seed is expected to finish 13 rounds from victory. Much like SPR, the final calculation here is simple subtraction, with the top seed as always expected to finish 0 rounds from victory. As such the upset factor for this match is 13 - 0 = 13. That, by the way, remains the highest Upset Factor ever recorded at an S-Tier event.
Next, consider an upset from Evo 2019. Prodigy, the 33rd seed, fell to Japanese Falco player Seven, the 1712th seed at the event (an unfortunate victim of the typical practice of random seeding after a certain point, usually top 192 or 256 for an event the size of Evo). Simply looking at the raw seed difference, this upset is orders of magnitudes bigger than Wizzrobe’s over Tweek -- a difference of 1,689 seeds for Seven’s upset versus just 96 for Wizzrobe’s. 
This is where Upset Factor can help cut through the noise, because it simply is not true to say this was a “bigger upset” in terms of their impacts on their respective tournaments. The 33rd seed is expected to finish 10 rounds from victory, while the 1712th seed is expected to finish 21 rounds from victory, resulting in an upset factor of 11. We can calculate this with a mathematical function that's more complicated than it looks. Click here to learn exactly how it works, but as long as you can remember that both Upset Factor and Seed Performance Rating are all about differences in losers rounds from victory, you'll be good to go.
This result—Wizzrobe's upset over Tweek coming up as the biggest upset of Ultimate history despite the major upsets that have happened at Evo and Evo Japan due to the random seeding issues discussed above—is what convinced me of Upset Factor's value. Sending a top seed to losers bracket early has a cascading effect, leaving players on winners side with easier matchups and the unlucky suckers in the top seed's losers bracket path with the last thing they ever want to see. A double-digit seed losing to a quadruple-digit seed is certainly an event, but the carnage is restricted to just one section of the bracket.
Over the next few weeks, look forward to more content on PGstats.com that will leverage these statistics and show the value they can bring to tournament analysis for both Super Smash Bros. Ultimate and Super Smash Bros. Melee. I have already found these statistics useful in helping me find surprisingly strong performers, players on the rise, and major upsets I missed from early Ultimate tournaments. With this data now available for Melee as well, I'm looking forward to diving in and seeing what its rich history has to offer.
Follow PGstats on Twitter @PGstats for updates on new content and website developments!