The SCII GOAT: A statistical Evaluation - Page 9

On June 21 2025 15:05 PremoBeats wrote:


@WombaT: I think Maru now won 18 and Serral 17 when locked tournaments are excluded, although both were tied in Premier Tournament wins at the time of the article. But context here is important: Maru won 10 of these tournaments (all his GSLs, StarsWar 11, Africa GSL) without Serral as well as mostly Clem and Reynor participating.

On June 22 2025 04:38 JJH777 wrote:


That's about as relevant as the fact that most of Serral's wins are missing at least half the top Koreans.

On June 22 2025 15:29 ejozl wrote:
Isn't tournament % and efficiency the same thing?, similar to win rate and aligulac rating. It seems strange why win % should count for 36%. Why not 5% aligu/5% win rate, 5% tournament%/5% efficiency, 50% tournament score, and 10% for the other results or thereabouts.

On June 23 2025 23:45 ejozl wrote:
I'm interested about efficiency for the longer career players if they're not punished by their long careers. Say, if you took rogue, serral, maru and inno and looked at their best period of a scope that's as long as life's scope. It's not entirely fair because life literally starts out his career royal roading, which the other players didn't, but would they be able to eclipse life in his best catagory?

On June 23 2025 23:45 ejozl wrote:
I'm interested about efficiency for the longer career players if they're not punished by their long careers. Say, if you took rogue, serral, maru and inno and looked at their best period of a scope that's as long as life's scope. It's not entirely fair because life literally starts out his career royal roading, which the other players didn't, but would they be able to eclipse life in his best catagory?

On June 24 2025 07:13 Mizenhauer wrote:


This is a poor argument. Other than Rogue and Maru (with a bit of Inno) all of these players played in tournaments the others didn't or played in said events when the format was different. Mvp played when GSLs were run almost monthly. Rain never got a chance to play with offline events with foreigners as good as serral, reynor etc. The Maru/Rogue/Inno/ Rain group never played in Code S during Mvp's era and Serral never played in Code S at all.

There would need to be some heavy adjustments to get this metric to a reasonable point, but I tried to avoid that/stay away from subjective era based adjustments.

On June 24 2025 11:06 rwala wrote:
The original analysis had many flaws and this analysis reproduces most of them (though somewhat amusingly claims all the flaws were all addressed). The author tries to maintain a veneer of objectivity and neutrality, but the bias inevitably reveals itself, often without the author being aware that it does. You can read the section "was the prime era harder?" and learn pretty much all you need to know about the author's perspective. Real quick on the main flaws tho:

1) This is not a statistical analysis. I can't see the results of any regression, monte carlo simulation, or any other modeling technique. I don't see any p-values or analysis of statistical significance. If you're not a numbers person, do not be fooled by the author slapping the label "statistical evaluation" on this.
2) If this is not a statistical analysis, what is it? This is a heavily-biased calculation designed to demonstrate that the author's favorite player is the GOAT. The author will say that he bent over backwards to nerf Serral and he is still the GOAT by nearly 2X points in his extremely objective and neutral GOAT calculator. Don't believe it. It all comes down to the fact that that the author chose to include specific criteria and specific weightings for those criteria that were designed to crown Serral the GOAT. I think everyone knows this but it's worth making the implicit explicit.
3) Many people pointed out that the Serral = GOAT calculator rated Rogue ridiculously low so clearly something must be wrong. The author's responses are interestingly defensive in how they justify Rogue's low rating, but many of us told him this would happen once he started ranking multiple players with the calculator. It's worth noting that while the calculator diminished Rogue to such an unbelievably low ranking, it's actually worse than simply rating Rogue lower than a guy like Rain or at a third of Maru's GOAT ranking. If you ran this calculator on all pros I suspect Rogue may not even be in the top 10 or top 20 GOAT list. That's how flawed the calculator is. Very good chance of even more bizarre results tbh.
4) Even if this were a proper statistical analysis, statistics do not explain GOATs. If they did, Shane Battier might be the NBA GOAT (seriously, look into it), and Muhammed Ali could not be the boxing GOAT (his 5 losses are simply too high for any objective weighted calculation to accomodate).

This point about Shane Battier and Muhammed Ali is the most important point, abstracted as follows: math equations do not understand greatness. They will always "undervalue" a player like Rogue in any GOAT convo because it does not know how to evaluate, for example, Rogue's jaw-dropping offline Bo7 grand finals performance record, which is among the greatest achievements in the history of e-sports. And they will always overvalue efficient and effective players that nonetheless seem to lack some of the incalculable qualities of greatness such as being clutch, resilient, or a great ambassador or leader for the game. Or even simply being lucky enough to be the right person, in the right place, at the right time to make history. With players like this, even when they seem to get lucky, it feels like a byproduct of their hard work.

Serral is indeed a great GOAT pick, just not for silly reasons like being high in Aligulac rankings for a long time. He's a great GOAT pick because he achieved what no one outside of South Korea was able to achieve, and is arguably the best player to ever play the game. At the same time, he simply never played in the most competitive tournaments and leagues, nor did he play in the most competitive era when over a thousand pros were spending 12+ hours a day trying to win premier tournaments and be the best player in the world. Lots of GOAT contenders achieved significant results in the toughest tournaments and in the competitive era, and if any of those guys are your GOAT, don't let any calculator say otherwise!

On June 24 2025 20:08 Mizenhauer wrote:


During one of my talks with Wax I asked if he thought the fact that soO was so high up was because he's my favorite player.

He told me that soO's ranking was consistent with my methodology, but that any person who makes a list like this probably creates a set of metrics that benefit their favorite player.

Something to ponder...

On June 24 2025 01:59 PremoBeats wrote:


Efficiency in winning tournaments or in achieving tournament score?

The problem is if you look at roughly 3 years (Life's career span), player 1 could have two phenomenal years, then one subpar year, followed by two phenomenal years and another subpar year, another two phenomenal years and one subpar year. Meaning 6 out of 9 phenomenal years. Player 2 could have had 3 consecutive phenomenal years followed by 6 subpar ones, meaning 3 out of 9.
Looking only at 3 consecutive good years, player 2 will outperform, despite his overall career being less impressive.

But as WombaT said: Serral outperforms here as well. Without looking perfectly at time spans, as I made the calculations on a yearly basis, these are the best results for 3 years for tournament win%:
Serral
- 2018-2020 (only non-locked): 37.30%
- 2022-2024 (only non-locked): 52.62%
Maru
- 2018-2020: 25.18%
- 2022-2024: 23.92%
Rogue
- 2017-2019: 22.75%
INnoVation
- 2015-2017: 34.09%
Mvp
- 2011-2013: 26.57%
Rain
- 2013-2015: 12.23%

If we only look at Serral's prime years, he sits at a win rate of 38.10% from 2018-2024. His lifetime win rate is better than even 2 year time periods of any other player and only Maru, Rogue, INnoVation, Mvp and Life in 1 year are above Serral's 2018-2024 result. As I said in the article... in such a volatile game as SCII, this is absolutely insane.

On June 24 2025 14:34 PremoBeats wrote:


I simply made it cause ejozl asked for it and it was kind of easy to collect the data from my set. In the original analysis I applied era-mulitpliers that seemed somewhat reasonable to me. I pointed out other flaws of this methodology too.

This small analsyis shows something pretty clearly nevertheless: Serral is way ahead of other players from his time (Maru, Rogue and to a lesser degree the overlap with a pretty good but not prime INnoVation). We could only look at tournaments, where all of them played as well to make it even more precise, as you mentioned though.



1. The term statistical analysis can mean a lot of things. In sports it simply refers to a structured evaluation of empirical data using numeric methods, not necessarily inferential statistics (Elo systems, composite indices like FIFA or ATP or even sabermetrics-derived rankins in baseball). Calling this analysis statistical is perfectly valid, as it is a descriptive, statistical framework, not a regression or simulation similar to Aligulac. It is about ordering based on performance data, not modeling outcomes. If you want a Monte Carlo simulation and significance levels you're looking for something different. I simply used statistical tools that seemed necessary (data collection and organization, analyzing, using normalization to a common 0-100 scale, interpreting).
2. Well. Where exactly do you see this suppossed bias or design tailored towards Serral? I am completely transparent with my methodology and even referenced external ideas like the ones from Miz' list. I laid out all weightings and calculations transparently, so you are free to criticize the exact ones, where it turned sour in your opinion. As a matter of fact, you made the same accusation in this very thread before (me not addressing past criticism). But I incorporated the critiques from last time:
- Including Rain, Life, Mvp
- Including team events
- Including a complete analysis of the deterioration argument.
I simply did not include your critique of "all statistical analyses are flawed", as I still think that these analyses are important. Well, I kind of did, as I explain in the article why I think my methodology is fine.
Serral's scores have been lowered and the most impactful decisions I made were against him. If that's bias, it's an awfully stubborn one.
3. What about pointing out that Rogue was inconsistent is defensive? Wouldn't building a model to reward Bo7 offline GSLs do exactly what you accuse me of doing under #2?
I even talked to Miz in this very thread and via DM and a quick reverse-engineering of his list made Rogue come much better (although Serral still is way in all different calculations and I still don't understand how he was a close 2nd in the first list and a close 1st after a couple of months and Mvp is below INnoVation in my attempt to reverse-engineer). The issue mostly was the weighting at the end, which I conceded many times was bad, as well as the era-multiplier, which Miz did not use.
4. Anecdotes and nostalgia are fine too. But as neither numbers tell the whole story, nor do such subjective takes.
In my opinion, as I already said in this thread, subjective takes might shift a GOAT comparison where the end result is close. But no single SCII player has a "better story" than others and in my opinion no "story" can rival Serral's numbers.

And I completely agree with your last sentence, as point 9 of my statistical evaluation made pretty clear...
"Because different weightings yield different outcomes, it's possible to justify alternative GOATs - if one prioritizes specific metrics disproportionately. So if you value...

… efficiency above all else, your GOAT is Life.
… career duration and sheer persistence, your GOAT is Maru.
… accomplishments in the prime era, your GOAT might be INnoVation.
… winning GSLs (or your name is Artosis), your GOAT is Rogue.
… any subjective skill, personal charisma, or emotional weight, your GOAT is whoever resonates with you the most."