A terran winning 4 games in a row means nothing

[Trias-2243]

@ZtAbhEt-1813

PvT win-rates have been protoss favored for years over tens of thousands of games.

Even with a 99% PvT win-rate, a Terran will still win 4 games in a row (albeit rare).

A 55% PvT win-rate will yield a 4-0 for terran once in every 24 matches.

A single PvT match means nothing about balance compared to tens of thousands of games.

If a single match is the deciding metric of balance, then Protoss is simultaneously OP (when Zest beat Byun) and underpowered (when Ty beat Stats), therefore a single match cannot determine balance since it gives contradictory conclusions.

I think I can agree with most of that : a single 3-0 or 4-0 match cannot be used to draw balance conclusions by itself, as it’s too small to make for a sample size, and the players’ styles and skills might counter each other in strong way (independently from the MU). They can be hints to pinpoint specific points though, but for me assessing balance requires to take in account multiple factors, including winrates with a decent duration/sample size (such as Aligulac’s monthly ones), leagues representation, tournament victories, games analysis at different timings, etc. etc.

Nevertheless, while PvT might’ve been often slightly protoss favored, it was far from being an insuperable advantage. Particularly at high level, as TY demonstrated it in this example.

@ZtAbhEt-1813

Here is a fun one. Shonix blasts the creator of the thread for using a “non randomized” sample when he uses ladder stats to estimate the number of smurf accounts. Shonix then proceeds to say he, and other “high-level” players, can recognize smurfs and estimates they happen once in every 20 games, which is much less than the OP estimates.

In other words, he is saying an estimation on ladder data is wrong because it’s non-randomized, meanwhile his anecdotal estimations based on his own play are MUCH more accurate. He does this while claiming to have a Masters in statistics. I kid you not!

You didn’t answer the question. What was your method to confirm smurfs ? Pro tip for your source then : If P = 1/x is the probability for a regular player to meet a smurf Then the probability of two smurfs meeting isn’t P, but P x P, in other words (1/x) x (1/x) What you did there with the highly factual “I know 'cause I’m higher level” completely non standardized biased method, wasn’t in fact evaluating the probability of playing a smurf, but the probability of a smurf playing another s…

He gets absolutely taken to the cleaners by Trias, then starts spouting BS like this:

@Cheezecake-1895

You are entitled to your own opinion, but as someone who has coached hundreds of people I highly highly disagree, lower leagued players are ** very ** bad at trying to analyse their own games, so they cannot identify these “small” differences in skill at all

A word about this.

When I started this thread, the population estimates I initially did were mere observational reports. Cheezecake (Shonix ?), highlighted that those estimates weren’t randomized, and as such weren’t independent from me as a player. His own estimate however was not even randomized, but also non standardized in the least, which allowed me to defend my work in a verbal bout.

However, though I still think that his vision is often needlessly black and white, he had a point in saying that my data was imperfect. And, later on, in another thread and without recognizing each other at first, he made a valuable suggestion on a possible workaround. Which, with some additional research, allowed me to perform randomized surveys I consider of higher quality.

So, while his argumentation had flaws as a verbal bout, in terms of statistics, in this very example, I consider he had a point.

As for the statistical debate you two had (in that thread ?), what little I saw on the purely statistical plan seemed beyond my knowledge TBH. ^^