And the fact the data pool sizes are so small they are insignificant to the overall game statistics as a whole. Least of all are those statistics even correct on Overbuff.
Otherwise, Iâm a GM Sombra main one trick, whom is better than 94% of the worldâs population of Sombra Mains⌠All because people say data pool size doesnât matter.
I struggle with the fact you donât either.
the only problem with overbuff is they âoverbuffizeâ certain stats, i.e just make up their own measurement that are not really 1:1 with how the game measures.
would be better if they stuck to elims/damage per 10 etc.
Just tell me why sample size is not applicable to Overbuff?
Overbuff is good as a general baseline but not at all definitive, mostly for those things you mentioned.
If a hero that has lets say 15% PR and 53% changes numbers with the meta to 14% PR and 51% WR its not really that significant and can be dismissed, even the data gathering being skewed doesnt realliy say much.
If that hero goes to from 15 to 2% PR and 57 to 40% WR then yeah, something is happening. It might not be brutal and huge and super drastic but you cant deny such overall decrease.
I donât believe so. Essentially, weâre looking at statistics when using over buff. You donât need to get data from literally every person to get accurate information for statistics to function. As long as there is a sufficient sample size and the data isnât being skewed (which can happen in say polling depending on how questions are asked for example), the data should be accurate.
A great question.
As for my thoughts on Overbuff, I am skeptical of it atm due to some weird outlyers it has had since the inception of 2-2-2. Like Ashe having 125% win rate for a bit. I think itâs been fixed and is probably okay now, but letting it settle and watch for trends.
Edit: This should also be noted, overbuff shouldnât be the sole basis for an argument. I think it can be useful to help support an argument, but using it alone isnât great as it doesnât explain why certain statistics and hero performance are the way they currently are. It just gives the data.
Most times people dont say overbuff stats as 100% true, but use as info on things we dont have any other kind of info on, such as pickrates, and winrates
Slightly. When profiles became private, there were no other significant changes on the server. Stat changes were around 1% or less, and usually leaned towards going higher. Which does make sense since people that put their stats up are generally actively trying to track their gameplay performance (i.e.theyâre trying to improve).
In other words, if Sombraâs numbers look bad on Overbuff, itâs worse on blizzards inhouse tracker.
They donât update that because it practically doesnât matter.
It has a good enough sample size with or without private profiles.
Other people have answered the question about statistics adequately and why you donât need 100% of a population to get accurate numbers. I want to address the second question.
Itâs a known issue they havenât addressed yet. They mentioned it in their Season 18 Role Queue blog post. Itâs at the very bottom.
KNOWN ISSUES
- Offense and Defense roles are still listed separately in some areas instead of being listed as Damage.
- Symmetra is listed as Support.
https://www.overbuff.com/blog/2019-09-04-season-18-role-queue-profile-changes
Thereâs not much reliable stats yet as itâs only the week⌠but it is decently accurate by the month when itâs not bugging out, and it seems to be okay for season 18 so far. Time will tell.
Welp, to generate higher pick rate for supports? i dont know, just seem like someone decide to not change that for some wierd reason.
Hard to tell. It strongly depends on how public and private profile players differ in behavior. If for example private profiles are predominantly one-tricks while public profiles arenât, it can have a very strong impact. However, itâs the best weâve got and using some statistic is better than just pulling one out your behind.
That being said, what keeps annoying me are all those people who only use GM stats of the last week. To have a representative and less biased sample size, you need to look at longer timespans first and foremost and preferably also at more ranks than the least populated oneâŚ
Eh, no idea, but it doesnât really matter much for hero to hero statistics.
Itâs still pretty unreliable. It shows Sigma as having a 16% pickrate when in reality thatâs impossible
The maximum pick rate would be 16.67%. Overbuff is telling us that basically no Grandmaster team is willing to play without a Sigma. This means that Sigma is way OP if true.
Why would that be impossible?
Because only 1% of the playerbase seems to matter eyeroll
That actually sounds about right.
Small interesting sidenote:
The quick ratings of players on OB is actually mostly decided by your competitive SR and not your QP performance.
I wonder if sites like OB periodically poll and refresh data for tracked players automatically? If not that would make even a sample set sort of useless.
It would be nice if Blizzard had an official way to do this. It is kind of fun to see how youâre doing compared to other players.
The sample size of people with open profiles is more than enough to support it as a sampling tool. There is bound to be some error, but its probably nowhere near as significant as people think.
Overbuff is kind of lazy about such things as Sym being in support.