You’re assuming it’s all apple pie. What if the slice You’re eating is apple pie, but the rest of the pie is blueberry?
That’s the issue with overbuff. The small sliver you see (public profiles) is apple pie, the rest is blueberry (private profiles)
Also take the data with a grain of salt. Its not 100% accurate, but its within a margin of error
2 Likes
But you don’t know the ratio. For all you know, there could be 20,000 apple pies, and only 10,000 blueberry pies.
The issue with Overbuff isn’t private profiles. The data may not be 100% accurate, but it is data, and it is as close to accurate as we’re going to get without Blizzard sharing stats.
Overbuff must have a large enough sample size for it to be as close to accurate as it is, considering the heroes who got buffs were underperforming.
2 Likes
They it know it far better than you lol. You’re such a joke tbh.
Overbuff is a valid sample size, you do have to be careful with bias in data but open profile isnt a bias.
2 Likes
What does that mean? That is not a good way to phrase it for analysis.
They do breakdown ranks and also comp and QP.
1 Like
Unless you can someone prove there is a connection between weather a profile is public/prive and the “blueberries”.
Like “all blueberries have private profiles” then you would need be concerned about the accuracy of the data.
But as far as we have been shown, there is not correlation, meaning that while the data is incomplete its still accurate to a degree.
Think of it this way.
In space there are different types of stars, they get classified as different types depending on factors. Our star is a type G star.
Type G stars account for ~8% of stars, with Type M accounting for ~77% of all stars.
However we have only observed a tiny tiny percentage of all the stars in the universe (theoretically we have observed a infinity small % of stars if the universe is infinite)
How ever as we document more and more stars, adding to data we have, the ratio of stars remains similar. Type G are still 8% and Type M are still ~78%. Even after adding Billions of more stars to the data set, the ratios have stayed similar.
We have a very incomplete data set when looking at every star in the universe, but the data is still valid and continues to be valid and has yet to be proven wrong.
3 Likes
That is actually a frustration of mine. So many people think winrate tells you balance. But it is arguably the WORST stat to look at for general balance questions.
I think you are going down a wrong path here. I do get where you are going. But, the question is not what is a hero’s current win rate, or even what the win rate is at specific ranks and/or time periods.
Win rate is just a VERY bad stat to look at for balance indication. This is primarily because the matchmaker largely negates the impact of general hero power on win rate, leaving other variables to swing it far more.
Think about what happens if a hero is stronger than others. Does that hero just have a higher win rate? No. Because the people playing that hero get ranked up until the extra power from the hero can no longer make up for their lack of skill. And they end up with close to a 50% win rate. The reverse is also true. If a hero is weak, those players end up at lower ranks where their higher skill can make up for what the hero lacks. And you end up with around a 50% win rate.
That doesn’t mean win rate is useless, it does help identify what the issues are. Take Sym for example. She is weak. And not just this season, though this season didn’t help her (quite the opposite). She has historically had a high win rate though. Why? Well, simple. There are some places she has been really strong. They are few, but they have been there. So, in general, people only use her there. Overall she is weak, so people generally don’t use her. And only using her in the places she is strong inflates her win rate.
She is a good example of what a low pick rate, high win rate means. Generally weak, niche hero. Win rate is useful, but not as an indicator of power.
Occom’s razor
So, if you pull slices of pie out and see apple it is logical to conclude the rest of the pie is apple.
Extraordinary claims require extra ordinary evidence.
If you are going to make a claim like that, which goes against all logical reason, you need to provide evidence. You need to provide evidence at least as good as Overbuff if you are going to make that claim.
What you are saying is the equivalent of someone saying: ‘I know NASA took some samples of moon rock, but they didn’t check the whole moon. So, most of the moon is actually made of cheese.’
4 Likes
Tom and jerry told me the mood was made of cheese, you telling me they lied?
3 Likes
It’s more just an example of how people use the default data to draw a conclusion about heroes they specifically don’t like.
Because winrate without pickrate is a worthless stat, and there’s also just a lot of nuance behind why a hero might have a high winrate.
FYI blizzard does use win rate, pick rate and unmirroed win rate as key data points.
Anyone here who thinks they know the stats of ow better than Blizz is wrong. They have a team of analysts that know what they are doing better than anyone in here.
Why would they even care about QP? it’s not like you are competing here for anything, elo, sr or w/e it is they use now days. Seems like a waste of code / development efforts.
1 Like
All this talk about pie makes me hungry! 
2 Likes
Neither do you… it could also be 50,000 apple pies and 1 million blueberry pies…
Overbuff has always had issues not just limited to private profiles, and I don’t know how you can argue the sample size is large enough when multiple “overperforming” heroes and multiple “underperforming” heroes according to Overbuff weren’t touched.
Open profile is a bias in the same way that Type A personality and Type B personality lead to bias. People who choose to make their profiles public are a certain type of person, people who choose to make their profiles private (or just keep it private by default) are another type of person. You can’t say there’s “no bias” when there clearly is in the intentional choice of opening a profile publicly.
It’s not on me to prove if there is a difference between public/private profiles lol the difference is already there. It’s on you to prove that the two groups are the same if you want to say that the public profiles on Overbuff represent the entire playerbase accurately
The easy answer isn’t always the right one. Just because you “assume” the rest of the pie is the same doesn’t mean it is. If I were to show you a picture of New York and the picture was of Manhattan you’d “logically conclude” that New York is a highly populated metropolitan state, but a large portion of New York is the opposite of that.
Such as claiming that the small amount of public profiles Overbuff (a non-official website) is based on accurately represent the entire playerbase which is by majority private. Yes, that extraordinary claim requires extraordinary evidence which you haven’t provided.
I’m waiting for your evidence besides “we can assume”, that’s not evidence.
What you are saying is the equivalent of someone saying: ‘I’ve been to Italy, I know what Europe is like’
No… you know what that part of Italy is like, not the entirety of Europe. You wouldn’t even be able to say that you know what all of Italy is like, nevermind all of Europe.
Ok what is the difference? If its already there there it should be easy to point to the factor that makes the data invalid.
As for our proof, We dont see a pattern that would skew the data, there seems to be no rhyme or reason for why some people have it public vs private. We have people in all ranks, people who 1 trick, people who flex, we have a random assortment of everyone.
And since we see no pattern, its safe to assume there is no “blueberry” that skews the data
1 Like
Overbuff is not even an official tracker app.
Why use a misleading app that will only bring biased information?
Do you understand how overbuff works? It litterally takes the public OW data. For the data to be false, OW own profile data would also need to be false
1 Like
Do you know what does third party data means?
Do you understand the difference between Third-Party vs First-Party Data vs Zero-Party Data?
Do you know which is most valuable?
I work in a game studio and my day to day work consist on working on SDks.
Do you know what does SDKs means?
So, your reply lacked a lot of context.
Just because it comes from a 3-rd party doesnt make data false?
Overbuff gets its data from OW profiles, its litterally just a website that gathers ow profile data.
You can do the same thing manually, you could open a excel cheet, go through every profile, white down the data overwatch has, and then compile it and get the same result as overbuff.
Saying the data is false/wrong becuase its from a 3rd party is dumb. IF I say e=mc2, is that now wrong because it came from me? a third party?
1 Like
You make the assumption here that Blizzard is actually trying to balance the heroes. They aren’t.
There are numerous examples of them making ‘balance’ choices around spectating rather than fair game play. There are numerous instances of them intentionally making certain heroes OP in order to force a meta they want.
Plus there is the time Blizzard directly told us they often do not follow the stats and instead just make gut choices.
Blizzard not balancing over or under performing heroes does not mean the stats are wrong.
There is no evidence in any way that the lean towards public profiles effects how well someone plays, hero preference, etc. There is every reason to believe they have the same spread as the rest of the population.
If you want to claim a bias then you need to show evidence of how that group would differ from the general population.
No, you are the one claiming that the choice between public and private profiles actually effects hero picks, performance, etc. There is no logical justification for that. So, it is on you to provide evidence to back up your claim.
Extraordinary claims require extraordinary evidence.
The more unlikely the claim, the more evidence you need to back it up. And your claims are the unlikely ones.
No, I would conclude that New York had large cities in it.
Your claim is more like claiming that the picture of Manhattan is just a few small buildings with a giant cardboard backdrop.
Do you have any clue how statistics work? Like any at all? Basically all of them are samples.
One thing we learn from history is that when dealing with large groups of people/things/etc. there are some pretty consistent trends. So, if you take a sample you can get a pretty good idea of how the whole picture comes out. That is how statistics work.
No. That is what it is like when someone claims to know because of their games. Which is usually what people are doing when they claim Overbuff is wrong. Sample size does matter.
More comparable to Overbuff would be that someone visited a couple hundred cities in Europe across multiple countries and then saying that you had a pretty good idea of what European cities were like. Sure it is not every city in Europe, but it is going to give you a pretty good reflection of what cities are like.
3 Likes
Its so funny to see so many people prove OPs point: there are people who can read the data and know what it means and there are people who cant even understand the fundamentals of statistics and therefore are unable to get usefull data out of it.
2 Likes