📰 “According to Overbuff..”

No. People will tell you otherwise, but in statistics, you need only a 10% sample of a population to make an accurate prediction with a 95% confidence interval. The amount of public profiles are higher than 10%, so it’s probably 99% accurate.

That said, no one on the forums knows how to use statistics, so they just use whatever data supports their claim, and bends/twists the numbers to make stupid interpretations based on whatever their current reality is. This is why anyone spouting overbuff statistics isn’t to be taken seriously - they generally have no idea what they’re talking about.

1 Like

It’s important to remember that a relatively small sample group is required for statistical accuracy - only a few thousand players’ stats need to be recorded in order for the information to be considered reliable.

Having said that, it’s always better to look at trends rather than the hard numbers themselves.

but I think thing the over all number of accounts and players has just tanked, making the issues of stat bias even worse. Like early OG heroes, before private profiles, will have QP rankings out of 300k+ sort of numbers. Moria? 175k, Baptiste? It’s out 12k.
So right there… are we looking at stats for 300k+ Rein accounts vs only 12k Baptiste accounts? So one Baptiste player’s stats weighted to the tune of 25 Rein players?

I really do think private profiles just wrecked the 3rd party sites and it was, tinfoil hat on, totally deliberate on the part of Blizzard to mask balance and match making issues.

I mean sure it’s all we have so there is some merit to it but I really have a hard time trusting the numbers anymore. It’s just an easy way to check “Oh my over all death rate is down, good.”

Polling data is garbage and only reflects the reality of a miniscule percentage which cannot be extrapolated onto the massive populace.

Any attempt to extrapolate data in such a way will always end in failure with 100% certainty.

But is the sample size selected from a uniform distribution? That is just as important if we are treating statistics of the sample to be an estimate of the statistics of the whole.

I’d like to see some data on the distribution of private profiles if any exists.

1 Like

None of the stats on Overbuff are important anyway. The only reason to use it is if you don’t want to or don’t know enough about the game to understand the meta.

Overbuff was always opt-in. And in order to make it function, you have to have a public profile. So there is an actual incentive for people to public their profile provided by Overbuff.

The sample bias is largely just people who care about stats.

People also think they number 1 for that said hero thanks to overbuff which is silly.

And yeah, Junkrat, Torb, Symmetra, Bastion etc are all still considered Defense and Support. Makes zero sense what so ever.

Still more valid and objective than all the baseless anecdotal “evidence” which has been circulating the Forums forever…

2 Likes

Well, even if overbuff isn’t every single player, it’s still a sample size which you can consider. It won’t be 100% accurate, maybe all the sym players have private profiles and her pickrate is actually 17% but overbuff shows like 7%.

I don’t think it should be used to back an argument, I think it can be useful information to add to an argument.

I remember over a year ago there was a plat Lucio player who posted a thread with one of those stat sites that said he actually should be a top500 player.

And that was the downfall of people trusting third party sites. Seems their making a return though.

The sample size on Overbuff is still large enough to be cited as credible. It may not be a truly random sample size, but it’s still good data.

1 Like

According to Overbuff Im a super torb (Or rather was). Git Gut.

PS Overbuff is simply inaccurate, thats it.

So can you inform me of what Overbuff says the damage increase was from Symmetra pre-bugfix compared to after? Because Blizzard said it was a 40% damage increase, which is why she was nerfed so quickly. Not because she’s hated, as many Sym threads wanted to say.

If Overbuff shows a similar increase in her damage output, I’ll believe it to be more accurate. But if the data you find does not support what Blizzard says, I will continue to believe that those who rely on Overbuff are simply grasping at straws.

Yes, and no. Typically those who use overbuff want to compare their stats to other individuals. They’re the people who will want to improve based on their own performance. Naturally that will curve the stats a little higher than what the average will be compared to new players or accounts who don’t turn off private profiles. With that being said however, you also need to consider that is still a highly competitive, and large proportion of the player base, with a good amount of the casual player base thrown into it as well. The sample size from that alone would typically level out the numbers, not to be truly 100% accurate, but still close enough to give you a pretty great estimate of the standings in what ever stats you’re looking for.

So yes, overbuff is skewed, but it is still more than reliable regardless of what people would have you believe because the numbers go against their arguments.

Same reason they still have the defense category. Old layout that never got updated. It really doesn’t hurt any stats but pick rate, and only when sorting in those categories. The overall pick rate across every hero is still highly accurate.

It’s not impossible. Mercy during moth showed us exactly why with her 17%. The pickrate percent is determined by how likely it is to have that hero on any given team. There are six slots on a team that hero can be in. So, a hero who is a must pick, will be on every single team, with a pickrate of 16.67. Basically, they’re guaranteed to be in one of those 6 spots on every team, on both sides.

Now if somebody on that team switches off that hero, and somebody on the exact same team then switches onto that hero… that hero has taken up two slots on a team through-out the match. People were doing exactly this with Mercy during moth. Now that same hero has a higher than 100% chance to be on any given team, given they were on that same team twice.

So, no. it’s not impossible for a hero to be 16%. And it’s entirely possible for them to go over 16%.

Yea that’s safe to assume. Especially without knowing the exact inner workings of the system, player-to-player comparisons are biased towards the thousands of now unused accounts. Personally I never trusted any of those to begin with because the most accurate system we could want is already in game, anything else that says “you belong in a different rank” is just going off of less data using what is probably a less sophisticated algorithm. (Not saying the in-game MMR is perfect, just better than what a third party site in it just for the ad revenue is capable of.)

But hero statistics and pickrates are different. Particularly on Overbuff, it only looks at matches from recent history. Meaning if you haven’t played for months, you won’t bias today’s stats. And if you only come in and play a few games, your influence is only as strong as those few games, its not like anything spikes as it takes year old data and suddenly incorperates it into today’s matches.

Unfortunately, Overbuff doesn’t keep a consistent log of stats like Damage, similar to how they do with pickrates. Because of that, I have no reference starting point to compare today’s stats with. Now, that’s not to say that the site hasn’t shown an increase, just that I can’t directly prove it since I don’t know of a record from that time.

Now, remember that does NOT prove the site is or isn’t reliable in other areas, just that we can’t use it to determine this one specific data point.

What I can say though is the few times where the devs have referenced things such as pickrates, they have matched what Overbuff said almost exactly. So as far as that goes, there’s little reason to doubt it beyond baseless conspiricy.

1 Like

Because only people interested enough to look up how well they fare against other people are in the sample group. This means people serious enough to care and know about it, which means the whole “casual” player base is not represented in the sample group.

The average stats of a casual plat Soldier is the same as the tryhard soldier, otherwise they wouldn’t be the same rank.

Not necessarily, since there are more characters than Soldier. Maybe the ones that uses Overbuff are better and worse with other characters than the average population. They would still have the same rank but they would not have the same stats on the different characters. Maybe for example the Overbuffers have high win rate with a niche character while the rest do not. Or that certain characters are more popular among the Overbuffers than the casuals. Or higher and lower winrates on all characters. I am not saying that all this is more likely than not. All I am saying is that we can not know for sure. Therefore we should always take these statistics with a grain of salt.

You should study statistics.