So I’ve seen Overbuff stats touted many times to explain why a character needs (or shouldn’t get) nerfs or buffs. I’ve also seen many people say Overbuff stats aren’t very accurate, but “they’re the only metric we have available.”
So what is the general thoughts on the issue? Does Overbuff completely ignore open que competitive? And with private profiles being very common there’s a lot of missing information within those stats, so how effective are they? Are there any other stats that are more reliable?
I think there’s more to pick rates vs win rates that should be involved with the discussion of balance, but it feels like a lot of the community boils it down to those 2 talking points. What other issues would you bring to the discussion of changing a character in the name of balance?
Statistically, Overbuff appears to have a sufficient sample size to give a data you can draw reasonable conclusions from; however, they have a notable presentation issue which limits data value.
It’s been quite some time and they still present Damage as two separate categories: offense and defense. This can limit the ability to compare and contrast in role data. Additionally, Symmetra is still listed as a Support, further complicating data analysis in the Damage and Support roles. Tank appears clear and useable.
Adding to this is the recent problem with QP data, and their use is dubious in many areas without significant improvement in presentation.
You only need a small sample of data to get a general idea of things. Overbuff is reliable enough to use, I’d rather the devs release actual data tho regularly.
The problem lies with the community.
Many bold claims are based on one dimensional thinking and are very flawed. Stats, PR and WR are always influenced by everybody else
Overbuff is accurate for pickrates and winrates when compared across all heroes. I personally wouldn’t use it for specific stat tracking, but at the very least whenever the devs have released information about characters Overbuff has matched it with a 1% margin of error.
It’s just an arbitrary grouping of these heroes with these other heroes. Of the 7 “offense” heroes, this one is picked the most, followed by this one, and then this one. The arbitrary grouping may be wrong, but the data that goes into it is still correct. People that look only at “offense” hero pickrates don’t usually take into account that its comparing the offense heroes against one another, not against the whole cast.
I’ve seen things like 17% PR ana… but ig it’s trying to compensate for private profiles.
If it was only showing public profiles, I’d be against it, but if it’s making assumptions from the private profiles, I’ll trust that OW has a large enough sample of open profiles.
I used to check Overbuff quite often when the game was released and no profiles were private. If these stats are accurate and the sample size is reliable then the difference between my stats pre private profile and post private profile should remain the same.
That wasn’t the case. My personal stats were off by 10% of what they used to be when the profiles went private. A 10% deviation didn’t look large at first. I did some quick math and for my profile to have deviated 10% a very large portion of the player count would have to have swapped to a private profile. More then Overbuff was willing to admit to.
I am not sure how bad the statistical deviation is for the joint information of all heroes. If you are attempting to use it as a metric to value yourself against other, you may as well just write down the information you want to see, and guess where you are among many.
Private profiles pretty much ruined Overbuff. It’s only good for trends now. Being certain roles or heroes come under more community fire and scorn than others.
So there is a MUCH higher motive for say Brig main or Mercy main to mask their profile. Players who mostly like QP but get crud for being low rank have more motive to mask their profile.
People with losing records and one tricks have way more motive to mask their profiles.
it’s bias all over the place…
huge stat breaking bias. The only people who don’t seem to understand this are those who cling to the generic statistical line of “Well as long as there is large enough sample size it’s fine” but that’s not true once you have a lot of bias.
It would be like trying to figure out the average height of men by way of an online dating site. As it doesn’t mater how large your sample size if every guy 5’10" or so is saying they are 6 feet tall.
The problem with Overbuff is that it a) gives general pick / win rates for heroes but no data on who they are playing with to get those rates. It could be that xx hero has the highest pick rate as they work well with a group of 4-5 other heroes that make it better. and b) it uses data that is hidden by default, in order for overbuff stats to exist the profile must be public and defaults to private. Especially at lower ranks where they may not even know the profiles exist they remain hidden removing a large chunk of that demographic from Overbuff’s statistics.
I trusted it reasonably up until private profiles and (likely untracked as it’s opt-in) alt accounts became so popular
I don’t really care about it that much lately, since way back when the devs dropped some pickrate stats about genji and soldier that didn’t quite match up with overbuff… I don’t remember how far off they were from overbuff but it was significant enough for me
Even then, it’s easy to lie or misrepresent a case in a seemingly correct way (even accidentally) so I’ve never cared much for overbuff stats in the first place. if anything I use it as a quick “what are people playing at a glance” and nothing more
I’d take what anyone here says with a grain of salt. I’ve had to explain multiple times over the course of a thread that having a hero with an incredibly high pickrate and winrate actually means that winrate is higher when adjusted for non-mirror matches. It’s not that I’m so smart and this is such a difficult concept so much as people don’t understand insanely basic concepts here.
I have no doubt that you can draw broad conclusions that pan out from Overbuff, but my experience is that most people here don’t know what the flip they’re talking about and use Overbuff to work backwards from their conclusion.