To me, I’d say that it is going to be fairly accurate. Exact, clearly no, but within an acceptable margin of error.
Now it’s been a quick minute since I’ve done any kind of statistic class, but I will say this. My company recently used a phone survey to ask about our company. They stopped at 400 results because that gave them statistically valid data within a 5% (I think) margin of error. Our population is approx. 300,000. So they only needed 400/300,000.
Thus, I’m assuming that overbuff pulls from a large enough pool of players that it should work out to be accurate enough for broad assessment of the information provided.
Just curious if I’m way off base or if it is more accurate than a lot of people seem to think.
13 Likes
It’s accurate unless it invalidates my opinion.
50 Likes
Depends on if you know the pop of overbuff results or not.
Correct. Statistics do not need the entire population as their sample size in order to be reliable or accurate.
24 Likes
Adequately reliable to make conclusions as long as you’re not looking at small-scale differences in data. Like, if you’re trying to compare a 4.17% pickrate to a 4.24% pickrate… okay, that might not be reliable. But like a 4.17% pickrate to a 3.67%? That’s pretty reliable. Keep in mind Overbuff still pulls from hundreds of thousands of accounts, so your margin of error is pretty small
Sort of and no. There are various factors that affect the values on the site that the site doesn’t even have or can even record.
1 Like
Assuming you’re not over-filtering the data (e.g. GM only for 1 week), it should be able to provide general indications of basic statistics.
There are caveats to that, though, in that gameplay doesn’t always map to the way things are represented on the site. For example, each character has a flat “pick rate” and “win rate” on the site, but many players play multiple characters per round/match, and there’s not really a way to determine which characters were most responsible for the outcome.
3 Likes
i don’t think so
apperantly im rank 248 on widowmaker and 252 on genji
yup not accurate at all
no, it’s completely accurate
you’re just one of the first players to play over an hour of those heroes in comp this season
1 Like
The stats being accurate or not doesn’t really matter since the stats themselves are often irrelevant.
i don’t think playtime is the reason because its clearly not just 250 widows and genjis that have played 1 hour so far
it is playtime
if you see your other heroes whom you dont have an hour or more on, it shows an ∞ symbol, meaning they need more data to rank you
also it doesnt factor in private profiles
ive been using this site for over a year now
1 Like
because ur a smurf ya dimwit
all your stats are basically top 1% so it calculates that you’re around there
Not since private profiles, no.
1 Like
Nope, not at all accurate. It only takes data from people who actively made their profile public. That in itself is a very biased sample.
1 Like
take it with a grain of salt, and look at all the context. stats alone dont give a full picture
1 Like
it’s most reliable thing we have.
It’s not time based but game based
10 games is a hero score. 10 games can be an hour or three. Depends on your matches.
(Quick play is apparently 30. Which now has to be wins since blizzard removed wr in qp)
1 Like
Its the only source we have
So reliable or not, we work with what we have
Overbuff is probably relatively accurate for general numbers like overall pickrates/winrates and average elims/accuracies and such (taking the right context into account, that is). However, when you start looking at the less-populated ranks like bronze or GM (which is a popular one to use in discussions) and/or shorter timespans (e.g. last week), the chances for skewed statistics become much bigger.
Also always take the numbers with a pinch of salt because the main reason that we don’t have full numbers is because of private profiles. Due to the nature of the system, private profiles are usually mains/one-tricks/etc, especially of the ‘throwpick’ heroes for example (i.e. the distribution of private profiles across the playerbase is not random), so it’s quite difficult to predict how big of an effect it has on the numbers.
Overbuff isn’t statistically accurate because of its sample size, but because the sampling isn’t random, which is an equally important requirement for good reliable statistics.