Really? The 2 is important. How about you ask the blizzard devs directly how accurate is overbuff.
Probably, but nobody cares in an election about non-voters.
Like I said you are allowed to doubt overbuff.
Yeah, no. Whatever.
This is a game, everyone who plays and purchases from the store is a “voter”.
People who complain on the streets are not necessarily voters.
I only got involved in this thread because the thread creator, wanted to prove something by using numbers and calculations.
My main message is that if you don’t like a hero, say it’s your opinion and don’t try to sell it as fact.
There are 9 supports from which you can pick from and there are 4 support slots, but any hero can only be played once per team or twice total if both teams play the hero.
The chance that any particular support hero is picked on a team is 2/9. 2 players can choose from 9 supports.
The chance that any particular support hero is picked on any team at least once is 4/9.
By default, just looking at statistics, you will have any support hero in a match at least once 44% of the time if my math isnt off. Facing an Ana like every second match would still be in the norm.
Jesus christ, why is a thread about someone being painfully wrong about basic arithmetic almost 200 posts long?
2 Likes
I intended to hijack this thread.
What do you think about overbuff? Do you think their data is accurate now that every profile is private by default and there is no Overwatch API anymore?
Do you think the people who log into overbuff with their BNet accounts are in any way representative of the Overwatch player base? Or do you think the data that comes from overbuff is biased?
The data collected is from the Overwatch API which is in-game data.
Now the data set is of course incomplete but will and always will collerlate to complete the data set.
Even tho things have changed a bit it still correlates
In this they say…
On Overbuff she has the highest winrate for Silver,Gold,Plat,Diamond,Masters and GM
- So still true and matches
On overbuff Kirko and Moria are… still in the bottom half. Moria has improved tho and Kirko isn’t dead last because of Lifeweaver
- So still matches somewhat.
On overbuff Ana, Kirko and mercy are all still the most picked supports
- So matches
Now don’t get me wrong some things in the article don’t match Overbuff now but it is an old article and things have changed since like…
Isn’t true now but he has been nerfed and you’d expect his win rate to decrease based on those nerfs which is reflected in Overbuff with him being 2-4th highest win rate tank now.
I don’t think so, why would they need me to log in with my bnet account then?
Do you have any source for this! Please! The official Overwatch API doesn’t exist anymore.
Nope, but there is an unofficial one that does which is called overwatch api on github (this is your source as well). It uses data in profiles.
Yes, I looked into it, but it doesn’t give you the ability to search the match history of players or even players who have hidden profiles. It’s parsing the HTML page.
https://github.com/TeKrop/overfast-api
Maybe this has already been touched on. I’ll admit I didn’t read through 190 posts to find out but…
You can’t really apply statistical probability to something like an informed choice. Saying someone has a 4.xx% chance of picking Ana implies that the choice is randomized. Outside of Mystery Heroes (where we don’t know the formula for determining picks), there is no randomizing factor in whether or not someone picks Ana.
All you can calculate then is “Win Rate%” and “Pick Rate %” (IE, how often is she picked, and how often does that team win). Those numbers should be fairly easy to figure out for Blizzard and for users that have access to Pick/Win Rate information.
It’s not a matter of statistical probability.
2 Likes
If gives total wins and losses. If you take 2 different snap shots for a time frame which overbuff does…
For that time frame you’ll see win and loss. What hero you played and win and losses on those heroes.
I don’t think overbuff does anymore than that.
Yes but how many players have hidden profiles? I would say at least 50%, my estimate is more like 75% since it’s the default setting and players who perform bad want to hide it. And those 25% with public profiles? How many of them do you think sign up for Overbuff to start tracking stats?
Let’s say we have 10% of the player base tracking their stats. I think more than 50% of them are in diamond and above.
The statistics are not very meaningful for players outside this group. (about 1.25 - 2.5%)
1 Like
We live in a community, where your opinion must well founded if you want to convince others.
An unjustified opinion is worth nothing.
So of course people will use facts to support their opinion.
Not every opinion is of equal value.
I wish that would be true.
Random sampling is never 100% accurate.
But if overbuff WAS at least somewhat accurate, the pick rates of all the heroes, when added up, would be pretty close to 100% combined.
Go add them up and let us know what they come out to.
1 Like
Let me word it differently.
The number isn’t for the probability of Ana being in a match on a singular team because neither would make sense. 14% wouldn’t make sense and neither would 140% make sense by that perspective.
14% applies the entire active monthly playerbase.
You don’t register these as something which cancel out the other. What you do is think of it like possibility. It’s the probability that one of the two or two of the four will be Ana. That is way more important than the perspective of “one can’t be Ana” because, in the end, we are still calculating someone on one team is playing her.
It’s not, it’s probability but also not probability. (Depends on what you are looking for.) It’s stating how often you will see Ana in a match and it’s quite common to enter a game and both teams have an Ana. Her and Kiriko dominated as the high-end supports for a majority of Overwatch 2.
Why do you think a number beyond 100% is too high?
Your own personal account tracks your Time Played. It’s public data.
Let’s say it isn’t just X. It’s X, Y, and Z.
Now X, Y, and Z have separate events. A, B, C. A is Ana, but you only get A from the subgroup X.
No matter what, though, all these subsets of data generated will generate on a table each time together and you, me, and everyone else have equal opportunity to choose X/A. You opt out of the ability to get A if you don’t pick X, but you could always pick and apply for X. The table generates when all priorities of five people fulfilling a full amount of subsets out of X, Y, and then Z.
Between you and me, each of us obviously have a choice of picking X and also A.
This applies to anyone of the dataset.
See, the probability isn’t about chance of us individuals playing the character. It’s not about probability, either, it’s about the amount you see Ana in matches. I word it this way because we include the possible AND the impossible. The data is there in a way to say 100% is impossible because you’ll never see two teams of five Anas every single game. Or ever. However, you will see one Ana on each team pretty often.
That’s why I mentioned perspective.
To say “Your probability of seeing Ana in a match is 14%” is obviously wrong.
14% is how many and how long the entire Overwatch community plays Ana in Overwatch games.
Your probability to see Ana is %140. That means not ONLY is your probability very high, you are also most likely to see both teams have one.
I understand we disagree, and I’ll be civil with you, I’m just here to explain why there is a differing opinion. How you label the percentages, what they are for, drastically matter.
Depends on what you’re using the data for really. My point was about the correlation between what the devs see with pick rates and the win rates it does.
You talk about sample size and selection bias which will lead to inconsistency in results but it depends on what you’re using the data for. 1% sample size can be meaningful and very accurate depending on how the sample is taken and the number of data points. If you want to know more I’d suggest you study or read up on it.
I was bored, so I did. GM competative win rates over the last 1 month add up to 168.68. Legitimately wonder if they actually are calculating swaps in some weird way that throws it off. But Ana’s current 19.19% pick rate seems a bit… idk, unlikely? Because that’s something like a 96% effective pick rate.
Pick or win rates?
If you’re not in GM go watch a Gm streamer- I’d say this seems about right. Ana is basically in every game on both sides right now.
Adding them up, I get ~120% for GM from the last three months for pickrate for all modes, and comp is way way over 100%. Seems pretty off.
Yeah, I get 212% for three months comp GM. Maybe I’m not inputting something correctly, but seems pretty far over 100%.
All skill tiers this month only comes out to 96%.
Maybe Overbuff just sucks.