The thing is Median and Mean pick rates don’t really matter in this case, they are different paths up the same mountain.
If you balance for EITHER across each hero, you still end up in the same place.
If you kept buffing heroes under the median, and nerfing heroes above it - and, say that power was the only part in determining pick rates (I know it isn’t right, but stay with me here)
You would STILL end up in the same place, where everyone had the same pick rate at the end.
I think median is actually the better target. Because, as you have pointed out, Mean is biased out in one direction.
It would require less steps if you kept targeting for Median, since your “target level of power” wouldn’t be moving.
The mean would change every time a hero was rebalanced, meaning the previous ones wouldn’t be in the same place relative to the target any more. The Median doesn’t have that problem.
But getting upset that someone is using Medians, is a weird place to be, since, the real issue is pick rates in GM are indicative of power, but isn’t a direct measurement of it.
Take my idea of balancing between roles. in THEORY you would try to balance for power or impact or something like that.
But in practice, Blizzard will eventually HAVE to balance for player base, because of queue times.
Practicality wins over theory every time. Median is a very practical target if you are going to use pick rates as a target in the first place.
I would argue you shouldn’t, but, I don’t think Genji would do any better out of the ones I would use myself.
I think what drives people crazy is Blizzard doesn’t have a stated balance philosophy, and even if they did, it is obvious they wouldn’t apply it consistently.
I’m not sure what post you are referring to, there are 324 posts in here so far so I may have missed it, but I do recall treb saying he doesn’t think genji is op or underpowered, he thinks he is balanced (just like i do). I don’t really consider anything he posted as a lie tho (or possibly missed it).
I don’t think many people consider you guys a joke, but making posts every single day about the same topic could be seen as annoying to some people. I’m not sure if you were here during the mercy raids, but it caused a negative stigma towards forum mercy mains for a long time.
First off median is a terrible metric to use and is probably the most intellectually dishonest and I suspect it comes from trying to use statistics to prove a theory instead of building a theory based on statistics. If you want to use that pickrate though, it needs to literally triple (not actually kidding) in order for him to reach a hypothetical balanced pickrate (aka if the game were balanced and all DPS were equally viable and popular all DPS would have a 2.06% pickrate whereas his is .68%.
The reason we use GM data is because that is the only area where imbalances between heroes actually exist. If you are in Gold you can pick ANY hero and assuming you are not being hard countered, your hero choice is NOT RELEVANT to the result of the game.
The game is balanced at those ranks (except for Sombra who is terrible). The hero is so inconsequential to the result that it creates a balanced game. Games are decided by who plays their hero itself and with the team better as well as various other things such as playing the map well, positioning, ult management, and enemy matchups (aka the basics of OW). The basics decide games at low ranks. Since no one at those ranks can do the basics well, they are all that matter and whoever does them best wins. At GM everyone is doing the basics well, or close enough to well, that suddenly that tiny portion of the game that is hero choice suddenly becomes relevant because everything else is so close in terms of skill.
Median is a pretty good one. if you were going to use a single number to balance.
Say we had a bunch of heroes, and their powers were…
1 2 2 3 4 5 5 5 6 7 8 8 10 20
You rebalance the 1 to a 5. Since you are after the median target.
2 2 3 4 5 5 5 5 6 7 8 8 10 20 - 5 is STILL the median. It won’t move.
You balance the 20…
2 2 3 4 5 5 5 5 5 6 7 8 8 10 - 5 is STILL the median. It won’t move again.
But each time you rebalance for mean, the mean THEN CHANGES.
This makes the Median the right answer for this. It isn’t someone trying to be dishonest, it is the right answer from a maths point of view.
Now, trying to use GM pick rate as your sole way to balance heroes isn’t going to work well anyway…
But I won’t have you bag on using Medians here, it is the right tool, and they are using it correctly for what they are trying to do.
Means would be the wrong tool, since it would charge the target every single time you adjusted a hero.
You can argue that pick rate in GM isn’t 100% related to power, sure, but get out of here with that “intellectually dishonest” stance. Median is the right tool for this job.
If that was the case, then all heroes would have the same win rate in gold.
Peoples picks STILL matter.
The spread of win rates in the lower ranks is just as much as in the upper ones. Their average skill, and the skill of the people they would be facing is different, so what is a good or bad pick is different, but make no mistake, it is still a big deal.
I basically did like a half revert on his Shuriken’s and Reflect. They did 28 to 30 damage for Shuriken’s, buffed his Secondary fire from 0.75 to 0.65 recovery, & made his reflect go from 1.5 to 2 second duration. Just went with the 50% revert to all of those, then threw in the Dragonblade change.
Median is a generally useless statistic that is rarely ever used (its right up there with raw average in terms of numbers actually statisticians consider useless). It has value as something that helps you parse the data but to analyze it has no real value.
If you are dealing with sub 1% pick rate heroes (aka heroes who are MORE THAN 50% BELOW their expected pickrate) you are dealing with bad heroes. Now maybe they are bad because of the meta or bad because of their power level and usually that is harder to tell but generally over the past 18 months with the exception of the 5-6 weeks when they made Genji completely broken, he has been underpicked by a wide margin regardless of the meta. The meta shifts and his power does not. That tends to point to an underpowered hero.
Second, win rates on Overbuff are known to be inaccurate and if there is one thing worse than statistics used poorly, its statistics that have been shown to be untrue so I do not use win rates (especially considering how flawed the win rate calculator for OW is anyway) when thinking about power level arguments.
As for why there is a divergence in Gold, that is pretty easy to understand. First, Gold/Plat players play every single comp like they are running a Rein/Zarya with an Ana healing them (I use a Plat toon as well as a low Masters toon to learn new heroes such as Ball who I was learning recently). It makes no sense to me. They can be running Sigma/Ball and they will still position themselves like its a Rein/Zarya and in OW positioning is EVERYTHING and if you do things that wrong you will get run over. There are other reasons, such as how win rate is calculated, some heroes like requiring more communication, good heroes like Widowmaker or Sigma/Ball who are built around high mechanical skill/positioning are going to struggle when those things are inconsistent etc. etc., but in general those numbers, with the exceptions of a few heroes such as Sombra, are actually reasonably close.
Edit:
Using a valid sample btw your numbers for this median are:
When dealing with an exponential model such as this, median is useless. What you look at is expected out come versus actual. If your expected outcome with 2 across the board and gave me those numbers, I would first tweak the extreme highs and extreme lows to see if they shift accordingly (given we are dealing with a zero sum game after all), but if after rerunning the numbers after 15 tweaks I was consistently getting the 8th and 9th heroes being that low and always being the same heroes, I would identify it as a problem with those heroes as well.
Basically outside of the Math OW has 3 good DPS choice 3-4 acceptable ones and the rest are just useless. DPS balance is not in a great spot right now.
Obviously you don’t work in statistics. The 50 percentile is almost ALWAYS better than the mean.
The problem with it is it is harder to calculate in a clustered environment.
No seriously, it is ALMOST always what you want. It doesn’t get wrecked by outliers.
Take this series. 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 10000000
Median tells you what you expect any one sample will be close to.
Mean is almost always horribly screwed by outliers, so statically, you don’t want to use it.
The TARGET for where a hero should be is at the 50 percentile.
Again, a quick demo will show why this is the case.
Say we are ALMOST there.
We have 4 heros, all at 1% pick rate, and one other hero at 40% and the other at 46% pick rate.
If you try to adjust the 46% picked hero to have a mean pick rate. (and you succeed) you are still placing them above all of the 1% heroes, where, you want to place them AT their level of power.
So when you fix the last hero, everyone is now balanced, rather than have to do the process again, and again and again…
They have different skills in different things, that is why.
They can’t hit other targets, and if they could move ball around properly, they wouldn’t be in gold.
Nope, it is STILL what you are after BECAUSE it is a power curve. See the above example on why.
If you don’t use Median as the target, you have to rebalance the same heroes over and over again.
When you have a skewed distribution, the median is a better measure of central tendency than the mean.
Looking at GM gives you a clearer picture when looking at what’s good or bad, as in lower ranks everything can work, and usually, hero popularity and how fun they are plays a major role, which isn’t something you necessarily want to base your balance thoughts around, yes, people in GM also play some things because they are fun, but you are much less likely to encounter those statistics that you are trying to ignore.
If hero A is bad but hero B is worst, it doesn’t mean that you can’t help both.
Pointing out that Genji is balanced because other heroes are bad doesn’t mean Genji doesn’t need some help.
but that is again, misleading, it paints a picture like Genji is doing fine - “Look he is in the middle”, and you are saying that he isn’t yourself now, heroes can be worse but it doesn’t make Genji better.
But that “most cherry picked sample” is one that doesn’t favor him, if it favored him then sure, but if it makes him look better and makes it seem not bad, it means that it is worse then what you are painting it out to be.
The problem with the median is that when it is given out of context it makes things seem worst or better, the average grade of 11 kids can be 49, yet the median is 90, it is an extreme case, where 5 kids have a grade of 0 and the other 6 have a grade of 90, but it shows very well how out of context the median is misleading.
Look, I have experience with you, you may have changed, but it seems to like you are avoiding arguments people have, like the whole thing with Tornado in this thread, I don’t think you are in the position to say that other people don’t want to hear your points when you yourself are doing the same thing to others.
Very true, but since you can’t achieve perfect balance you’d be better off trying to make heroes be in line with heroes that you consider as balanced, those heroes are likely to fall around ~2% pickrate (~1.5% for Overbuff right now), so that is where you should be aiming, Blizzard could very well say that they consider Genji has their balance point and work out from there, but I think that it would place DPS as a much weaker role overall, it seems to me, at least from the way Blizzard has been doing changes for the most part that they are working mainly on powerlevel, rather than Pickrate, Winrate, and stuff along those lines, but what it does do is show us a trend, a hero is much likely to be weaker when they have a lower Pickrate, as fewer people would pick them, and the same thing for stronger heroes.
I generally think that it is the best way to balance a game, not by stats alone, but by looking into extreme cases that the stats show.
I think that using both Median and Mean Pickrates is important, but when asking the question of if a hero is balanced, or how well they are doing, I find Median to be misleading, as it doesn’t mean a lot about the state of the hero themselves, but rather the overall balance of the whole role, as if the median is lower than the average it means that the extreme cases are mainly above, but that doesn’t say a lot about the hero.
The whole thing with using the median to judge how a hero is doing, it is not a “lie” in its typical form, of saying something that is incorrect knowingly, when saying “lying with statistics”, means that the person is giving something that is theoretically correct, but in reality, it is just misleading.
Things like “9/10 dentists recommend this toothpaste” is lying with statistics, it is correct that they recommend it, but it makes you think like this is their number one recommendation, when in reality they are just saying a list of kinds of toothpaste, and only 1/10 does not put that specific toothpaste in his list, again, it is not incorrect, but it is using statistics in your favor to show a different picture.
There are a lot, I don’t know how to really show it, but there is a lot of negativity against Genji main.
As I said before, most people don’t do this, I for one am pretty sure that I have made about 2-3 posts since August about this all thing with his balance and state, it comes back to that whole thing of people generalizing - If one dog bit me I’m not going to hate all dogs.
I was there, and it is not very comparable to the situation now, as the Mercy raid was planned by an individual, there isn’t some sort of secret Genji mains cult that is similar to what they had back then (at least that is the most I can tell you about it ), but in all seriousness, this is not the same thing at all, people are having a problem with their hero and want him changed for the benefit of everyone, they just post when they like to, certain people do spam the forums some times, and need to be stopped, but that is it, you can’t really blame a person for wanting to show the world his idea on how to fix the hero, he is not doing as part of a raid so his voice would be heard above other people, but they want to be heard, to show people their thoughts, if someone doesn’t post every day about this, there is no problem with it.
I note the Mean is just as awful here. Trying to argue Mean over Median and use that example? It isn’t ANY better.
If you were going to use that distribution, and say, well, the next kid is going to do the exam, what is the score they are likely to get? Well, the Median MAY be the better answer, since they either get 90 or they totally bomb, and MORE people pass (just) then bomb.
There isn’t any way that mean is useful at all here either - it is even more misleading. NOONE got close to the mean. No one at all.
But in overwatch, as a target for balancing you CAN show it is the right tool.
I’ve been pushing out examples above.
Median is the right tool here. IF you were going to use pick rate as a proxy for power.
Except that, if you were going to, the ACTUAL balanced hero you would try to target WOULD be the median one. Targeting above that would always lead to power inflation.
Every time you rebalanced a hero to have a 2% pick rate, most of the heroes would be badly balanced. If they THEN applied that again, to the next lower hero, they the last one you had wouldn’t be at a 2% pick rate again.
It would be a forever treadmill of making heroes more and more powerful.
If you target Median, you don’t get this, you hit a point of stability.
Does anyone actually take Ergyn’s threads seriously?
Either he’s a troll, in which case everyone who engages with him is an idiot, or he legit believes Genji is broken, in which case he shouldn’t be taken seriously anyway.
I can definitely confirm this. I have been here for a decent amount of time and there is a lot of blatant bias towards certain players and heroes.
For example, when someone even thinks to bring up something about Doomfist and how he’s suffering they “Good riddance, I hate him so he deserves to be bad” and there will be a whole slew of people liking that comment
Though when someone even hints that they might think that about, say, Brig, they are mass flagged.
Then you have the stigma that people put on Genji players saying they are all “Whiny 15 year olds.”. People will just say “Oh, you’re just a Genji main” as if that somehow invalidates what they say.
In Overwatch, you have a cap on how much a hero can be used, which is something you can show in those examples, the maximum DPS hero can be used is 16.66%, and there is always going to be another slot that a DPS hero can be played in, the average of DPS Pickrate in Overwatch would always be ~2% as long as they are not going to change 222, which means that your average is consistent, and that your maximum is capped, so balancing around something that is a const would always be better when trying to balance.
Examples that do not have a consistent average with a capped maximum (including mine) are irrelevant when it comes to Overwatch balance as they can have extreme numbers which can affect what the mean is, leading to things like your example happening.
I mean, this example was more to show how much those can be different rather than showing that one is better over the other, In reality, you need to use both to get something that is more accurate.
You know that it needs to go both ways, you are not just going to buff heroes until they are good, you are also going to nerf heroes that are too good, nerfing the top and buffing the bottom would lead to balance, but you need to have a point which you are trying to put them at, that point needs to be a hero (that you consider as balance and don’t change) as well as the average Pickrate, without using both you are going to lose the anchor point, as both can be easily shifted, but together they are a const, which is what you want to have as your middle ground and go-to powerlevel.
), but in all seriousness, this is not the same thing at all, people are having a problem with their hero and want him changed for the benefit of everyone, they just post when they like to, certain people do spam the forums some times, and need to be stopped, but that is it, you can’t really blame a person for wanting to show the world his idea on how to fix the hero, he is not doing as part of a raid so his voice would be heard above other people, but they want to be heard, to show people their thoughts, if someone doesn’t post every day about this, there is no problem with it.