Blog post thoughts

We could debate what’s the bigger mistake: a) taking into account part of data that has bigger quality or b) taking more data of lower quality and scaling it…

If you can do scaling the right way than it is definetly better to go with option b.

But scaling process is far from perfect. You have to imagine what 1K Paragon guy would be able to do at 5K Paragons. And you hope you took everything relevant into account. Very easy to make a mistake there.

Honestly I’d really just like him to take his numbers and put his efforts into bringing up weaker classes/sets.

No one will hold their breath though.

With great data, comes great responsibility.

Just look at the ladder. On EU for barbs level 130 starts from 289, for wizards - 60, and for monks 90. On US, respectively, 112, 29 and 35, on KR 160, 25 and 42. It is obvious that the barbs build is stronger, and if you get the same average values for them, then this means that the methodology is erroneous.

Averages outside the ladder are corrupted by bots that closed only low levels grifts in fast run. (For example, many players with high paragon barbarians have records of 110-115 in 3-4 minutes, especially at the beginning of an era.)


Big ups for Nev and Matt and the rest of the team for mentioning stutter stepping, thanks a bunch!

1 Like

And that is why the correct thing would be to balance around the top level, as we would have more accurate data.

1 Like

Just wanted to chime in and say that I appreciate the transparency and wished this would have been released 5 years ago to better understand nefts and buffs from the devs point of view. Thanks for writing it though :slight_smile:

Since the statistics are based on players (see Nev’s post) you can rest easy. -They don’t use bots.

You aren’t balancing a class or a set at top level you’re balancing a player. So no, that’s not the correct path. You factor it into an average. Re read what Nev wrote, from a statistics standpoint it’s correct to do this the current way. People aren’t going to like it because it removes a lot of potential for hyperbole and dishonesty. The ‘feelings not fact’s’ which includes people who like to try to use skewed data to present an erroneous argument crowd.

More accurate data with biases and skews? No, less accurate. Especially if people at the ‘top end’ aren’t pushing as hard on some classes as others for whatever reason.


They should look at how build performs by playing it… not by looking at statistics only. I am not sure if they know their own game.


I understand you cannot give exact numbers however 5k paragon NS still seems to be on the high side to be considered an ‘average’ check point. From the actual live bodies playing, wonder what %age of players are 5K and above vs 3k - 5k? IMHO, 3K seems more ‘average’ however that’s just my 2cents.

1 Like

You do realize the D3 team is small aka they don’t have hundreds of internal testers for practical verification of data, right?

They don’t, but they are trying. Let’s give them the needed support with productive comments and discussions instead of bragging that some class is not part of the meta. There will always be a meta and classes left out of 4s until the core design isn’t changed so that players don’t have to follow the meta all the time.

But they can make char with 5000 Paragon, infinite keys, and what ever gear they want in minutes.

Like we do for years now??? Bragging about group play is bragging about ability to play the game at all for me.

why would somebody ever play solo to farm XP??

Honestly, I am a bit disappointed.

A common misconception is that we balance solely around 4-man groups. While we certainly take it into account, group play is not our only focus because not all players enjoy playing in groups. We want to make sure that the content we’re designing can be enjoyed by the most players, so our design decisions should take both styles into account. If we were to design solely around one style of play, the other would be severely impacted (and likely a lot less fun).

I understand your reasoning. But… It doesn’t shock you to see always same classes in META? Like some does’nt really have their place in a 4 team?

Sure, balancing is good, but we really need build diversity ATM.
I don’t really see how you want to adress the major flows off the game, which is kinda worrying IMO. (When we see some other games really invests a lot in this point.)

I play solo for everything. I really really don’t like playing with other people. I get why the game is balanced to prioritize group play as most efficient. I don’t care though. I still do everything solo.

My biggest wish is that they make solo bounties less painful.


What they can do is change the PTR with preset characters that can’t be modified in addition to normal characters, so that the testing phase relies on less variables.

No you exactly deserve that

They can make any char they want with any stats in minutes. And test what ever they feel like testing.

I remember when, by mistake, some weird clears showed up on PTR. Clears done by Blizzard or people hired by them.

1 Like

Have to agree, it’s laughable how much more time it takes to complete bounties solo than it does in a split farming group.

1 Like

It’s not that simple - you need more data for proper balancing, which means more internal testers that they don’t have.

The best way to balance the game properly (if that is their main goal now) is to improve the PTR functioning.