Blizzard i hope you learned your lessen about

Remi-tichondrius · October 31, 2023, 1:17am

you implied with no context that this season will pass them with time

what you meant to say was that the inaccurate source you use isn’t updated and they’ve already been surpassed likely 3x over

Kennie-malganis · October 31, 2023, 1:21am

Gotta remember that it’s people above threshold in a season with a week versus a season that ended, so like I said, it’d catch up and overtake them in time. Anyways, curious to see how s3 goes.

Remi-tichondrius · October 31, 2023, 1:23am

it already overtook them

you’re being unclear when answering a question out of your depth, your answer is explicitly “my sources are out of date and as of the last time they were updated x”

not sure 50% chance tank/fw isnt playable so it’s another season you miss glad without me selfplaying you

or 50% chance tank/fw is playable and you trip over 4,000 player glad on accident again

Kennie-malganis · October 31, 2023, 1:25am

Fingers crossed. Hope aug also perma meta in keys, too.

Mondazi-bloodhoof · October 31, 2023, 1:26am

Hey thanks!

but scratch this part, I just went linkdead in a lobby and lost all the points. fml

Kennie-malganis · October 31, 2023, 1:27am

Oh, that’s dreadful. Had you lost any rounds? Maybe your MMR is still somewhat intact.

Mondazi-bloodhoof · October 31, 2023, 1:40am

Probably left it 1-2 if my character stayed in the match and lost the follow up round/s.

Too disheartened to requeue at the present.

Berzz-frostmourne · October 31, 2023, 1:46am

Manfighting:

I actually would be interested in this, but I think you are seriously underestimating the significance of the top-down nature of the API limitations.

If Blizzard’s goal were to prop up the part of the ladder where the smaller pool of healers would naturally cause their ladder to be deflated, a lack of difference would be evidence of inflation, because the expectation would be deflation. Were distribution to remain consistent (i.e. healers had the expected 19% representation at every tier) that would be evidence of them wholesale having a different algorithm for MMR distribution that increased the rate of rating diffuseness. If, on the other hand, they were being propped up using the single lever that Blizzard uses to inflate all their MMR ladders, you would expect the most correct representation to be at the highest tier and for healer representation to increase as you move towards the DD’s median, because Blizzard would have to overinflate the ladder overall to combat the depressive effect of the smaller healer population towards the outer parts of the ladder.

This latter possibility is what we are actually seeing, unless you want to claim that the real healer population is ~25% and it’s purely a coincidence that healer representation at the highest reward tier is hovering around the last verified number for the healer population. I would attribute the closeness of the overall data sets we have access to to the fact that those data sets are preferentially showing the part of the ladder that is being intentionally mirrored. If we had the full data from both ladders, without them being touched, you would expect the healer ladder to just be a diminutive version of the DD ladder. If we had the full data and the healer ladder were inflated, you would see a significant overall rightward movement, but which would still yield the DD ladder extending further; the largest and most apparent gap would be on the leftmost side. This leftward data is hidden from us, however, so what remains is specifically the part of the ladder that has been matched.

If there is a consideration I am missing, please let me know.

Its not deflated or inflated based on healer or dps. They are 2 different ladders in operation. Deflation or Inflation would mean the histograms or boxplots are shifted completely to the left or right when compared with each other(they aren’t). As I said before there is almost no difference in the medians/means for people not in placement matches. The population for healers is just a shrunken version of the DPS ladder due to less players like you have said.

What you say about the full data is incorrect, there are two effects causing the data to be non-normal. The first is placement matches, where people have stopped queuing after a small amount of games on the least played specs(this drags the mean and median down). The second is people who stop queuing at 1600 and 1800 who have gotten their mogs(drags the mean and median up).

Untouched the histogram and boxplots looks like this for the same date:
https://imgur.com/d73Ibzo
https://imgur.com/cUlewle

Kydrav-khadgar · October 31, 2023, 2:37am

I actually feel like RSS is pretty inflated rn, could be wrong but I geared a fresh mm hunt/boomie and I’m already 2100 mmr on both with little to no games played.

Swolé-emerald-dream · October 31, 2023, 2:41am

Remi a savage

Felrynpr-malganis · October 31, 2023, 3:03am

this season is inflated.

Berzz-frostmourne · October 31, 2023, 7:38am

May as well post other findings after grabbing the API again today:

Both means for DPS and Healers increased in rating for non-placement games(> 12 wins). There is still almost no difference between the roles though:

DPS rating
Min. : 456
1st Qu.:1614
Median :1803 (+63)
Mean :1790 (+48)
3rd Qu.:1952
Max. :3014

Healer rating
Min. :1053
1st Qu.:1682
Median :1811 (+58)
Mean :1828 (+60)
3rd Qu.:1956
Max. :2848

Dextructive-tichondrius · October 31, 2023, 11:47am

I don’t think anyone can dispute this season started deflated because of a 2400 MMR cap that wasn’t released for 2 months. Besides, I think it’s fair to at least partially attribute the smaller number of players this season to… just that. Despite its systemic ease of access, DF unfortunately has had rather poor player retention so far.

Manfighting-westfall · October 31, 2023, 11:55pm

The difference between the means staying similar between the two scrapes is very compelling. If the ladders have been untouched then the ~19% representation of healers at 2400+ is, in fact, a wild coincidence, and we can infer that the current healer population is up to around 24%.

The question persists, to me, of why it appears to be the case that the bottom of the holistic data set being cut off irregularly isn’t having a more pronounced impact on the calculations being done on what remains. If you compare a popular and unpopular spec, how far down the API limitations allows you to reach is significantly different, yet between healers and DDs it is not?

Edit: Doing the calculations with the assumption of a 76/24 split between DDs and healers (ignoring the marginal tank population) yields about 3.4% per spec for healers and 2.9% per spec for DDs. Doesn’t necessarily answer all of my question, but the specific number of healing and DD specs relative to their population seems to be producing a very similar spread on accident, which could be normalizing the impact of the omitted data.

In any case I appreciate the correction.

Beedie-malganis · November 1, 2023, 12:26am

OP, not dont do the same thing again, but please do keep mixing stuff up… Make some random spec Gladiator OP for a week whatever, just keep the craziness and laughs rolling.

Haunting-nerzhul · November 1, 2023, 1:16am

damn bummer

Beedie-malganis · November 1, 2023, 1:44am

Learn a lesson about what exactly? They made Ret Pallies the most OP class in the game by a huge amount, and did not notice for a week lol.

Berzz-frostmourne · November 1, 2023, 2:34am

It’s due to specs being lumped together in the calculation, per spec they are wildly different, however when grouped together by role the means are the same. The limitation is top 5000 players. For example, Retribution paladins would have their histograms cutoff at the 1800 rating mark, where as rogues have the whole ladder visible due to lack of players on that spec.

https://imgur.com/hILFhQ4 - Assassination Rogue
https://imgur.com/i8wZggD - Retribution Paladin