Real Legendary drop rates and other statistics for 9785 chars [Spreadsheet]

[impending doom]

Update March 23:

Some basic drop rates for the last week:



Full spreadsheet here: https://docs.google.com/spreadsheets...it?usp=sharing

The method used to track mythic/M+ dungeon drop rates here is to check if a legendary is looted within 60sec after killing the last boss of a mythic dungeon. This number includes any normal mythic and M+ last boss kills, so it's just an average, but I'd guess that a good part of these are 3 chest runs. Based on this I'd say that mythic+ is still a good source for legendaries, at least if 3 chest.

Also now there has been 5 weeks since the blue post about legendary drop rate changes, so I've little more data to see how these changes have affected the KP statistics. However keep in mind that these calculations have been done with a relatively small sample size, and the KP weights used are not accurate anymore either, but I've used these same weights and similar methods as in the last calculations to make it easier to compare the numbers. So this is still just a very rough estimation:



Conclusions:
- Average KP between legendaries (after having at least 2) in the last 5 week seem to be around 600-700 KP. There may be an error margin of 100-200 KP though, depending on calculation method and the type of content used.
- The Bad Luck Protection does not seem to have much effect at all in the first 300-400 KP (or roughly 100 kills), but after that it starts to rise fast, the effect is more noticeable than in the earlier calculations.

These stats are still based on those few 1000 characters which I started tracking in october. I've also almost finished a new code that can track more characters more efficiently, but patch 7.2 is coming very soon and it looks like it will have some major changes to drop rates again, possibly making the bad luck protection much more complicated.


UPDATE Feb 22, 2017:

Here's Nighthold drop rate statistics from the last 3 weeks:



NOTES - There are several things that affect the drop rates, which makes it harder to directly compare different groups:
- People with less legendaries have higher average bad luck protection(more "killpoints"), which also increases the drop chances in these groups.
- People with more legendaries do more mythic/HC raids, which should have have higher base drop chance.
- Blizzard is regularly changing the drop chances.
- There is always some amount of statistical error.

Either way, it appears like the drop rates for people with more legendaries have increased just 1-2 week ago.
I will try to improve the tracking process and kill points calculations, maybe start to track more characters.

Full spreadsheets:
Feb 22: https://docs.google.com/spreadsheets...it?usp=sharing
Feb 15: https://docs.google.com/spreadsheets...it?usp=sharing
Feb 8: https://docs.google.com/spreadsheets...it?usp=sharing


---------------------------

Blizzard confirmed on Feb 17. that each time you received a legendary, the starting chance (base chance) to get the next legendary did get lower (aka diminishing returns):
https://us.battle.net/forums/en/wow/topic/20753149009

But they also said that they changed this recently, so that NOW each legendary after the 2nd one should have the same starting chance.


Because of these changes all the statistics below are "outdated" now, in a sense that they don't represent the current drop rates, however they still show how the system worked in the past.

---------------------------


Update Feb 6. 2017:

First some clarifications:
- The data we can get from armory is somewhat limited and there are a lot of variables, so the results here are a combination of real numbers and some estimated numbers. Therefore it's hard to say exactly how accurate these results are.
- These statistics have been showing some kind of diminishing returns, at least for the first 4 legendaries.
- The average effort(KP) needed for the next legendary appears to be around 40-50% more than for previous one. (it's not "twice as much", that is false/old information)
- It's still mostly random and statistics only give info about probabilities, in the end some people still get more lucky than others.


Here are new drop statistics from Nighthold bosses for the past week. This time I was able to sort the results by the number of legendaries that those chars already had, and even though the sample size could be better, it again looks like the drop chance is lower for people with more legendaries.


Full spreadsheet with more details: https://docs.google.com/spreadsheets...it?usp=sharing


I also did some new Kill Points statistics, with little more data and better tracking for Mythic+, PvP and emissary quests. Example: at 2100 KP the average legendary count is 3, however this is only average, which means you have about 20% chance to have just 2, 58% chance to have 3, 20% chance to have 4 and 1-2% chance to have something else. Please note that the Kill Points model is just a tool to estimate droprates, not scientifically accurate proof of something (see the spreadsheet and rest of this post for more detailed information).


Full spreadsheet here: https://docs.google.com/spreadsheets...it?usp=sharing (work in progress)


Extra speculation - Theoretical model for legendary drop chance:

Drop_Chance = Base_Chance * BLP / GLP
->
Drop_Chance = (0.0018 * KPevent) * MAX(1; 1+((KPcount/250)-1)*2) / (2^LegendaryCount)

Not to be taken too seriously, but when simulated it gives quite similar results as all the statistics in this thread. This formula is based on info from this thread, with numbers tweaked to fit the statistics.


---------older stuff below----------


Update Jan 4. 2017: I made one more stat spreadsheet, now for top 500 EU characters sorted by AP, to see that the legendaries are still dropping quite regularly also for people who have farmed mythics more than anyone else.

This time I pulled more detailed armory data to have better tracking for mythic+ and other things. However with this data I can't track bad luck protection in same way as before, also I have no way to see how many legendaries these characters already have, but I'd guess most of them have 4 or more. So this data can't be directly compared to earlier KP statistics as is, also using just the averages does not necessarily show the "true" drop rates on such unlinear system influanced by BLP, but it gives at least some idea about how often they drop.



Full spreadsheet with more details here: https://docs.google.com/spreadsheets...it?usp=sharing


UPDATE Dec 15 - Latest stat spreadsheet now for the 10th week, and more analysis on how much already looted legendaries and bad luck protection affects the drop rate.

1. Drop rate vs how many legendaries looted

In the Nov 24. update I showed the average KP needed for next legendary which doubled after each legendary, but it didn't take into account the future effect from bad luck protection, and therefore it only showed the average of current situation, but not the real expected average. Also using just the average does not tell the whole truth. Anyway here's the new averages from total 10 weeks of data:

edit: Just to clarify: the "avg KP needed" in this chart is NOT useful for predicting the true avg, because it doesn't count in BLP, so the avg numbers are misleading(too high), but I leave it here for reference purposes: http://i.imgur.com/BklpMEy.gif

To explain the KP system a little: each time I find certain activity from feed data, that can be raid boss kill, weekly chest loot, or mythic+ epics, I increase the KP counter for that character. There is also alot of untrackable activities like emissary quests, HC dung's, pvp chests, world boss etc and to balance those I give 2 KP daily to everyone, but keep in mind that these are all high ilvl characters, so they get majority of legendaries from raids and mythic+. All these weights are based on average drop rates (see chart at bottom) but it obviously isn't very accurate for those untrackable activities or for mythic+. But for now it gives a Rough estimation.

BUT as said, this average KP above isn't very good for predicting exact drop rates, which are dynamic due to BLP. Also there isn't enough data to see what happens to drop rate after 4th legendary, except that it seem to be higher now, 5 chars confirmed to have 5th, while 3 weeks ago there was none.

Anyway the following chart may be more informative:





Here it shows that to have 50% chance for a drop, you'd need 210KP for 1st, 430KP for 2nd, 590 KP for 3rd and maybe around 1000KP for 4th legendary (total 2230 KP). So maybe it's not as bad as needing a double effort to get the next one, but it still shows noticable dimishing returns. Also of course it's still possible that someone gets very lucky and gets 4 legendaries with first 400 KP, while 6% haven't got even 1st yet.

-> Full data Google Spreadsheet here - Includes also a sheet where you can examine results for only certain subgroups (need to copy spreadsheet to edit it).


But... analysing these kind of dynamic drop rates is a bit complicated and would need further verifications, there are still few big issues that may cause inaccuracies:

1. Perhaps the biggest problem is that the activity feed doesn't give data for all activities and isn't always 100% accurate. Even if I'd save the data daily, it doesn't always show all the loot. Could this cause noticable bias in results? I'm not sure, but I might need a more accurate data source to improve the results.
2. Is the sample size enough big and the formulas good enough? Also the KP values are just estimated, especially for mythic+.
3. This is just pure speculation: but different activities might increase the BLP differently. Maybe only mythic+ has dimishing returns? Maybe there's weekly limit for BLP?
4. Blizzard can change the algorithms any time, and already did couple times during the collection of this data. Soon all these numbers may be irrelevant.

So I wouldn't take all these numbers as cold hard facts, but they should provide a better estimation of the drop system than pure quessing or a small group of people.




2. Bad luck protection

Here is a chart that shows drop rate for 10 KP (equals roughly 2-3 raid boss kills or mythic+'s), in relation to how many KP collected since last legendary, and how many legendaries looted before. The effect of BLP is quite noticable. (full spreadsheet in link above)





3. General legendary drop rate statistics



The drop rates for mythic+ at the end are still estimations, due to limited data, but it should show a rough idea what the real %'s are.

-> Full data Google Spreadsheet here



Extra: List of 159 legendary items with ID, name, and how many times each has dropped in last 4 weeks + source of drop

Old links:
Nov 24 BLP/KP spreadsheet here
Week 7 spreadsheet here
Week 4 spreadsheet here
Week 3 spreadsheet here
Week 2 spreadsheet here
Week 1 spreadsheet here
Stats for the last 24h (Oct 27) compared to previous weeks reset days, to see if 7.1 had influence to drop rates


Method used:
1. I saved the names of top 10k characters of Draenor-EU from wowprogress.com, sorted by ilvl.
2. Fetched item and activity feed from Blizzard API for all these chars few times per week.
3. Used PHP code to parse interesting parts of loot data, insert it into spreadsheet and calculate some %'s.

[Ggurface]

I wonder if I delete my useless neck legendary, will that put me back in the 0 legendary drop rate bucket.

[thunterman]

I wonder if I delete my useless neck legendary, will that put me back in the 0 legendary drop rate bucket.

The neck is so undervalued it's hilarious. It can prevent death several times in an encounter, it could save you 200 wipes in Mythic progress. It's a very good second Legendary for an entire raid team to use IMO.

[Deleted]

The neck was my first legendary still waitign on my second tho.

[Deleted]

Why are characters with less than ilvl 847 and characters that haven't gotten 5 pieces of mythic+ loot not taken into account? I believe "theres still a chance some legendaries dropped elsewhere" is an understatement. Good job though.

[Huntaer]

I wonder if I delete my useless neck legendary, will that put me back in the 0 legendary drop rate bucket.

Just put it in your bank

[Alvito]

Perfect so weekly 77% of the no legendary people should get one.

[Tiphess]

While this spreadsheet is nice, people need to keep in mind that despite the bad luck protection(the spreadsheet shows its there), its pure RNG. Trying to find some sorts of patterns or whatever is completely pointless.

[Deleted]

Oh boy, shitstorm incoming.

- - - Updated - - -

bad luck protection

He fell for the meme.

A MEME!

[Einsz]

Thank you for sharing this, would be intresting to get 2nd wave of data from you in 3-4 weeks atleast regarding 0 legendary people to see if the drop rate is higher and how much "bad luck protection tool" affects droprate over time.

[Deleted]

While this spreadsheet is nice, people need to keep in mind that despite the bad luck protection(the spreadsheet shows its there), its pure RNG. Trying to find some sorts of patterns or whatever is completely pointless.

You're wrong. With a big enough dataset you can determine drop chances in mythic, mythic+, heroics, WQ caches and other places. This may help people who are trying to farm for a legendary.

[Sim]

The neck is so undervalued it's hilarious. It can prevent death several times in an encounter, it could save you 200 wipes in Mythic progress. It's a very good second Legendary for an entire raid team to use IMO.

The problem is that the Shield isn't reliable. Its on its own fixed timer and CD. You have to be lucky the shield wasn't low from chip dmg or the huge dmg came when it was down. Sure its better than nothing at all, but to say it can do anything on purpose is misleading.

[thunterman]

You're wrong. With a big enough dataset you can determine drop chances in mythic, mythic+, heroics, WQ caches and other places. This may help people who are trying to farm for a legendary.

Raids by far seem to be the best, I've seen 4-5 drop in one heroic run last week.

[Count Zero]

The neck is so undervalued it's hilarious. It can prevent death several times in an encounter, it could save you 200 wipes in Mythic progress. It's a very good second Legendary for an entire raid team to use IMO.

Yep. If our entire raid had this, I'd argue it would have prevented more wipes than everyone having their best dps legendary.

[impending doom]

Why are characters with less than ilvl 847 and characters that haven't gotten 5 pieces of mythic+ loot not taken into account? I believe "theres still a chance some legendaries dropped elsewhere" is an understatement. Good job though.

I chose only top 10k characters because they are most active, and if I wanted to take everyone above ilvl 800 the amount of data would be too massive to handle. The "5 legendaries from Mythic+" part is used only for the last number, as an attempt to estimate the ratio of legendary loot for people who do mostly only mythic+. But as can be seen in the results, the "drop rate" is obvioulsy different depending on where to compare it to.

[Propainn]

Perfect so weekly 77% of the no legendary people should get one.

Not at all.

6555/9785 didn't have a legendary when he started the sample. Out of those 6555 who did not have one, 906 obtained a legendary, so it's closer to 14%.

[Tiphess]

You're wrong. With a big enough dataset you can determine drop chances in mythic, mythic+, heroics, WQ caches and other places. This may help people who are trying to farm for a legendary.

Wrong.

Of course the most played and exploited content will have the higher droprates when putting it on paper, especially combined with bad luck protection.

If you say something like 10 000 people dropped Invincible after 100 LK kills, while the lets say 5 000 other people dropped it after 5 to 10 LK kills, you will conclude that you have more chance to get the mount on the 100th kill.

See how that doesnt make any sense?

Pure RNG can't be predicted.

[Deleted]

Pseudo RNG is still pseudo RNG. I have seen 6 dropps so far, 5 of them were on the last bost of the daily hc, and one on a hc raid boss. Don't try to see patterns where there are none.

(No drops for me so far though )

[skitzy129]

The neck is so undervalued it's hilarious. It can prevent death several times in an encounter, it could save you 200 wipes in Mythic progress. It's a very good second Legendary for an entire raid team to use IMO.

It's not undervalued at all.

It's value is only relative to its competition, some of which are MASSIVE DPS increases which dwarf the absolute shit out of it.

The second factor is that it's a piece of jewelry. A legendary Helmet will give you ~700 more primary stat over a typical 850 helm. A necklace will give you...nothing.

Stop making excuses for this insanely horrendous design Blizzard stuck with. It should have never ever been possible to get bad legendaries to begin with, but here we are. And before some other dickhead chimes in "herp derp at least you actually have one", that doesn't change the FACT that shit legendaries are in the game and should not be.

[Deleted]

Wrong.

Of course the most played and exploited content will have the higher droprates when putting it on paper, especially combined with bad luck protection.

If you say something like 10 000 people dropped Invincible after 100 LK kills, while the lets say 5 000 other people dropped it after 5 to 10 LK kills, you will conclude that you have more chance to get the mount on the 100th kill.

See how that doesnt make any sense?

Pure RNG can't be predicted.

It can be predicted. Your example just doesnt make any sense and will not happen in the real world.

If for example out of 15000 tries 1000 mounts were obtained. You can predict the chance of getting a mount the next try is roughly 1 in 15. With a margin of error dependent on the sample size. This is basic statistics.

Also, using percentages automaticly corrects for the amount of times Different types of content is being cleared. (If 1% of heroics gear dropped is a legendary and 2% of mythic gear is legendary. You can then say mythic gear dropped is twice as likely to be a legendary.)