Stats review one month after rework
So it’s been one month since the big rework and for fun I wanted to take a look at the stats on the aggregate site I follow and see what’s changed statistically. 🙂 (Just to head off the question I get sometimes, the mods don’t want links to the site posted since a lot of aggregating sites use Steam API which the game doesn’t officially support. They don’t seem to mind me posting the numbers from it though.)
- For reference, the site gathers endgame screen data from voluntary uploads of something like a couple thousand games a week or two, which while being a small fraction of the overall games is still large enough to typically have interesting correlations with the actual numbers.
- Kill rates and killer pick rates are based on the players who do the uploads, so those numbers are going to be skewed toward the skill bracket you’d expect from people who are interested enough in tracking their own stats to bother using this type of website. So for example Nurse has a much higher kill rate on this site than in the overall charts the devs provide because newer and more casual Nurses haven’t quite gotten past her initial learning curve.
- Perk usage percentages are based on the perks used by the uploaders’ opponents, not their own perks, so those numbers are actually from a more randomized sample of players and therefore presumably a little less biased toward the website user base than the kill rates are.
- The average overall kill rate over the month prior to the midchapter patch was 48%. The average kill rate in the month since then is 57%. Because of the skew toward higher MMR users my guess is both of these numbers are probably about 5% lower than the actual total overall average kill rate in the game since historically I think lower MMR killers have slightly better kill rates because it’s slightly easier to get kills on survivors who aren’t quite up to speed than it is to escape killers who are still somewhat new. So it wouldn’t surprise me if the overall kill rate across the whole game is around 62% over the last month.
- Another factor to keep in mind is the first week on the post-patch month had a slightly higher average kill rate than what we currently seem to have now because survivors were still getting used to all the changes. So that 57% net is actually something like 62% the first week and 55% the following weeks. There’s also been a bit of a rash of people giving up early which is probably inflating the numbers slightly too, so combining that with my guess above that the website is about 5% lower than the actual I’d guess the actual number over the last week is around 58-60%, meaning the game is slightly killer favored but definitely still winnable by survivors.
- There are no stats on the breakdown of kill rates against solo only groups versus groups that include swf members, but historically the dev stats have put the gap at about 3-5% additional escape rate per person in a swf. So if the average overall kill rate is 60%, then solo survivors are probably on the order of a 65% kill rate while opponents that include swfs range from 50% to 62% depending on the number of members. This is why the devs commented in January about their desire to buff solo survivor play with additional information so hopefully once they catch up on the current roadmap they’ll be able to get back on track with those solo changes.
- The highest kill rate killer over the last month on the site is Plague at 69% followed by Pig at 68% and Nurse at 64%. Note that Plague’s rate is inflated though because of the initial rework of Thanatophobia which has since been nerfed so I’m expecting her kill rate to go down in the next month. Pig, despite the common perception on the forums of her being weak, has always had a very good kill rate since they changed how her boxes work a couple months ago. And generally speaking her kit is best against average players and weakest against really solid players on comms, so I think her overall kill rate across the entire game is significantly higher than her kill rate specifically in the top MMR brackets, but that’s just a guess on my part.
- The lowest kill rates are Twins 46%, Ghost Face 45%, and Trickster 44%. Twins and Trickster likewise have had the lowest kill rates in the devs kill rate charts they released so that’s not surprising. Ghost Face’s low kill rate is a little unexpected since he’s normally around average, but I have a feeling that he’s having a little trouble currently because of how easily he can be broken out of stealth at the moment. If the devs were to tweak the mechanism a bit so survivors have a little harder time breaking it he’d snap back up.
- Top three most played killers on the site this month are Huntress, Legion, and Blight, all 7%. Legion will probably fall off a bit though next month now that Thana is nerfed.
- Lowest three pick rates this month are Hillbilly at 2% and Hag and Twins at 1%. Not unexpected for them probably.
- Over the last two weeks, the top five survivor perks are Windows of Opportunity 22%, Prove Thyself and Lithe 21%, and Sprint Burst and Adrenaline 19%. Windows of Opportunity is the only maybe unexpected entry there, I think survivors are finally realizing just how useful it is to know exactly where to run to get to a pallet or vault. (It’s certainly much more useful info for survivors than Zanshin Tactics for killers.) Dead Hard is number 6 at 18% so it’s still popular as well, though not nearly the juggernaut it was at 37% usage prepatch!
- Top killer perks are BBQ and Chili 22%, Scourge Hook Pain Resonance and Jolt 17%, Save the Best for Last 16%, and Corrupt Intervention 13%. Some people thought BBQ and Chili would never be used after its bloodpoint bonus was removed, and it did used to have a whopping 37% usage, but this demonstrates it still has a core base of fans who were already using it mainly for its aura and the bloodpoints was just a nice side benefit. Likewise Pain Resonance remains a top picked perk despite its nerf. Also Pop Goes the Weasel is still decently picked at 13%. Again, not as high as the 20% it used to enjoy but still in the top ten perks. That’s because although it’s weaker it still gives good enough regression when you really need it against gens that already have a lot of progress. (Which is also the key feature of Pain Resonance, regressing the gens that are the most progressed and in need of it.)
- And as an anecdote personally for my part I only recently started tracking games, I’m at about 70% kill rate over 20 matches the last couple of weeks (I only play 1 or 2 matches a night and always just play whichever killer I have a daily ritual for, or Ace when I on rare occasion play survivor) so I feel like the patch has definitely helped my kills. 🙂 Also the Prestige rework gave me a massive perk windfall, I went from having every killer at Prestige 1 or 2 with 30-40 perk levels in their inventories each to all of them having every single teachable perk at level 3 which was fantastic! 😄 It’s allowed me to finally experiment with some builds I was curious to mess with but didn’t have everything unlocked on everybody like Backpack builds and all four Scourge Hooks and builds around smashing pallets. (The one I’m trying this week on various killers is a Bad Medicine build - Coulrophobia, Unnerving Presence, Septic Touch, and Distressing. Survivors in your very large terror radius get healing checks more often and that spin more quickly and which take twice as long to do, plus they’re Blind and Exhausted while and after healing similar to Fearmonger for gen repairs. Got a 4k with Clown using that last night, good times! 👍)
Comments
-
Cool data.
Yea I been running Unnerving, Distressing, Coulrophobia, and Merciless Storm on my Trapper and it's pretty decimating.
2 -
Clearly trickster needs to be nerfed :p
1 -
Ok, but, My stats are
Killer pre patch : 73% kill rate VS 72% kill rate pos patch, so, survivor can play in high MMR and SWF, buuuttt, look my SOLO SURVIVOR STATS
Pre patch: 45% escape rate VS 28% escape rate now.
This game already needed Solo survivor improvement helps before, now its so much necessary. I am scary about silent from DEVS about this. Play as solo now, is so much boring, pain and worse than before. Play as Killer is fine now.
5 -
Keep in mind Nurse is unusual because her kill rate graph is probably an upward slant with low mmr Nurses having atrocious rates and very high MMR Nurses having really amazing kill rates against equally good survivors. So her average kill rate on this site and on the overall stats are probably lower than the rate she has if you look specifically at the top end, and I’m guessing there are a lot of prolific posters who are more in that upper bracket and see really excellent Nurses more often.
But yeah, for average players and even above average players Nurses won’t be playing literally perfectly and missing Blinks and her slow movement speed are punishing enough that even small misplays really hurt her. That’s why she’s in the top 3 above but not quite the top position, the players posting her results are above average but not all the top, top end.
0 -
Right, like I said above the solo survivors always have a pretty significant gap in escape rates so I wouldn’t be surprised if they’re hovering around 38% escaped, maybe even lower. That’s why I said I really hope once the devs catch up on the current roadmap they get back to what they were working on for solo survivors.
2 -
What's the name of the site?
1 -
The data is useless, the amount of games uploaded does not represent the actual population results. These uploaded games are from such a small sample of the population. Also these people are biassed since almost no1 will spend time on uploading this unless they want to prove a point (so data uploaded will be in their favour). Reading these stats is like measuring the average height of 10 males that voluntary do this (so will probably be above average) while measuring each person 50 times. It gives no correct data that you can/should analyse.
4 -
Very nice stats my friend. Good job.
Not at all surprised to see Hag and Twins still being in the lowest pick rate. I think any sample size of data would show that. They have always been my least seen killers.
I am surprised BBQ is still one of the top picked perks. I never really used it myself as its not a perk I find particularly useful on Mikey, but I always assumed the major attraction to it for most people was the BP and not the info it gives. Clearly I was wrong.
2 -
You’re definitely both underestimating how many games are included and also not understanding how the data works for the perk usage.
2 -
It doesn’t include any console player data whatsoever. That in and of itself creates a lot of skepticism. Additionally, because it’s voluntary, there are likely many players who don’t spend any time uploading data. I’m sure it’s probably a microcosm of what global game trends are, but I wouldn’t put so much stock in this.
1 -
I'm always surprised about Ghostface and Trickster's kill rates. I typically 3 or 4k with them. Really sad about Billy's pick rate. So dumb what they did to him.
3 -
I just touched the basis on why this 1 thing i mentioned already made the data completely useless to analyse. Its already breaking 1 of the first rules of statistics, i dont find it worth to spend my energy in mentioning everything else why this data is bad to use, i would be writing a few pages.
However, this 1 single thing you mention about perks usage is something indeed a bit more useable, yet it isnt because of the small amount of people uploading this (as a killer its not hard to just add games that include certain perk usage that you hate to try and prove a point, again, this data is extremely sensetive to bias, and we all know how bias completely dominates every discussion on this forum)
I think youre underestimating how big the sample has to be because of the so many variables in this game that cant be controlled.
1 -
Again, it’s not true there’s no console data for the perk usage because those stats are based on opponents’ using the perks, not the people uploading it. So the perk data actually is probably fairly accurate for the MMR bracket the site population generally falls into.
Also I said at the very beginning that the kill rates in particular are a skewed sample because those are based on the players who voluntarily do the uploads. However their relative upward movement in kill rates is probably not far off from what the overall database sees. (I.e. The site’s beginning and ending kill rates are I think lower than the overall data but moved up a proportional amount.)
0 -
I disagree on the size not being sufficient, particularly for the perks. Once you’re above a few hundred entries in a sample your confidence level is already fairly good for an informal result and the site is collecting thousands of essentially random opponent samples in the periods discussed. And I seriously, seriously doubt there’s some bad actor intentionally trying to upload extremely cherry picked opponents from multiple games trying to skew the data.
Feel free to ignore the post if you want, though, that’s fine by me.
1 -
Hmm stats are sus, you're saying Nurse is at 64% killrate but she was lowest killrate in january 2022, Pinhead was highest but now suddenly he's not even top 3 ? He still has same oppressive passive he should be higher
0 -
Nurse is the lowest in the overall ratings but as I mentioned above Nurse is an example of how the site’s kill rates are skewed toward people who play a lot more regularly.
As far as Pinhead goes he’s number ten on the list at 58%. So basically average. Also note that the stats you’re talking about were in mid December to mid January and Pinhead came out in that September so was still somewhat new. On top of which Pinhead is generally considered stronger against weaker players who don’t necessarily really get how the box works so his high kill rate in the overall stats is probably a very high rate against average to below average players but drops down as MMR increases.
2 -
Its hard for me to ignore a post that has so many flaws in it regarding statistics. My background/experience in statistics gives me the urge to correct someone that is completely out of touch with the subject and doesnt know how to use/read data, especially when data is being used that shouldnt be used in the first place. Ill stop trying to make you understand why this is bad since you already disagree with the simple parts and clearly dont comprehend the basics (you also emphasize using the data is useable for perks, note that this is just a small part of your post, also the representability for perks is still questionable to say the least). Those people that upload arent random people, these uploads are biased, this is just how humans psychology works but bias like this is countered in actual research with for example different techniques in which way the sample is gathered.
I can only ask you to do more research on what makes data representable enough and what makes it reliable, and trust me this is alot of research since in some cases a sample of 5 people is already enough and in some cases 10 000 isnt enough (which brings is back to the start since every upload in your data is not done by a unique person which makes the conclusions youre trying to make completely unreliable).
Ok, Ill give 1 last simple try for you to comprehend a basic on why this is such bad data to use for the conclusions that youre trying to make:
Does the average population like apples? you ask 10 people, 5 people like it and 5 dont. In your data you ask the 5 that like it 10 times while the people that hate it you only ask once. RESULT?: you got 50 people that like apples and 5 that dont, and this is used in your data to represent THE WHOLE POPULATION that ever ate an apple, and on this data youre going to make conclusions?
Just the simple FACT that this dbd data includes subjects that are measured multiple times MIXED with subjects that are only measured once makes this COMPLETELY flawed in representing a population.
0 -
Im surprised to not seeing Freddy one of the least played killers. But also not surprised that Billy being one of the least played killers. Good job devs, you gutted most balanced killer.
0 -
pretty sure its nightlight
0 -
So you agree the game is even more killer side that really suckes
0 -
- I’m a math major myself and use stats all the time in my job, I fully understand the difference between random samples and voluntary polls and even said in my original post and my replies that I expect there to be a skew in the results due to the voluntary uploads.
- The perk usage was literally half my post, not “a small part of it”. And I clearly spelled out in the intro the difference between how perk usage is calculated from opponent data versus kill rates being from the uploaders’ data.
- Your counter example is clearly a much, much smaller, greatly exaggerated scenario compared to the set of data I’m looking at.
- I get your complaint that, for example, if someone is maining Pig and they play 30 games and they’re the only person to upload it but all the Twins games are different players and they each upload 1 game each then it can give an odd bias in the kill rate results. (It wouldn’t change perk rate data though since those are from the opponents.) That voluntary bias is why I mentioned the disclaimers in the first place and why I didn’t say something like “the devs should be balancing around these numbers”. But just because the potential skew exists doesn’t make the numbers less interesting to look at, nor does it mean there aren’t actual underlying correlations that they point toward looking for in the full data set. Pig having a really high kill rate might be a fluke in the voluntary uploads, sure, but it could also be an indicator of something more fundamental going on with her as a character and it’s that possible path of inquiry that’s worth examining.
As an analogy, the forum has polls posted all the time of “What’s your favorite map?” or “Who’s your favorite killer?” etc. And I still enjoy responding to those polls and looking at the results even though I am completely cognizant of the fact that they are extremely skewed by their voluntary nature and very limited population. Not because I expect to use those results directly for making decisions but because the results can bring up hypotheses that can then be examined more formally using better data sets. Informal polls are a first stepping stone to finding the questions that might be worth pursuing more analytically.
2 -
Personally I don’t suspect Hillbilly is actually weak, he still has great map mobility and the potential for instant downs, even though it can be tricky landing them. I’m not a fan of him though myself just because his chainsaw steering and movement is so awkward midchase. There are other character with similarly useful or even better map mobility I like better, and if I want instant downs Bubba chainsaw is easier to land. So for me it’s a style thing versus thinking he’s weak.
0 -
Well per the devs they felt the numbers were slightly survivor sided before the rework so one of the goals was increasing the kill rate, which is what’s apparently happened. As I was discussing with someone else in the post above, the kill rates from this site are from the equivalent of a voluntary poll so you have to take them with a grain of salt, but given the other data devs have previously released and looking at the changes themselves I’d be a little surprised if the actual numbers didn’t go up more or less in line with the overall average from the site.
0 -
The problem is like you said is he is so awkward and he is not good as to compensate his difficulty. The devs are still doesnt fixed his animations for literally 2 years, his chainsaw much louder, a roar and a pointless overheat mechanic. If they at least make him smoother, im pretty sure more people will play him.
Post edited by Silasy on0 -
Voluntary self-submission is biased.
0
