More global statistics

Feature Requests
1

Hello,

This website is a great project and I am grateful for all that has been done.
However, I do believe there are many interesting statistics to deploy with the available data, which is a gold mine in my opinion.

If I understand correctly, current global statistics show :

  • Killers statistics per pick rate and kill rate
  • Perks and builds statistics per usage

Well, killers statistics look good for me, but I guess perks and builds statistics could significatively be improved. The major reason why is, in my opinion, “popularity” statistics are not necessarly correlated with efficiency.

How about the following :

“Perks and Builds” page

  • Perks and builds statistics per individual escape rate (“Sacrificed / escaped / used hatch” distribution ?) / kill rate ?
    This would provide great analyzes in order to identify over and underrated perks and builds. I insist that point looks really important for me.
  • “Top Killer Builds” per killer ?
    Because some perks could be more or less efficient combined to killers’ powers.

“Killers” page

  • “Top Builds” per killer, and eventually split them by most popular / most efficient ?
    Same reasons as above.
  • Kill rate volatility ?
    It could be estimated over kill distribution and this should help to know which killers make “stable” outcomes or not. We could also condisder confidence interval so we know how trusty is that kill rate based on data volume.

Miscellaneous

  • “Maps” page where we could see kill rate per map, and eventually “killer x map” kill rate ?
    So we could know which ones are killer-sided or survivor-sided. But, if I am not wrong, that suggestion has already been made.
  • Maybe same “Killers” page for survivors ?
    Because I guess not everyone has unlocked all perks over all survivors so these statistics could show where people priorize their survivor builds.

Anyway, I could continue making statistics suggestions on and on, as those available datas give endless analysis possibilites. Like really. But I believe those ones above could greatly help us to learn the game and see interesting or even unkown aspects.

Thank you for reading, and I am open to any discussion on this subject !

2

Hey, thanks for taking the time to share your ideas and I’m glad you like the project.

The community stats are something that has largely been driven by the amount of data I have to work with (there’s only so much room to break things down before each stat is only made up of a couple of matches). It’s always been a bit of a chicken and egg situation as I know many people would love to see more there and that would likely result in more users/uploads/data to work with. Over the last few months the number of uploads has increased significantly (to the point where we’re usually over 1000 uploads a day) which has been allowing me to work on adding more/better ways to view the data. Map stats were the initial focus but as usual I’ve had some other ideas and the scope has increased slightly…

All that is to say there is a decent bit of overlap with what I’ve been working on and your list so this might be one of the faster feature requests I do :smile:

3

My pleasure !

Well, I see what you mean with that chicken and egg situation, and I guess this tends to happen when a project like that grows.

Great news about uploads per day ! But I presume you will still confront volume issues ? I imagine that most of NightLight statistics take very small data segments, especially if 95% of uploads are sort of same meta builds. So, in that case, I keep suggesting you to add mathematical indicators such as confidence intervals, as I mentionned previously, so we could know how reliable are results we are looking for…

Anyway, I do believe there are some work to do around data management ? I don’t know how it works exactly here, maybe this is fine, but I guess data optimization is strongly required. Today is a thousand per day, how will it be managed when it will double ? Maybe you already have an answer, otherwise I wish I could give a hand about that but I am not in data engineering stuff… But still, I keep being curious about that !