Jump to content
Sign in to follow this  
Dakado24

[WN8 Expected values] Process of automatic generation

Recommended Posts

Hi, since I do not have right to post in mathlab section I am posting it here:

I got a database of currently 113896 active wot players and 1517 clans updated daily and after 30 days of collecting data I have already 1757386 records in it. The database is growing every day as visitors of it search new clans and new players. I could publish a table of avarage values per tank which could help with generating wn8 expected values. Let me know if anybody is interested looking into my data.

Size 

Share this post


Link to post
Share on other sites

The issue that was occurring with wn8 which led to V30's class averages was all of the new tanks being released, rebalanced, and re-tiered. It would still take a while to get accurate numbers generated for new tanks, and wn8 is just not worth investing so much time into since recents are easily padded and per tank isn't relevant since DPG exists. If the only reason it existed was to analyze players applying to clans, then imo it's better to force recruiters to do a more in depth analysis of a players ability rather than just going off of a 4 digit number.

Share this post


Link to post
Share on other sites

Daily is somewhat useful for immediate data after a patch I guess (plus showing people how they're doing daily) but ~115,000 players really isn't a lot to generate expected values. How are those accounts chosen exactly? I assume it's just people who are in clans that are searched or specific players that are searched for, in which case the data would be slightly biased upwards (although it was biased in that direction when data from vbaddict was taken as well).

I'm still taking 30 day interval data (hoping to eventually automate it down to 14 days) of all active players worldwide (as of 3 months ago, hope to eventually get it down to worldwide active players monthly), which is ~8.7 million accounts. The nice thing about so much data is that you can somewhat accurately linearly regress expected assisted values for each tank from the account overall averages.

I'm on the 3rd interval now, I expect eventually to send the data through Eureqa and come up with a WN9-like formula to use as a 14 day recent w/ 3 month weighted average rating (much more about that in the WNR topic), but I just don't have time this summer. I expect to get something working maybe mid September or October.

I would share my data but each run is a little over 22GB.

 

The other thing to remember is that expected values aren't the same as the true average values on each tank. It's linearized by the upper 50% of players per tank, then adjusted if the top 10% of players in that tank fall to much above or below the top 10% of players overall.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...