Jump to content

How Stats Are Gathered And Archived


No replies to this topic

#1 Congress Hating Ur Freedoms

    Member

  • Pip
  • The 1 Percent
  • The 1 Percent
  • 10 posts

Posted 16 April 2018 - 11:20 AM

Hi I'm a new-comer to big data development and I have to say... jeesh... in big corporations it seems like friggin no one actually knows what they're doing. I'm one of the only former app devs on the team who does tech stuff as a personal life choice too, and (with no one to ask) I'm wondering how big-data volumes of data can be made available to real-time apps like a website or a game client. So what has PGI's approach been to data archiving, and data exposure over the years?

I take it based on how match maker 'works' that
1) Sure, this is a complex game with matches that have difficult to predict outcomes based on player rosters and/ or,
2) PGI invested more time and energy in areas other than in-game metric archiving such that you can't at this point easily apply AI/ ML techniques to help balance game rosters before the match and really give players that adrenaline rush they only feel from participating in those few-and-far-between "close ones."

I totally get it, it's 12 v 12 games... that in and of it self leaves a lot of room for error in outcome prediction, I mean, I try 99% of the time, but I totally observe that to NOT be the case when I drop with friends, lol... Hopefully smaller teams can clear this up without creating the disturbingly toxic community somehow cultivated by CS:GO's psychotic 5v5 match maker formula.

But anyway, I was wondering if I could get a response from the devs regarding any of these 4ish questions about data handling.

How much detail in the match is registered by the client, but doesn't get all the way to the server? (this is probably just a few things like mouse look.. holding ctr, ummm.. maybe weapon grouping stuff? but probably very few things.)

How much detail is registered by the server, but is lost in the 'archiving' process? This is probably (but hopefully not) stuff like knowing which components of each mech survived the match; knowing their exact remaining HP per component; knowing which player did what damage to which component; knowing how many pieces of equipment actually survived, etc. etc.

My last question is how much additional detail can be added to the private lobby match API that isn't there already (such as the stuff described in the last paragraph).

I ask these questions because really I'm curious whether you're using purely relational-databases behind the scenes or if you're using message queues and send the data in various directions so that it can be archived, applied to match maker, and also deal with the event stuff with out you needing to limit the volume of data that is being gathered.

Speaking of with, I guess my last question is, remember back in the day when there used to be a large delay in events being marked completed, whereas now-adays there's virtually no delay? What's the story behind that, anyway?





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users