Player behaviour and balance in League of Legends

Tuism

There was this article that started with the Red Team vs Blue Team discussion, which is the age-old is White stronger than Black sort of discussion. Then this part really interested me:

One interesting finding was that showing a red message about negative behavior during the loading screen lead to a much larger decrease in toxic behaviors (in terms of attitude displays, abuse, and offensive language) than did the exact same message in a white font. Additionally, showing a blue message about positive, cooperative behavior during the loading screen also lead to a decrease in a negative behavior, while no effect was observed for the same message in white. And, interestingly, when the question "Who will be the most sportsmanlike?" (a positive behavior message) was presented in red, the toxic behavior metrics actually all increased.P

This kind of stuff really comes out from big data research. Man I need to learn some analytics :)

The article:
http://kotaku.com/yes-the-blue-team-has-an-advantage-in-league-of-legend-1578533860

dammit

In considering starting my masters in psychology next year and taking up where I left off on behaviour research in online gameplay, this is the super interesting and inspiring stuff I would love to do.

Anyone want to let me experiment with their live game?

EvanGreenwood

We'd love to better foster co-operative behaviour in Broforce... But... Just getting the analytics data itself is a fair bit of work that we don't have much time for. Maybe if Steam integration makes this easier we will spend some time on this kind of thing.

Tuism

I'm really curious as to what the best and easiest way to get analytics would be from our indie perspective. Heck, especially er my perspective... With next to no experience in coding outside of Unity, Gamemaker, Flash AS2, etc.

Is it just plain... difficult and I shouldn't bother until I'm more experienced?

ShaneProcrastinates

That's pretty interesting stuff, although in that case it also depends on what skill bracket most of the stats where gathered from.

A higher percentage could come from mid-low ranked games due to general map awareness with the hud blocking partial vision during the laning phase, but at higher brackets this is not really an issue.

One reason blue was stronger than red last season though was due to the fact that you could kill the 2 jungle Golems before going to your botlane, having half a level up on the enemy teams botlane.

Another interesting thing is that it depends on the jungler habits in certain ranks too. Low-Mid ranked games usually start at "Blue Buff" every single game, and due to this the blue team gets to gank botlane first which could potentially get 2 kills as opposed to 1 at toplane.

Oh snap im rambling...

But yeah, mining those stats yourself is quite tough (And time consuming) as those title usually dedicate a division or person to "Research & Development"

EvanGreenwood

@Tuism We were saving stats everytime someone finished a level for a while. Within a couple months we had over 5 GB of stats and it was causing a severe lag when the stats were read... (Eventually we just started deleting the older stats so there's only a few thousand records saved on the server)

And we were only saving the time and the number of enemies killed (and one or two other stats) only when someone beat a level. It didn't give us nearly as useful information as watching a few Youtube videos of players and it was a lot of hassle to set up... We could tell which levels players were giving up on, but we couldn't tell why (and the level players gave up on the most is the level that is many players' favourite).

To really get insightful statistics we'd have to have set up a LOT more metrics and spend a LOT of time adjusting and studying them. And hooking in A / B tests like the red text / blue text tests LOL did (which is another system that requires maintenance and engineering).

And the information has to be more useful than Let's Play videos by enough additional utility to justify the hours spent tweeking the systems.

That said. There are Unity packages out there that save information to clouds and have a lot of the standard features you'll want. Getting analytics isn't hard, it's getting particularly useful ones that I think is a challenge.

I'd say, if you have a question that you really want answered (and only analytics can answer it) then consider implementing analytic systems... otherwise don't.

Although... If I were making a competitive game like LOL then I'd certainly consider analytics worth the effort. Analytics are really good at outputting data about the balance of the game, but not as good at outputting data about the player's experience.

Sorry about the unnecessarily long response. Long story short, I was more excited about analytics before I tried using them (and then found that using them properly was more work than I anticipated).

dislekcia

We collect quite a few stats for DD. There are some others that we've been collecting since the beta as well, but those are much harder to turn into user-facing data: I have to run SQL queries against them to get anything useful. Some of these stats have been helpful in determining preparation balance during the beta (and trying to find answers to the "did you really want to die?" button question), but we still got most of our balance information from player feedback and observed play.

I feel like A/B tests are only useful when you want specific outcomes you can measure and choke points in your game's flow where specific "interventions" make sense. The colour of loading text is a pretty strange thing to want to optimise, but Riot's been on a player behavior push for a while now, so anything that they feel can help that gets attention.

Fun stats fact: The stat collection database got so large that when loads of new players were trying to access it at the same time during the sale last week, it was constantly returning errors - this didn't break the game, but it did slow down the cloud saving and some other server functions. @Aequitas did some optimisation magic and sped everything up again, despite the player load not lightening for a couple of days.

@ShaneProcrastinates: I haven't read the source for that article, but I strongly suspect that Riot is controlling for player skill levels somehow. Blizzard does that when they talk about SC2 balance, treating different player skill levels as different strata of game balancing concerns. Riot's pretty slick, they probably do something similar ;)

Miltage

Tuism said:

I'm really curious as to what the best and easiest way to get analytics would be from our indie perspective.

I know of http://www.gameanalytics.com/ but have never tried it. Might be worth looking into if you want to track player stats.

ShaneProcrastinates

Actually I stumbled upon this article about Google Analytics in Unity , could be helpful to someone as it seems quite an interesting experiment to implement.

@BlackShipsFilltheSky: The more people play your game the more of a nightmare data mining becomes it seems xD.

@dislekcia: That's probably why they havn't made too drastic changes to the 2 different sides of the map yet, as it would impact every single skill bracket instead of just the one having an issue. Makes sense :)

EvanGreenwood

@ShaneProcrastinates To be fair, we went into datamining with zero experience. It's a field that some companies employ multiple specialists just to do.

dislekcia

ShaneProcrastinates said:

@dislekcia: That's probably why they havn't made too drastic changes to the 2 different sides of the map yet, as it would impact every single skill bracket instead of just the one having an issue. Makes sense :)

At the risk of turning this into a LoL discussion, they've definitely been doing a ton with trinkets recently. The whole 4-0 quick push fad and the mandatory buff swaps happened just because of a few small changes. Messing with the entire layout of both sides of the map is going to be a tricky thing to pull off elegantly.

Also, I feel like side advantages could actually come down to emotional loading too - especially in solo-queue. Wonder if they're planning "blue vs green" any time soon. I'd do that as a control point ;)