This response pleases me.

I will just add that I do urge you to also consider a system with "stick" aspects other than Tribunal.

Tribunal is great for removing those with an overwhelmingly toxic attitude, but it seems like kind of a slow, heavy, expensive hammer not well-suited to dealing with players who aren't "the absolute worst of the worst". Even if you're periodically culling the most toxic 0.1%, the most toxic 10% can still have a pretty serious impact.

I'm glad to hear that Lyte and company are on the case, though. I have high hopes!

Totally agree. We're going to play in the positive space for awhile and do some heavy exploration and iteration, but we have a few things in mind we want to experiment with on the 'punishment' side as well. The Tribunal is ideal for targeting the absolute worst players in the game, but there's a lot we could do for the somewhat frequent toxic players that are never bad enough to get sent to the Tribunal.

Hey Lyte, while you're posting, I have a really random question I've been dying to ask for a while:

In general, how do peoples' normal Elos compare to their ranked Elos? So, like, if I had a 1250 normal Elo, would my Ranked Elo probably be 1250 +/- 50 ish? Or is it more likely to be a lot lower?

We're figuring out the flaws of the current Champion Select Lobby and I know how frustrating it is to navigate trades in the current game flow; however, just letting players trade pick orders is a pretty costly feature to implement and has quite a few concerns.

For example, do we give players X seconds at the start of a lobby to exchange pick orders first? This extends every single game by X seconds which actually adds up over time.

Do we allow players the same amount of time, but they can swap anytime they want? If the first pick swaps with another player but only 25 seconds are left... does the new first pick only get 25 seconds, or should we add more time?

There's a lot of small nuances like this that when you think it through makes pick order swapping seem like a very inefficient feature.

All I can say is there is a surprisingly high correlation between the two Elos

Huh, that's interesting. I always just had the feeling that a person's Normal Elo would be slightly inflated. Ignoring games with sub-30 players, I would have guessed it had something to do with the ratio of normal to ranked games played, especially with regard to the number of players who play like 75%+ of their games in either ranked or normals. But if they're generally very close I guess everything must be more balanced than I ever thought it would be.

Being a numbers person and not being able to see the numbers is tough

At any rate, I'm definitely not the first and certainly won't be the last to say this, but thanks for keeping this thread going / replying to other threads around here. Getting some insight on matchmaking and player behavior and such is pretty neat.

It's my understanding that '1400' in normal isn't the same as '1400' in ranked, and so it's not that their numbers are about the same, but they're positioned in the same relative position. Of course I could be wrong, but that's what I've gathered from various posts.

A multiplicative or additive adjust would still count as correlation. If you have someone's Normal Elo, it seems like you ought to be able to guess something about their Ranked Elo, and vice versa, provided you have a large enough dataset.

My guess is that the complicating factor lies in "approach to Normal games". I'd wager that players fall into a few different categories as to how they play Normals, e.g.:

(1) Serious Business -- play Normal like Ranked
(2) Casual -- use a broader range of Champions in Normal, and test out new builds, but otherwise play it a lot like Ranked
(3) Experimental -- play Ranked most of the time. When playing Normal, are just about always trying something bizarre, like triple-mid or Surprise Tank Teemo.
(4) Team -- play Ranked solo but Normals as part of a team, or vice versa

Ahhhh. All right, that's more in line with what I was expecting Lyte to say then. As you picked up on I just kind of instantly took "coorelation" to mean "the numbers are similar," as opposed to "the numbers have a pattern."


Yeah, definitely a good point. I thought of the serious business/casual distinction but I hadn't really considered that people would exclusively play premade or exclusively do weird builds in normals. (I usually do my weird builds in Dominion, ANYTHING works on Dominion).

Also, +1 for Surprise Tank Teemo, I kind of want to try that now.

hate getting stomped on by a duo