Red Post Collection: More on Team Builder, FeralPony on unpopular items, EUW Issues for March 29th

Posted on at 5:12 PM by Moobeat
[ Update ] Added in more Team Builder discussion!

This evening's red post collection features Lyte with more Team Builder discussion - including the recent experiment and a fill option, FeralPony chatting about a few of the game's least popular items ( Ohmwrecker, Atma's Impaler, and Executioner's Calling ), and Rincewind with context on the recent EU server problems.
Continue reading for more information!

Team Builder Discussion

Following the recent Team Builder Experiment that forces captain's to invite a least one friend, Lyte returned to the forums to chat more about queue times, a fill option, and more!

For starters, here's a repeat of the TB experiment announcement:
"TLDR: Trying a new experiment to reduce Team Builder queue times. Starting today, players will have to invite at least 1 friend to their group to become a Captain. 
--- 
We hope everyone has been enjoying Team Builder so far; we’ve been closely monitoring the system and are currently seeing that there are too many Captains, and not enough Solo players available. Although about 95% of players are having queue times of 5-10 minutes or less (even doing unorthodox or creative comps!), this means that 5% of players are having unacceptable queue times. Players who are high MMR or low MMR tend to see these problems more frequently, and players who play during off-peak hours see this problem as well. 
In light of this data, we’re going to try an experiment in Team Builder. Starting today, players have to invite at least 1 friend to their group to become a Captain. This should greatly improve the Captain-to-Solo ratio, and reduce queue times overall in Team Builder.
While this experiment is running, we’ll be working on a feature to show Solo players what teams are currently looking for to fill their teams.

See you in Team Builder,"

He continued, reiterating the goal of this experiment:
"Right now, we're seeing way too many Captains--there are already too many trying to find other players to run a non-meta team comp.

This is an experiment to see if we can greatly improve queue times and the experience, and if it doesn't work it's a simple switch off and we'll try something else."
and:
"The problem wasn't that Solo Captain groups weren't filling up per se, it's that we have too many Captains. If there are too many Captains, there's simply not enough Solo players to fill the groups.

This is just an experiment. We're going to see how this works and impacts the system, and we have plans for ways for players to be "Solo Captains" if there aren't enough Captains--but right now, we have 2x as many Captains, so some teams can never be filled."

He continued, explaining the solo aspect of the queue has not changed:
"If you are solo, you can still play by joining other teams. 
This is an experiment, we want to see how this affects queue times and match quality. If this is positive across the board, then we'll keep this flow. If we see some huge red flags, we'll go back to the drawing board.

We expect that the average queue times will go down for every Captain and Solo, so this isn't just to help the 5%."

To address concerns about this experiment, Lyte reminded :
"As a reminder, we'll be doing our due diligence here. We'll be assessing match quality, queue times, and overall player experience. If there are any red flags whatsoever, we can revert the change and stop the experiment."

As for the popular concept of being able to select "fill" as your role , Lyte commented:
"We're working on a way for players who like to "fill" to see what is currently in demand by teams and specifically choose those position-role combinations to join those teams.

Regarding letting Captains change roles/positions in the middle of a lobby, it's something we're considering but more complex to do."

When asked to comment on solo players and the input they have on what team they join, Lyte noted:
"We're interested in giving Solo players more input into what types of teams they join, but haven't figured out exactly what we want to do there yet. We'd rather give Solo players more input, than have an influx of Captains.

We are currently working on ways for players who like to fill to see what's currently in demand by teams though--this should improve the queue experience as well. At the same time, we'll be firming up the accuracy of all the estimated wait times for every champion-position-role combination, so players have the right expectation of how long they might have to wait to get into a team."

Lyte also dropped off a note that the wait time estimation & approximation are current inaccurate:
"An additional update: Estimated Wait Times and Approximate Wait Times are inaccurate in Team Builder. We're working to actively fix this to be more accurate. As the system collects enough data to get proper rolling averages, we should see these estimates become accurate and better match player expectations."

Speaking of problems, he also commented  on Team Builder games not adding to your normal wins total:
"Quote:
Though one thing I've noticed... any and all Team Builder matches don't add to your Normals stats on your profile page.
We'll be fixing the Normal stats in a future patch."

[ Update ] Lyte's back on reddit talking more about the upcoming fill feature!

To kick things off, Lyte replied to someone curious if the team plans to expand team builder to ranked play:
"We'll take what we've learned and apply it to other queues. But, let's first nail down the design of Team Builder's current ruleset."

When asked what the hardest thing about making Team Builder was, Lyte commented:
"Hm, there were two difficult aspects of the system. 
One, doing due diligence with the research and determining what the actual problem spaces of Champion Select were instead of immediately jumping to the 'obvious' design solutions for such a feature. Ranked Team Builder is in a similar situation where there is an 'obvious' design that might work, but if you deep dive through the system design and thoroughly consider it... it's actually quite a complicated space. 
Two, from an engineering perspective, creating a system that requires the level of scale necessary for League of Legends is always a daunting task."

In response to the suggestion that the queue times for each role should be listed individually, Lyte commented:
"Our designs are somewhat similar to this already. You'll see shortly!"

Lyte also commented on how Team Builder's MMR works intitially:
"Initial TB MMR is seeded from Normal MMR, and then becomes a separate MMR."

When asked about the possibility of queuing with multiple roles / champions, he noted:
"We're interested in discussing these ideas and are brainstorming on them internally. 
Pretty quickly, we realized that multi-champion select is pretty bad for numerous reasons. However, there's potential in multi-role selection helping with queue times."

As for a feature allowing players to "fill" a specific role instead of preselecting, Lyte elaborated and also commented on multi-champion selection for the same role:
"The fill feature will allow players to see what teams are currently looking for the most, and select that spec to join those teams. 
We don't want to allow multi-champion selection for the same role because it hurts champion diversity. Imagine this: as a Captain, you might be OK playing with Shaco, Lee, Pantheon or Elise as Jungle. But, when presented with all the options at once, you're forced to think about what the most powerful Jungler is, and you might feel bad for choosing one of the other options. 
That's a decision making process that isn't fun, and if many Captains are compelled to choose the 'strongest' jungler, whatever that might be, we'll just decrease the odds that some of the fun Junglers that are just as viable end up in games."

He continued, mentioning the dangers of rewarding players for queuing as a specific role:
"We're working on a way of showing players what is currently in demand; however, we have to be careful with rewards. 
By giving out rewards for playing something like Marksmen-Bot, you're changing the player experience quite dramatically. You're making some players think, "Well, I don't want to play Marksmen, but, I will for the reward." By doing something like this and changing the decision-making process, you lose a lot of the advantages of Team Builder. 
You want players to play something because they want to, not because they are being paid to."

When asked if Riot plans to launch Team Builder in China, Lyte commented:
"We're actively working on a version of Team Builder for China. As we make progress, I'm sure there will be news about it for all Chinese players. 
As you can imagine, there are some unique challenges to working with a platform as popular as League of Legends is in China."

Unpopular Items

FeralPony has been seen galloping around the community beta, chatting about the state of a few unpopular items - such as Executioner's Calling and Atma's Impaler.
"We agree they aren't in a great spot though straight buffing any of these has it's own set of problems. I would imagine we'll see changes to these at some point but not in the short term. Our focus on itemization atm is primarily Support and Marksman itemization."

He continued, sharing his opinion on Executioner's Calling and it's Grievous Wounds passive:
"GW isn't super effective against sustain lanes since you need to keep it applied at all times which makes it generally an ineffective counter. It is most powerful and useful against characters with high burst in-combat sustain like Swain and Mundo. I know the design team as a whole isn't entirely happy with how the mechanic is behaving and used. 
It is the most relevant item of the list to be addressed sooner rather than later though as it has some overlap with Marksman/Carry itemization and those are more our focus atm. It does need some adjustments but it's almost more related to how we want to handle LS/Regen champs like Mundo, Swain, Warwick, etc rather than a problem directly tied to itemization. 
I think EC feels more needed than it is actually effective, which is another problem, but generally I feel it's a bit weak and also ineffective at it's anti-sustain role. 
Overall I rate it a :("
When probed about some previous comments on edging GW out and lowering sustain in general, FeralPony commented:
"I think is an approach something Zenon and several other designers feels passionate about (which is not surprising in Zenon's case as he's working on Warwick atm). I'm not personally in this camp but I also don't feel GW is a huge problem on live atm as some of them do. If this is an issue they want to tackle/address I certainly welcome it, I think there are a lot higher impact things I should be focusing on."
Continuing the EC and GW discussion, he commented on the problems of shifting it into a counter item for laning phase:
"The situation with Ignite is that despite having a long cooldown it is pretty much always available when a large fight or multi-person fight occurs. This is partly why characters like Mundo and Swain are extremely durable in smallscale skirmishes but pretty much always have GW active when a teamfight erupts. 
Where I think the miscommunicaiton occurred is you're viewing it from a laning/early mid-game perspective while I'm looking at it primarily from mid-late. 
In terms of making EC a lane counter item we'd have to make it exceptionally cheap, and rushable which has a bunch of other side effect such as limiting what stats we can / should put on it. In addition, I'm not sure how often you would actually need to rush anti-regen counters in the current scene or even how useful it would be. 
Not ruling anything out mind you, just trying to think through some o the side effects. As for the GW internal debate, I don't feel particularly passionate about it in either direction, I'd like the more invested designers try and figure out what direction they want to take the mechanic before I do any substantial item work."

As for Atma's Impaler, he commented:
"Atma's is not only a case of priorities but an issue of what problem(s) do we solve by "fixing" or reworking that item? In the case of Atma's if we buffed it to the point where it competes with other fighter itemization paths all we've done is substantially buffed the fighter class which is the dominate role atm. Some of the itemization adjustments this patch were designed to better equip the traditional counters to fighters - Heavy AP champs and Supports. 
While the goal of make every underpurchased item better is a reasonable one, we don't want to have it fight against the larger goals and problems of the game currently."

When asked about Ohmwrecker, he commented:
"Yes, but again you need to look at the high level goals and problems in the game. We could certainly add a defensive option like fortify to Ohmwrecker . However, when one of the most common high level complaints is that games are being stalled out and people have an inability to close out games adjusting this item seems to make this issue worse not better, which is why instead we chose to look at Boot Enchants patch since Homeguard seems like a likely culprit. It obviously isn't the only factor but if we adjust it we can see what else is at play. 
If we had the opposite problem of people closing out games way too quickly, diving with impunity and breaking towers with ease I think looking at Ohmwrecker would be high priority as it would offer a solution to a current problem. Atm the solution you propose would accomplish the goal of "Make this underused item more used and interesting" but it would fail at making the game better as a whole and could actually make it much worse. Though of course this landscape could easily change in future which could then make your approach incredibly valid. 
It all depends on the current context of the game. While I agree with your goal, none of these decisions or priorities should be determined in vacuum, always try to keep external factors in mind."

EUW Issues: Update March 29th

Here's Rincewind with an update on the EUW problems that surfaced on March 29th:
"I made a 'post the other day' about the recent issues of EUW. Here is what we have been up to for the past couple of days: 
1) We made a few tweaks to the startup scripts of the memory caching software to prevent automatic restarting of the out of memory nodes. 
This change has already effectively saved us from having a full platform outage when nodes go out of memory and is now limiting things to a smaller outage that lasts a couple of mins. One downside of the small outages is that there is still a large login queue that lasts 15-20 mins before everyone is able to get back in. 
2) Last night we did a brief maintenance to trim down some excess capacity that we added on March 6th. We didn't remove all the excess capacity, we just brought it down by a notch to bring things closer to the way they were last month.

3) We have identified an inefficient code path that indirectly causes the login queue by dropping players when the memory issue happens. We are refactoring that code path now.
Picture a juggler with many items in air and he/she is interrupted, causing him/her to drop some items. That's kind of what's happening here. When the system freezes due to the memory issue, the inefficient code path becomes a blocker of all messages from the players and therefore dropping them resulting in a login queue. 
4) We are running a load test of a new version of the memory caching software and some additional improvements of the code for the next patch for more than 24hrs straight in a test environment that matches the production environment of EUW to make sure they are ready to be released. We want to make sure they don't cause new issues before we put them out. 
5) We are staggering quite a few changes over the next few days. The reason for staggering the changes is to understand the effects of each change, rather than lump all of them together and not be able to pinpoint what change caused how much of improvement or potentially cause even more problems.

I will post periodical updates like this when we have them. I will also be around for sometime to answer your questions."

When asked about how the Amsterdam datacenter is coming along, Riot Draggles commented:
"I'm working with Errigal and our localisation team to get an update out from our March progress soon!"

As for problems affecting EU on today, the 30th, Foro commented:
"Hi Summoners! 
Over the past few days we have seen a sharp increase in Denial of Service type traffic incidents towards EU West. These incidents have not been widespread enough to impact the whole of EU West, but enough that small groups of players may have felt their effects.

We've been working tirelessly to mitigate these attacks and have put some more safeguards in place. I'll keep you updated as we progress."

No comments

Post a Comment