Boards›[ARCHIVED] Miscellaneous

Future of ranked?

T RexHasTinyArms··2 votes·349 views

So I posted a question about some of Lyte's comments recently regarding ranked team builder and the feedback I received was that I probably read it right that because ranked team builder will be based off current ranked MMR, it will probably be the same ladder so they don't have to code a new ranked ladder in.

There have been discussions about how ranked team builder will impact role selection and wait times; if support will have an incentive or be worth playing more for the shorter que time, etc. However, a different question occurs to me if they are looking at making ranked TB the same as ranked instead of a separate mode. Does the community believe that the different roles have close enough to the same influence that it won't skew ranked? I was thinking how it might be worth playing other roles for a shorter que, and then I realized that I wouldn't play something I thought might risk dropping my MMR and win rate. It kinda follows that unless you're much better at a role that has less influence, the only role that would be worth queing for when you're seriously trying to climb is whatever role you believe has the most impact on the game. If the community largely agrees that mid for instance is the best role to carry from, then people who que for other roles may be split between mains of those roles (who think they can get higher rank w/ less influence due to greater skill in a less influential role) and people who just aren't concerned w/ influencing the game as much as they can and are maybe less serious in their mind set right now. Do you guys think this might be an issue? Also how frustrated would you be if people qued for roles they weren't good at to practice, or just to play really off meta stuff for fun, because they know they can just spam their main role and gain the LP back if they lose?

11 Comments

Matthias91192 votes

That's a post that needed more than one paragraph break. You changed topics like eight times in that second paragraph.

Do you guys think [some roles having a bigger influence on the game than others] might be an issue?

I think that if they go with a straight-up "you can always queue for exactly the role you want" system, it may be an issue at some MMR ranges. At lower ranks, it's easier to carry games by playing snowbally champions and exploiting your enemies' mistakes, so it may be easier to, say, climb out of Bronze spamming Yi jungle or Zed mid than, I dunno, Leona support. But I also think that this evens out at higher ranks, and people tend to underestimate the impact of a good Support even at low MMR. Really good supports can basically auto-win botlane when they get rolling.

Also how frustrated would you be if people qued for roles they weren't good at to practice

At some point you have to bring other roles into Ranked play, unless you plan on being a one-trick-pony forever.

or just to play really off meta stuff for fun, because they know they can just spam their main role and gain the LP back if they lose?

People can do that now, but yeah, it might be more of an issue. But it may even out because there will be fewer players being forced into roles they're not good at in lower MMRs. I'm not sure if Riot will apply the same sort of per-role / per-champ MMR adjustments that they do now in Team Builder, which might help with players who have a very different MMR in different roles. Or even rank you by role or something crazy like that.

T RexHasTinyArms1 votes

:D yeah it was a bit stream of consciousness, my bad. You should have seen how much I had to edit my 80 page master's porfolio for that kind of stuff.

I had hoped if they did ranked TB they'd do ranks by role bc that would be the most amazing thing ever, but I think if they are even considering (or being haphazard w/ their statements enough that they imply they might) making ranked and ranked TB the same ladder then there's no way they'd do that. The only reason I can see to make ranked and ranked TB the same ladder would be to save effort in having to make a new ladder or address the difference between the two in any meaningful way. If I'm right about that then it is unlikely that rito would put in the extra work to give us the nice features you mention at the end there.

Lastly, my point was that the way it sounds from Lyte's comments on ask.fm is that you could just always que for your favorite role. Technically I think they said you'll que for 2 roles as primary and off role, and if it gives you your off role too many games in a row the system will give you higher priority for your primary role. That being said I just spammed Heimer top and mid and it got me to low gold even though I probably play 20-30% of my games in other roles, and some of those games I just get crushed by someone who mains the role and knows the match up far better than I do.

If you drop those 20-30% then the climb would have been easier for me, and maybe I'm just a dick, but knowing that I won't drop from gold w/ silver MMR I'd play stupid stuff support that is technically doable until my MMR dropped to mid/low silver. It wouldn't be the same as now, bc I'd know I can get my MMR back up fast if I want to by just only playing my main role and having no off games. You'd be surprised how much winning in large streaks speeds up MMR and LP gains, so not dropping games in off roles would actually make a huge difference in how easy returning to a higher MMR would be if you play things you are bad at and drop.

Matthias91191 votes

The only reason I can see to make ranked and ranked TB the same ladder would be to save effort in having to make a new ladder or address the difference between the two in any meaningful way. If I'm right about that then it is unlikely that rito would put in the extra work to give us the nice features you mention at the end there.

There's more to it than just the devs being lazy. They're slow in delivering things sometimes, but they're not lazy. You really don't want to screw things up badly when you have a hundred million players.

Fragmenting the Ranked SR queue is a real concern. That's where the glory is, for better or worse, so if there are multiple queues that give the "same" rank/rewards then people will try to exploit any differences between them. Queue times are already pretty bad in the upper tiers SR, and splitting the pool of players into multiple queues makes that worse. Unless they have a way to pit a "team builder" team up against a "solo queue" one. But that seems unlikely to be workable unless "ranked team builder" is literally just letting you discuss strategy/roles with players before going into the same draft as you do now. (Which gets suggested all the time and does not fix many of the things that make people frustrated about solo queue now.)

You'd be surprised how much winning in large streaks speeds up MMR and LP gains...

I'm almost certain that win/loss streaks have no effect on your MMR gains, and that LP gains/losses are determined by your MMR. If "streaks" mattered, there would be all sorts of weird corner cases where Riot would have to manufacture "bonus" MMR out of thin air, and doing that over time would cause inflation of the ranking system. It also makes streaky players bounce around between ranks MUCH more, which is undesirable if they're still winning at around a 50% rate overall. It's much simpler not to do anything like that, and not really beneficial except maybe for players who you think are far more skilled than their current MMR indicates (ie, smurfs).

T RexHasTinyArms1 votes

I had forgotten about the first issue. It is something I had thought about as an issue with making ranked tb in the first place bc I think playing different games in one mode will be an issue and it changes what is needed to succeed, so in some ways it is a different game. Really good point. I hadn't meant to imply lazy, but the extra stuff would take time to not be bug free so it probably wouldn't be done. If it was it would come later.

I should have left lp out of the second statement, bc it is entirely MMR that determines it. However, streaks definitely influence MMR. However, it has to be a statistically significant streak. I have a second account from leveling up with a friend who started playing.bi placed both of my accounts in bronze at the start of the season (out of practice and was legit bad). I started playing the second one again after improving and getting to gold on the first. I was playing with mid silver players while in bronze 1 bc my MMR was high from not losing much (so the system moves it up fast to see if I keep winning more than I should and it takes time for rank to catch up. If your MMR drops low but rank can't drop then you just have to win in a row greater than is statistically even an unlikely possibility to be chance. That is why not dropping a single game is important to fast MMR gain.

Matthias91191 votes

However, streaks definitely influence MMR.

Correlation is not causation. Most Elo-like rating systems do not use any kind of "streak" system, because it's not possible to do that and guarantee that rating gain/loss is balanced in every match. Everything I've seen about MMR from Riot suggests that they attempt to avoid rating inflation.

Also... if you were playing against mid-Silver players while in Bronze 1, you should have had LP gains/losses like +25/-10 and been in promos constantly. You almost always hear stories the other way around -- it's easy to gain LP and "rank" quickly in a win streak, but after you jump 1-2 divisions it gets harder and harder to climb because your MMR only adjusts by so much per match.

T RexHasTinyArms1 votes

Perhaps I misunderstand how MMR works. I had always assumed it works similarly to how the computerized gre gives you questions. For the gre you get base line questions and when you answer right it gives you harder questions. When you answer wrong it gives you easier ones. Eventually it figures out the level you are at fairly accurately. To be efficient it moves you faster if you keep getting them all right or all wrong because it assumes its estimations are farther away from your skill and it can safely move you farther without overshooting. From everything I have seen discussed and experienced it seemed logical that a system based on balancing skill for players would work similarly. If you never lose in 100 games and the system doesn't put you against significantly harder opponents then there is an issue. Then again I wouldn't consider anything less than 10 wins a real win streak, and less than 15-20 probably isn't outside the realm of being common across as many games as are played each day. If you aren't against at least 2 divisions higher in terms of opponents at the start of the streak and the end, the game is adjusting too slow bc it is just letting you win and not making the game harder.

Matthias91191 votes

The MMR system they used is (AFAIK) a modified version of the Elo system (https://en.wikipedia.org/wiki/Elo_rating_system). Typically these use an initial "provisional" period where your rating is adjusted up or down fairly rapidly, and then rating adjustments are based on the difference between the rating of the opponents in a match or event. But there are many ways to modify it.

For instance, the TrueSkill system that Microsoft uses for XBL is similar to Elo in principle, but has an explicit "confidence" factor that adjusts how quickly a player's rating can move up or down. Beating players much higher in rank than you or losing to players much lower in rank than you will decrease the "confidence" factor of both players involved, causing their rating to adjust more quickly.

Ranked League matches, however, rarely pit players more than +/- 2 divisions against each other, unless there is a duo further apart in rank than that. And the MMR difference between the teams overall should be very small, so unless there is some sort of "streak" bonus, your MMR should adjust up/down by nearly a fixed amount per match.

Like I said before, I think they don't do a "streak" bonus because there are situations where it causes explicit rating inflation. Like if several players on one team are on a winning streak and nobody on the losing team is on a losing streak, then there would have to be a net increase in MMR from that game compared to one where nobody was on a winning/losing streak. Maybe it would balance out between players on winning streaks and players on losing streaks, but that's hard to count on. It also feels like it would reward/punish short-term winning/losing streaks too much -- unless it only started to kick in with extremely statistically unlikely streaks (like 15+ in a row as you said). Even with a 50% chance of winning each match, about 1 in 1000 players will win all 10 of their next 10 matches.

If you never lose in 100 games and the system doesn't put you against significantly harder opponents then there is an issue.

High-rank (like high Plat or Diamond) players doing "climb out of Bronze" challenges typically have an 80-90+% win rate in Bronze and Silver and are back in at least Gold in 50-75 matches at most. e.g. here's a player who did a "Bronze to Diamond" climb and it was only about 300 games total: http://www.reddit.com/r/leagueoflegends/comments/1xz6j2/the_conclusions_of_my_bronze_5_to_dia_1_journey/

T RexHasTinyArms1 votes

The second part was kind of my point, in that high win % moves up pretty fast (so if you screwed around and dropped your MMR it would come back up that fast especially because there is no block of promos to get through or anything, just getting back to playing against people in division 5 of whatever tier you're in rather than lower tier players). I'm kind of surprised it takes that long though for high plat and diamond players to move up given how crazy a 90% win ratio is.

The firs part is really cool, and thanks for the explanation, that makes a lot of sense. It's not winning in a row that makes a streak and pushes you faster (that's just in promos where it might do that), but that by never dropping a game you move up by X each time and never down, so the additive effect creates the illusion of moving faster than you expect. Although I do want to point out that the 1 in 1000 players number is slightly misleading. It assumes a constant 50% probability to win, which in an ELO system (if I understand it right) would in theory only exist at the perfect ELO system for someone. If 10 wins in a row bumps you up maybe a full division (low silver 4 to low silver 3 for instance) because you only add MMR for 10 straight games, then the probability of winning (if MMR actually does relate to skill) can't be consistently 50% unless we are considering average and the player is below perfect elo at the start and above at the end so they swing from 55% win chance to 45% (which I think would still mess up the stats). Therefore, assuming a player is "where they should be" at the start of a streak then the X in Y stat would be a lower proportion than could be assumed with a 50% win ratio, because as they climb above true skill representation they should become a weight to the team and have a less than 50% win probability that increases as skill difference increases. It sill would be too probable though sub 15 win streak though at a minimum probably (depending on how much MMR is gained per win, obviously if 10 wins in you're up two divisions from where you were, you should expect a loss very soon unless they are much more skilled than their previous MMR).

Matthias91191 votes

I'm kind of surprised it takes that long though for high plat and diamond players to move up given how crazy a 90% win ratio is.

Unless you're way above the target MMR range for a division, you're only going to gain ~25LP per win, and then you need to win two in the promo series. So it's something like 5-7 wins per division at a minimum (unless you skip one or more along the way). Even at a 90% win rate you're looking at probably 25-30 games to climb a full tier. And obviously once he got to Gold and above he wasn't winning nearly that many matches, which is why it took 3x as long for him to go from G5 to D5 as it did to go from B5 to G5.

Although I do want to point out that the 1 in 1000 players number is slightly misleading. It assumes a constant 50% probability to win, which in an ELO system (if I understand it right) would in theory only exist at the perfect ELO system for someone.

Yes, you're absolutely right. No matter what the player's starting rating, winning multiple matches in a row will put them up against harder opposition on average, and this would tend to decrease their chance to win. So they might be 50% to win in the first match, then 49% in the second, then 48%, etc. I don't really feel like dragging out a statistics textbook right now, but you can mathematically work out the likelihood of things like that, or work backwards and try to estimate what their rating should be given a winrate over time. They've given some indication that they do things like this when trying to detect brand-new smurf accounts at low levels, but it seems unlikely they would want to monkey around with Ranked too much.

T RexHasTinyArms1 votes

Ha, yeah the lack of desire to do stats is why I was being general. I would have felt like an asshole if you actually did even a remotely accurate approximation of something like that because no one should spend time doing that.

You seem pretty chill man. If you ever want to play ranked (or bug me to play something other than my normal ranked, I've kind of been enjoying the aram I had to do for event icons), hit me up. You can add this account (T rexhastinyarms) or the other one I have from leveling up w/ a friend (T rexhaslonglegs). Thanks again for taking the time to teach me a little about the MMR system, and have a friendly back and forth about it.

Snowball4u1 votes

I hate this idea of ranked tb

If i could get my main role every game i would be high gold, low plat by now.

But im not very good at the other for and i just do meh every time im put there so ya...

More from [ARCHIVED] Miscellaneous

Open Letter to Parents of League of Legends Players
[ARCHIVED] Miscellaneous860 comments
2,611
votes
Rewards for positive play
[ARCHIVED] Miscellaneous19,712 comments
1,337
votes
I don't like Boards
[ARCHIVED] Miscellaneous324 comments
809
votes
STOP ALLOWING LEVEL 1 ACCOUNTS TO SEND FRIEND REQUESTS!!!
[ARCHIVED] Miscellaneous291 comments
679
votes
Wanting to do something special for an awesome friend. Riot, please help? <3
[ARCHIVED] Miscellaneous220 comments
614
votes
People that are telling others to kill themselves has to stop! Please help me get this noticed!
[ARCHIVED] Miscellaneous605 comments
600
votes

More by T RexHasTinyArms

Even if you don't do anything else Rito....
Gameplay46 comments
142
votes
You know what champion sucks?
Gameplay53 comments
30
votes
I salute you C9
Esports10 comments
28
votes
You know what's so terrible about season 6 changes?
Gameplay13 comments
11
votes
Can we get a "damage from champions" stat?
Gameplay2 comments
9
votes
Am I the only one disappointed?
Esports8 comments
8
votes