Question on LP Gain and Loss

ZaionX·Jun 13, 2019·1 votes·1,538 views

I'm in Silver and when I win a game I get between 19-20 LP when I lose it could go between 16-20 LP. That seems fair enough you loss and win about the same amount of LP, but you gain slightly more winning. I have a friend in bronze though and when he wins he gets about 14-16 LP while he could lose about 16-20 LP.

I'm not very sure how LP gain works here. Naturally when your in a lower rank gaining points should be more easy the lower you are. When I was in bronze I remember getting as high as 26LP. What confuses me is how my friend is getting less LP then I when I am a higher rank. I'm just wondering if anyone would be willing to explain how LP gain works is there a certain system; like LP gain depends on how many games you play. I've heard of MMR and that is supposed to play a role in LP gain, but I have no idea how much MMR is alot or where to find out how much my friend or I would have. If anyone can answer these questions that would be helpful. Thanks ahead of time.

3 Comments

Kai GuyJun 13, 20192 votes

Im gonna start with the TLDR. Look at your current tier and compare it to your Op.GG "average team MMR" If your under or over, your gains are probably slanted. The more dramatic the gap between your current tier and match tier the more slanted you see your LP gains.

I have an assumption that LP gains lower with K factor. I cant prove it other then showing the logic behind it.Sadly riot does not provide much info on their LP system.

Now for a standard Kai guy wall of text about MMR.

I like Elo, and MMR. I studied up on both. I will provide my opinion built from what I know on the two common system types and explain what seems logical to me to implement.

Tiers mean nothing unless they are attached to a MMR range. With out some Range then there is no interaction between Tier and Skill. Player tend to believe there are 2 common ways to build Tiers to skill. Option 1 is tiers are made in relation to population size. You put X % of player base sorted into different Tiers. Option 2 is Tiers are made off Static MMR ranges. Players can move thru them any direction.

I favor option 2. It naturally will sort players by skill due to MMR being built of probability theory. Being a 1200 Rated player means nothing alone, in the context of an Elo system built off probability it simply means the system has expectations on what the odds of you beating other players are, this is made off their ratings. It wont care if its wrong. Being wrong is part of how Elo/MMR adjust, it expects high uncertainty (Higher K factor) on new players with no games played. Think of it as the system having a calibration period before it starts being accurate. I like this method, the advantage of this is that population never matters as extremely because your not looking to force a % of the ladder into any 1 division.

If I wanted to use option 1 and I had 100,000,000 players im looking to split that into population %s and that means I try to shove players below and above the range I set to be the 50% marker. My issue with the first option is i feel this system gets negatively impacted by players quitting or an influx of new players because it wants to readjust the ladders ratings by current population. Its sloppier and does not inherently represent skill better then probability's will.

So personally I would have each Tier represent X range of MMR. You can be tricky and adjust how wide that range is if you wanted to move % of players into a specific Tier with out ever being impacted by current active populations. You also have plenty of opportunity to make dramatic adjustments for inflation or deflation during Soft resets every year. MMR is next to impossible to make zero sum in a game like league far as I know.

Now. The advantage of this LP system riot uses? Many of them are perception. Players do better with titles then numbers. IM DIAMOND feels far better then im 1700 MMR (im using 1700 arbitrarily no idea if that's accurate to riots system.) Riot can punish dodgers with out giving them a lower raiting of skill. The one I think matters most thou, Riot can have a more Dramatic K value early on an account. That means a new account can go up or down the ladder faster, putting them out of a range they make mismatches. Promotions and demotion protection help remove some of the impact luck makes. A lot of folks bitch about promotion matches but they function very well in stalling Luck reliant players, which is any player who does not rely on their personal skill.

So your LP needs to move slower then K because high uncertainty means low confidence in your systems estimation of skill. So LP gains are slowed to not give some one the reward of a title they don't inherently represent. So you see it slant to react to your hidden MMR. Least, that's how I would run it.

ZaionXJun 16, 20192 votes

I think I get it, so in theory if my friend performs better and wins more often than maybe he can get more pts. Still seems odd to me though, usually with games if something gets too hard they give you some leeway or don't change anything. Its a little odd to see LP gain go down when that player happens to be losing multiple games. I do understand though if you are going to be in a certain rank you have to prove you belong in it. So by lowering pt gain and making it harder it ensures that if you make it to the next tier you're likely good enough for it. One concern I have though is like you said that sometimes people get lucky. Just as someone can get lucky someone can get unlucky and its kinda a shame that some people's LP gain may be affected because of a bunch of unlucky games. Though I suppose part of being a good player is working with what you got even if it means having bad luck. But hey what do I know I'm only a S2 right now, I still have a bit of a way before I can truly consider myself a good player lol. Anyhow thanks for the reply I appreciate it.

Kai GuyJun 17, 20191 votes

You don't just lower PT gain thou. You accelerate it as well.

Overall For Tiers to have any meaning to MMR they need to reflect a Arbitrary MMR range picked by the devs. Its an extra step added to MMR systems. It has some upsides in that you can punish or reward players with LP with out impacting MMR inflation.

So if your MMR is below your assigned Range. Expect lower gains higher losses, its trying to make your tier match your MMR. If your MMR is above your current tier, higher gains lower losses. You should have more or less balanced gains and losses.

My assumption is LP changes positive and negative get a bit lower with K factor, but I have 0 ability to prove this because I don't know where Riot lowers their K and by what values they do It at.

Far as luck? don't worrie to much about it. Probability theory is how you make MMR so its increasingly difficult to sink or climb off pure luck. direct impacts on match results then shift to move players to a better range that reflected the outcome.

No idea what Riots looks like but in a Elo system 1 class interval skill gap looks like 77% odds (200 Elo) and 2 class intervals is 91% (400 Elo). Higher rated player being favored to win, Lower rated would be expected to win 23% or 9%. Out of 100 games your looking at -54 and -82 for the MMR balance assuming players matched their probibilitys. If its wrong then that's where MMR is made to self correct and would shift raitings.

For example if we had those players just grind out games the system would start moving them till their MMR raitings matched their games results. If after 100 games it was 64% to 46% then the raiting should look like gap between 112-119.

the Exact values don't matter because they just imply the probability in comparison to other values. a Class interval is the same assumption of win odds per point gaps so that Ladder ratings reflect the results of accounts.

Most players never make it to far away from the starting Range, its a majority of the populace in the starting class because that's where normal players fit best. Not because the ladder is made from Population percentiles.