I do not understand why you think this would be a good system.
It favors 1 directional movement. It rewards players with low # games. Its strongly prioritizes early streaks. Then later it rewards mediocre players to just grind longer.
Where do you account for the change in skill over ladder?
Where do you account for player experience in the que?
Where do you account for uncertainty on a new account?
Why did you use your numbers? I don't get it. Lets make the formulas.
Winning adjustment formula.
LPN=LPO + (20+WR)
LPN is New lp.
LPO is old LP.
WR = (W-L)2
WR < 51
WR > 4
You did not provide me a soft cap for MMR loss. Im going to assume its inverted scale so max gains of 4 and max loss of 50.
The loss adjustment would be
LPN=LPO+(20-WR)
So. Hard caped LP gains at 15 wins over losses always. # of games wont matter. Once I have 15 as my whole number out of the W-L formula I see +LP=50.
215-200 for WL? 45-30? 17-2? All will see 50 LP gains off a win. The scale is not the same for losses. I cap at my losses at -16. So 8.
What's this mean? Even if I go 50% **forever **after I have 8 wins. I will be climbing ranked ladder. My LP loss is 4. My LP gains are 36.
Your talking about a loss representing 4% of the 100 to top of a tier. But a win is worth 36%... and all I ever have to do is just break even from the moment I have W-L=8.But you can generate Negative values from W-L. L being higher then W. This is then multiplied in your system by 2. So a win becomes -LP at the value of 11 games. Win? Lose LP. Lose? Lose more LP. Fairly discouraging to your players, especially given how small 11 games can be for a dedicated player.
So... To abuse your system. All I need to do. Make a new account or wait for preseason. Do my best to do badly in provisional games to get placed lower. Go on a streak of 10 wins then spend the rest of the year going win loss win loss. Because my LP gains are 10X higher then losses are and this holds true the rest of the season even if I just break even at 50%.
Every 2 games = 36 LP gained. Promotional matches don't adjust LP but as I break even im going to not see any change to my 40 LP gains off wins. So 6 games get me to promos. 2-3 Wins gets me out. 9 games to make it to next division. Add 1-2 extra game to return to promo after every fail + 2-3 for a pass.
Realistically i am climbing at a rate of Division in 20 games at a 50% rate, simply because I front loaded my W-L at the start of the season.
I could do it with luck as quickly as 9 per title. So.... Challenger after 244 games if I can just break even all season best case.. less if I do better then 50% after my first 10 games.
If I wanted to break your system all I need to do is try to abuse my starting games, the ones where inherently I am expected to be vs lower Tier players. And make as large of a win stock pile as I possibly can because then I can then try and rely on luck to carry my Slanted LP gains to the very top.
Pretty much any level of skill can do this and try to rely on luck afterwards.... so.... luck starts showing up across your Ranked ladder over skill.... which kind of defeats the entire fucking point of a rating system. Few early games give any one an opportunity to make it farther up the ladder then they have any right to deserve.
MMR is pretty simple concept. Skill impacts ability to win.
You use probability distribution. If an account matches that distribution it gets the respective rating for that expectation. For example. Chess and Elo's system. a 1200 vs 1400. The 1400 player is expected to win 3 to 1. End of the matches if that is proven to be untrue then the system provides an adjustment to reflect this. That way when players are accurately placed to the range of their skill you see their win%s start to reflect your systems expectations. By sorting them to a value that reflects their match results you eventually end up with their account reflecting system expectations. 2 players with a high # of games will increasingly represent the % between the two accounts your system expects. Thus Low uncertainty 1200 and 1400 will consistently represent your 3 to 1 odds.
Here is an abstraction for if you want to make a rating system.
Its like sorting coins. A small coin sinks into any gap it can, larger ones need a bigger hole. The smaller coins need to be sorted out first so they don't end up with your large coins. One of the reasons MMR does not start players at the very bottom of the ladder. Give small coins the time to sink.
you need to build your system so that the first Gap does not allow any coin to fit or your attempt to sort your change is pointless, everything piles into the first slot. Obviously your looking for skill not size when you make a rating system.
Tiers are just supposed to represent MMR ranges when accurate. They mean nothing when its on a high uncertainty account.
