how i would make ranked if i was riot

A variation of elo matchmaking but taking in account invidual performance inside a team, and a more rewarding ranked, let me explain : every players would start at 1200, nothing new there with a basis k variation of 20 ( something similar to lp), a win would be ''player elo''+k(1-prediction) and a lose would be "player elo"+k(0-prediction)...

Prediction would be calculated like this : 1/[1+[10*(("team a elo'' - "team b elo''))/100]], well something probably similar than the current matchmaking, but then...

k would vary each game depending on how well the player did compared to the playerbase in his tier for his champ in the role, 50% being the average, gaining +1 or losing -1 to his value for each 4% of variation on all variables averaged to a percent score , the variables calculated being :

kda(death in teamfights and executions being worth 0,5), cs score(not calculated for supports), enemy laner cs deficit, contribution to objectives (calculating damages from the lanes minions to the enemy turret), cc effects applied, vision placed, vision denied, win or lose of the lane according to matchup ( for example, if 80% win lane, count 70% if you win and 30% if you lose), here some additional rules :

A teamfight is counted if at least 3 players of a team gather in the same spot and fighters and tanks gain an additional variable for damage done vs taken by the team when there is a teamfight going on. Also a minimal ping would be needed for ranked. Now here's an additional variation on k:

Win and lose streaks would change the value of k but this would be easily movable to prevent tanking, here some examples of k increase and decreases according to match history, max values being +11/-29 or -11/+29 :

+20 +21 +22 -18 +22 -18 +22 -18 -19 +21 +22 +23

+11 -29 +11 -29 -29 +11 +12 -28 +12 -28 +12 +13 -27

In placements k would be different and win streak/lose streaks wouldn't have an impact, i would also do season start placements and mid-season placements, the second having less impact (oh also, it would go with the elo you had when you last played ranked, no resets), here's the k values for each 10 games in order season start placements : 300 150 145 130 115 100 85 70 55 40 mid-season placements : 150 75 70 65 60 55 50 45 40 35

and then k reset at 20 without impact of earlier win or lose streaks

I would remove the fact that tiers are divided by 5, these would just be in higher number and a little bit smaller in elo margin, each would be worth 149 elo and this would be some kind of lp without promos and demotion shielding, each would give a different border with the percentile of the player toward the tier playerbase (even the lowest) and a different victorious skin ( highest elo would also get the skins of lowest tiers) herse how i would divide this : wood : 1-150 stone : 151-300 copper : 301-450 bronze : 451-600 iron : 601-750 steel : 751-900 silver : 901-1050 electrum : 1051-1200 gold : 1201-1350 platinum : 1351-1500 crystal : 1501-1650 emerald : 1651-1800 sapphire : 1801-1950 ruby : 1951-2100 diamond : 2101-2250 legend : 2251+

oh and yeah, fuck the decay ;)

I know i should work at rito xd