[enhancement] Start with correctly preordered in-game scoreboard to match lanes via machine learning

Salut girls & guys,

coming from a data science/machine learning background I'd like to propose a fun little enhancement. Higher elo players seem to love the new ordering feature of the in-game scoreboard. Usually they arrange it intuitively according to the meta (with home/opponents mirrored):

Top Jungle Mid ADC Sup

Although this is a task completed under 5 seconds, it has to be done every time a game is entered. So why not just write and train a classifier doing all this for the player by default. You guys have enough samples and feature engineering should be easy as well (e.g. summoner spells, champion chosen, team composition, first item bought, previously played lane). Training is done offline. Classifying would be cheap (this scales easily).

I wonder if you are not already doing all this for internal metrics to match players/champions to lanes. Judging from the careers page, you do. :) So this maybe is just a flick of a switch and an API endpoint.

Would love to see some open data of league games to work on some insights aswell (randomized partial set). Why not make a Kaggle challenge? You maybe gain some new talent.

Cheers