Quote:
Originally Posted by JSkelts
Hi guys, hopefully this isn't too eSports for everyone but I'm looking for some guidance on improving my player projections for LoL DFS.
Essentially scoring is like this:
+3 Points per Kill
+2 Points per Assist (given to any player who helps in a kill, and there is no limit on number of teammates who can assist on a kill)
-1 Point per Death
Theres other stuff but this is what makes up 95% of scoring.
I have projections for a teams total kills and their likelihood of winning, as theyll only produce a useful score if they win.
Currently my projections are this, for each player:
1. Take their teams projected kill total (either from my model or Vegas)
2. Multiply that by a stat known as "kill participation", which is the number of kills + assists the player has over the course of the season divided by their teams season total for kills. I.e. if a player has a 75% kill participation they have a kill or an assist on 75% of their teams kills.
3. Multiply that number by 2 + the ratio of Kills / (Kills + Assists). Since Kills are 1 more point than Assists, this is just a way to approximate how much more often they get the kill than the assist given their stats on the season.
4. Subtract the expected number of deaths for the team multiplied by the players % of the teams deaths they account for the course of the season.
5(?). Multiply by likelihood of winning - Does this even make sense? In my head this is just an EV calculation of Score when they win * % chance they win, but maybe its baked into their projected kills already.
This is what I have right now, but I'm wondering if it might not be robust enough or might not not account for the outlier performances we're looking for to win GPPs.
Thanks for any input
1. You can get individual team kills over/under on pinnacle. No need to worry about win-probabilities then.
2. To convert the 50-50 over/under team kills to estimated kills for the team you would need to add a small margin like 0.6 as kills aren't normally distributed.
3. Do kills, assists and deaths seperately. What is a players kill percentage, what is the assists percentage (kill participation - kill percentage), what about death percentage?
4. Then multiply each of the above metrics with team kills (or team deaths) and sum up expected points.