by L-Jimmy Thu Feb 10, 2022 2:15 pm
They're completely objective on raw data (which is good), but fail to take into account any context around scoring (injury etc), likelihood of changes to role or whether they may be selected etc - which pretty much renders them close to meaningless IMO
Any prediction before team lists isn't worth the pixels you read it on.
There is also very little point in predicting a change of role now. Because in most cases it isn't a prediction but a guess.
I wonder if they will revise predictions before lockout. Given the price they charge for coach, it would be nice if the data you were given had some use to it.
If you care about predictions just wait for a day or 2 before the round starts and hit up the daily fantasy predictions for a better idea.
It's also slightly interesting to note that fantasy predicts 445 players to score above 0 in round 1. That is equivalent to each team fielding 27 players.
The coach predicted score looks to be essentially ye olde pareto weighted moving average forecast
Last score * 1
Last-but-one * 0.8
Last-but-two * 0.6
Last-but-three * 0.4
Last-but-four * 0.2
Add all these together and divide by three and hey presto
Yeah, it's a dumb forecast. But Milch and ET are both right in that a) it's meaningless before TLT and b) depends on opponents, role, minutes etc.
The WMA forecast is a useful, very useful, proxy for autocorrelated series when there's lots of weird unknown factors. Autocorrelated just means that tomorrow is going to be quite like today, and the role a player gets (inc minutes, injuries etc) is probably going to be the like one they had last week. So WMA is a sensible, probably good, predictor for mid-late season.
But don't pay for coach, just plug the 20% bands into excel yourself: =(J19*H19+J18*H18+J17*H17+J16*H16+J15*H15)/3
Last edited by L-Jimmy on Thu Feb 10, 2022 2:16 pm; edited 1 time in total (Reason for editing : excel-ent)