Read on Twitter
Not sure I can do this justice, but: "predicting points": a thought
Productive players are more likley to be more productive
That's it. That's all we know. We're trying to find who was productive, and who wasn't
"Why does it get so complicated and why is it so hard then?"
Productive players are more likley to be more productive
That's it. That's all we know. We're trying to find who was productive, and who wasn't
"Why does it get so complicated and why is it so hard then?"
Consider why one of the best analyst's @ScottBarrettDFB spends so much time telling us to pay attention to what describes points
His chart
His chart
Scott emphasis that expected points (volume) is very important
Why? Because it does a better job of saying who was getting the opportunity for points
It essentially tries to adjusts for "luck", why?
Why? Because it does a better job of saying who was getting the opportunity for points
It essentially tries to adjusts for "luck", why?
Fyi: Ppg is one of the stickiest stats (explains the variance in itslelf and other stats more then most things"
Rounding up, let's say it tells us 40% of what will happen next year
Ep or fpoe ect tried to tell us who got lucky (the we argue if it's because they are good or not)
Rounding up, let's say it tells us 40% of what will happen next year
Ep or fpoe ect tried to tell us who got lucky (the we argue if it's because they are good or not)
Fyi somewhere a nerd is screaming eveytime I use rsq as "predictive" instead of "explaining varience"
What's the other *60%*?
variance, basically
variance, basically
Ppg (or a better stat) is 90% of what we know, and it's only 40% if what will happen
Conversations essentially split between what is "really" the best description of that 40%
And what that other 60% will look like because...
Conversations essentially split between what is "really" the best description of that 40%
And what that other 60% will look like because...
Because career arc:
@AmItheRealBlair has shown that Efficiency early on is a good bet to encourage more volumethe next season, if it's on a threshold of touches
@AmItheRealBlair has shown that Efficiency early on is a good bet to encourage more volumethe next season, if it's on a threshold of touches
Because regression:
From @TJHernandez redzone Efficiency in points to @friscojosh's airyards model or @RotoVizDynasty
Fpoe
We try to find players whoes stats may (more likley) move up the next year/week through underperforming volume
From @TJHernandez redzone Efficiency in points to @friscojosh's airyards model or @RotoVizDynasty
Fpoe
We try to find players whoes stats may (more likley) move up the next year/week through underperforming volume
Because team:
"vacated opportunity" and "team situation" changes or intention
We are worse at this, which is why I've made good impressions (I think) specifically pointing out how and why we are bad at this
"vacated opportunity" and "team situation" changes or intention
We are worse at this, which is why I've made good impressions (I think) specifically pointing out how and why we are bad at this
But that's about it
We know being productive is the best indication of being productive, so we try to find the best way to identify who was productive
And guess at who will be (rookies and team changes)
We know being productive is the best indication of being productive, so we try to find the best way to identify who was productive
And guess at who will be (rookies and team changes)
So, when some stats are proven better then other's
Dominator @BlakeAHampton or yptpa @LateRoundQB for ex
That's essentially saying they describe that 40%, better then ppg or yards ect
Dominator @BlakeAHampton or yptpa @LateRoundQB for ex
That's essentially saying they describe that 40%, better then ppg or yards ect
Tape analysis also does this, imo
because "he was better then his stats", or "his oline was bad" or "their d line is terrible" folks are trying to describe that 40% better and add future regression on top
because "he was better then his stats", or "his oline was bad" or "their d line is terrible" folks are trying to describe that 40% better and add future regression on top
But next year will also have its own luck
This is where a small sample sport like football gets a little more tricky, a larger % is going to look like, or remain, just...varience
Idk if this is helpful, but it seemed like a good thought when I started this thing
This is where a small sample sport like football gets a little more tricky, a larger % is going to look like, or remain, just...varience
Idk if this is helpful, but it seemed like a good thought when I started this thing
We are trying to describe who was "better" by finding things that describe who was more productive more accurately
Say 8-10% or so is regression?
But that leaves 50/52% that is a monster problem, and some of it we will always just be wrong on
Say 8-10% or so is regression?
But that leaves 50/52% that is a monster problem, and some of it we will always just be wrong on
Yeah...idk I explained that well at all
