HamiltonNHL
Parity era hockey is just puck luck + draft luck
- Jan 4, 2012
- 21,156
- 11,700
One of the flaws in this methodology (and all methodologies do have flaws) is that it it appears to be impossible to trace "why". Why is Jamie Benn so low? Why are Henrik and Daniel Sedin so far apart?
What I'd recommend in this specific case is to take a non-intuitive result (the Sedins would be where I'd start) and go through the model, looking at (and documenting) each instance of where Henrik and Daniel change rankings (relative to one another). If I'm understanding correctly, it would only occur when precisely one of them is on the ice.
It's not totally impossible... I think I know what is going on here.One of the flaws in this methodology (and all methodologies do have flaws) is that it it appears to be impossible to trace "why".
Why is Jamie Benn so low? Why are Henrik and Daniel Sedin so far apart?
type Sedin Minus Plus Plus_pct
1: Goal Both 122 187 0.61
2: Goal Daniel 31 20 0.39
3: Goal Henrik 27 41 0.60
4: Shot on-goal Both 1551 2029 0.57
5: Shot on-goal Daniel 305 267 0.47
6: Shot on-goal Henrik 401 370 0.48
Correct.If I'm understanding correctly, it would only occur when precisely one of them is on the ice.
They are ignored.How are "Hits" factored into the algorithm?
It's not totally impossible... I think I know what is going on here.
True but the Sedin and Benn examples point to a positional situation especially when Benn is factored in.
They are ignored.
This really hurts certain players. Suggest looking at the records of opposing RHS defencemen, Karlsson, Subban etc, vs Ovechkin and the Capitals last two seasons.
His aggressive forecheck and hitting severely impacts their ability to transition the puck or skate it up ice. True for other physical forwards as well.
Hits are the most inconsistent reported stat in the league. Whats recorded as a hit in one arena isn't the same as another.
How the recording of hits varies from building to building definitely matters - to present an extreme example, if Building A only counts "impactful hits" as "hits", while Building B counts all hits as "hits", then you're going to reach a skewed conclusion as to the true value of things leading from a hit.
However, these biases can be corrected for, so it's not a dead-end by any stretch of the imagination.