Why are the Stats model so off on the Kings?

S

sactowndog

All-Star
#1
#1
it is interesting to look from an analytical standpoint why the experts were so far off. 538 still has the Kings at only 37 wins which now seems low. Stats geeks please chime in with your hypothesis. Here are mine.

Reason 1: the net points for/against can be very misleading. The metric isn’t controlled for outlier games where a team may go to their bench. A perfect example is the recent Minnesota game versus New Orleans. The Kings on the third game in 4 nights and a long Dallas to Minnesota back to back waived the white flag. Outliers should be removed from the metric.

Reason 2: Clutch isn’t random. The current metric assumes clutch play is random and anyone who has played can tell you it’s not. The Kings were one of the most clutch teams last year as they are this year. The guys doing game predictions don’t accurately factor in clutch.

Reason 3: it takes minutes to score. Buddy isn’t scoring at a better rate this year than last year. He is just playing more minutes because Jeoger isn’t trying to look at all players. Put Buddy in Ben Mclemore’s minutes and strangely enough both his and the teams offensive numbers will go up.

Reason 4: Pace matters and last year the Kings had the slowest pace. With a Fox at point that was unlikely to continue.

Many of the factors still come into play with 538 projecting the Kings with 37 wins and likely to finish 7th.
 
E

ESP47

All-Star
#2
#2
I think part of it has to do with the fact that when they lose, they seem to lose by a ton of points and when they win it's not usually in blow out fashion.

I'm confused by your 3rd reason. What do you mean when you say put Buddy in BMac's minutes and both his and the teams offensive numbers will go up?
 
LPKingsFan

LPKingsFan

All-Star
#4
#4
I may be way off on this, but I believe 538’s model is based on its model for individual player projections. So, if the model is off on a few key players, it could miss on the team as a whole quite a bit. I haven’t checked, but I’d guess that a few key Kings are overperforming their projection from 538. The whole also seems to be more than the sum of the parts.
 
S

sactowndog

All-Star
#5
#5
ESP47 said:
I think part of it has to do with the fact that when they lose, they seem to lose by a ton of points and when they win it's not usually in blow out fashion.

I'm confused by your 3rd reason. What do you mean when you say put Buddy in BMac's minutes and both his and the teams offensive numbers will go up?
Click to expand...
Last year Jeoger played a really deep bench and no starter averaged more than 25 minutes. Buddy’s numbers adjusted for minutes aren’t to different and his numbers are much more productive than whoever used those minutes last year. No one accounted for that change in how deep we played the bench.
 
S

sactowndog

All-Star
#6
#6
LPKingsFan said:
I may be way off on this, but I believe 538’s model is based on its model for individual player projections. So, if the model is off on a few key players, it could miss on the team as a whole quite a bit. I haven’t checked, but I’d guess that a few key Kings are overperforming their projection from 538. The whole also seems to be more than the sum of the parts.
Click to expand...
Good catch. I missed that. The pure Elo model is actually 41 which is much closer to what I would expect. I think my reasons would still factor into player projections however and could drive a number of projections to be off thus causing the cumulative miss.
 
pdxKingsFan

pdxKingsFan

So Ordinary That It's Truly Quite Extraordinary
Staff member
#7
#7
kingsfan1984 said:
Also, we must have played one of the hardest schedules or strength of opposition thus far to date.
Which means 2nd half of season will be easier eand team could surge in standings or at the least stand a very good shot at holding ground in standings
Click to expand...
I think our 2nd half is easier but I don't think it is a dramatic change, because I am pretty sure the difficulty of schedule rankings also account for things like back to backs and miles traveled and the kind folks who made the schedule gave us either the first or second most in the league this year.
 
Capt. Factorial

Capt. Factorial

trifolium contra tempestatem subrigere certum est
Staff member
#8
#8
sactowndog said:
Many of the factors still come into play with 538 projecting the Kings with 37 wins and likely to finish 7th.
Click to expand...
Wait. 538 is projecting 37 wins to be 7th in the West?!?

(Checks standings)

Yep, right now 37 wins (.451) would be good for...14th in the West.
 
pdxKingsFan

pdxKingsFan

So Ordinary That It's Truly Quite Extraordinary
Staff member
#9
#9
Capt. Factorial said:
Wait. 538 is projecting 37 wins to be 7th in the West?!?

(Checks standings)

Yep, right now 37 wins (.451) would be good for...14th in the West.
Click to expand...
I think that means 7th in the lotto.

But I don't know.

Also as an early fan of Nate Silver who appreciates his efforts to bring political polling into the 21st century, I don't understand why so much respect is given to his sports analysis. Especially since his political polling has regressed now that others have adapted and caught up. I mean usually you hear about how "538 gave the team a 92% lock to win" as the SportsCenter highlights show the team blow it.
 
kb02

kb02

All-Star
#11
#11
Some fast food research thoughts:

3 yr weighted average for a team with a bunch of first contract players means the sample size is off as it relates to the Kings—Bags, Fox, JJ, Giles, Mason, Bogie don’t have three yrs to measure.

The stats aren’t adjusted for position. See Fox’s stats. His height and weight are considered bad, like in the lowest percentile bad. Fox, as we all know, is a premium athlete at the 1 spot.

If true, the second variable means 538 favors bigs n fails to factor in modern day differences in positions—pg > centers.

Note: the above comments are based on a cursory view of the 538 data. Subject to change if I do a deeper dive on their adjustments.

See here: https://projects.fivethirtyeight.com/carmelo/deaaron-fox/
 
F

fnordius

All-Star
#15
#15
I think the reason's mainly the same as why most Kings players have bad looking RPM stats relative to their usual play. I can sum it up with the last 5 games.

Mavs: W 120-113 +7
Timberwolves: L 132-105 -27
Thunder: L 132-113 -19
Grizzlies: W 102-99 +3
Pelicans: W 112-117 +5

Add that up, and you get a 3-2 team which loses the average game by over 6 points. If continued, we could have a team that heads into the playoffs with a crummy ELO and a whole roster full of ugly RPMs. Memphis is the only other team which has a winning record and a negative net score, but their gap is relatively slight (.515, -0.2) compared to the Kings (.545, -1.3).

Outside of maybe chess, I know of no sport where ELO actually matters, but the team occasionally taking a night off is confusing 538's algorithms, and whoever plays on those nights can be expected to take a big hit to their plus-minus stats.
 
C

Cojc

Bench
#26
#26
Well Lebron left the Lakers game in the 3rd quarter with what looked like he pulled a groin muscle. Did not look good.
Lakers still taking it to golden state.
 
You must log in or register to reply here.