2018-19 team win projection contest

Home for all your discussion of basketball statistical analysis.
Crow
Posts: 6208
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 team win projection contest

Post by Crow » Tue Aug 28, 2018 11:26 pm

Ok.

I was agreeing with 538 but my lower projection wasn't primarily on that basis. Age, lineup management and player synergy all were part of the pessimism.


A lot comes down to Carlisle's coaching. Last season, his actual wins were 9 short of expected by Pythagorean estimate. I've never seen worse.

Is he going to try to win or tank all or half of season? I dunno. Last season probably was a tank. He averaged a plus 1 in previous 6 seasons. But another tank, half season tank at least, seems likely. The 8th seed will be probably be too far a reach; so there will be some logic for eventually becoming casual or focusing on player development or whatever.

jgoldstein34
Posts: 240
Joined: Fri Sep 30, 2016 6:38 pm

Re: 2018-19 team win projection contest

Post by jgoldstein34 » Wed Aug 29, 2018 12:48 am

I'm aware it wasn't based off that, was just noting what I saw from a quick glance that stood out as main contributors.

I think Carlisle will be fine, he's a good coach. Last year was intentional, they'll try for half the year and see how close they are. I think there are real scenarios where they get the 8th seed that they'll play 50-60 games for.

Crow
Posts: 6208
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 team win projection contest

Post by Crow » Wed Aug 29, 2018 3:41 am

As of about 2 weeks ago, you gave Mavs about a 14% chance to make playoffs. I would probably go 2-5%. Yeah, it probably will be more like 60 games before they re-assess.

Jinxed
Posts: 26
Joined: Mon Jun 13, 2011 9:53 pm

Re: 2018-19 team win projection contest

Post by Jinxed » Fri Aug 31, 2018 4:02 am

It seems like the Luka Doncic comparisons and low rating (from 538) are based simply on biographical data:

"This year, we’ve introduced simple projections for European draftees based on biographical data: specifically, their age, draft position, height, weight, position "

Crow
Posts: 6208
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 team win projection contest

Post by Crow » Sun Sep 02, 2018 9:37 pm

Eminence, I don't recall you mentioning if your projections were model based or strictly or if RPM / RAPM played a role or a dominant one. You may or may not to answer on parts on any of this; but if you are willing, you've been asked.

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Mon Sep 03, 2018 2:08 am

Crow wrote:
Sun Sep 02, 2018 9:37 pm
Eminence, I don't recall you mentioning if your projections were model based or strictly or if RPM / RAPM played a role or a dominant one. You may or may not to answer on parts on any of this; but if you are willing, you've been asked.
I do have a model, though I don't have it set up to spit out an even 1230 wins (within ~10 under normal circumstances), I tweak those last couple wins to 1230 by hand. It's team factors based, with a directional variable (-/neutral/+) that I assign subjectively. Individual stats are not included anywhere.

Crow
Posts: 6208
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 team win projection contest

Post by Crow » Mon Sep 03, 2018 2:33 am

Ok, thanks for the reply. I'll assume it is raw 4 factors. If no individual stats are used, I assume your last stage manual tweaking is the main adjustment to account for roster and role changes.

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Mon Sep 03, 2018 2:39 am

Crow wrote:
Mon Sep 03, 2018 2:33 am
Ok, thanks for the reply. I'll assume it is raw 4 factors. If no individual stats are used, I assume your last stage manual tweaking is the main adjustment to account for roster and role changes.
3 factors actually. But yep, that directional variable accounts for roster changes and whatnot.

Crow
Posts: 6208
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 team win projection contest

Post by Crow » Mon Sep 03, 2018 2:52 am

Do you make your marks before looking at other projection sets or do you look at them and consider them in the background for making you marks? Or 2 step it using both without and with looking?

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Mon Sep 03, 2018 3:15 am

Crow wrote:
Mon Sep 03, 2018 2:52 am
Do you make your marks before looking at other projection sets or do you look at them and consider them in the background for making you marks? Or 2 step it using both without and with looking?
Hmm, I don't really go out of my way one way or another for the directional bit. Like I look at other projections as they come out, but I'm not consciously considering them when assigning values.

shadow
Posts: 218
Joined: Fri May 29, 2015 3:38 am

Re: 2018-19 team win projection contest

Post by shadow » Thu Sep 06, 2018 2:20 pm

Pinnacle win totals are up now.

Code: Select all

Atl 24
Bos 58
Brk 31.5
Cha 35
Chi 28.5
Cle 30.5
Dal 34
Den 46.5
Det 38
GS  63
Hou 55
Ind 46
LAC 36.5
LAL 49.5
Mem 33
Mia 41.5
Mil 46
Min 44.5
NO  44
NY  30
OKC 49
Orl 31
Phi 53.5
Phx 28
Por 42
Sac 26
SA  44
Tor 54.5
Uth 49
Wsh 44

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Fri Sep 07, 2018 3:34 pm

shadow wrote:
Thu Sep 06, 2018 2:20 pm
Pinnacle win totals are up now.

Code: Select all

Atl 24
Bos 58
Brk 31.5
Cha 35
Chi 28.5
Cle 30.5
Dal 34
Den 46.5
Det 38
GS  63
Hou 55
Ind 46
LAC 36.5
LAL 49.5
Mem 33
Mia 41.5
Mil 46
Min 44.5
NO  44
NY  30
OKC 49
Orl 31
Phi 53.5
Phx 28
Por 42
Sac 26
SA  44
Tor 54.5
Uth 49
Wsh 44
Could you give me a basic explanation for 'juice' not really familiar with gambling/the terms.

shadow
Posts: 218
Joined: Fri May 29, 2015 3:38 am

Re: 2018-19 team win projection contest

Post by shadow » Fri Sep 07, 2018 5:13 pm

I think I edited out that part shortly after posting because I couldn't quickly track down the formula I used last year to find the true projection with even juice.

Basically, if the price on both sides of the wager aren't equal, then the value of the line doesn't represent the true projection for the number of regular season wins according to the betting market.

The Pinnacle regular season win total lines can be found here:
https://www.pinnacle.com/en/odds/specia ... etball/nba

If you look at Minnesota, the current line is 44.5 wins, but the price on the over is -150 compared to +126 on the under. (You may need to toggle from 'Decimal Odds' to 'American Odds' at the top of the page to see the same odds format).

If you take the -150/+126 odds and plug them into a calculator like the one linked below to remove the vig, you'd see those odds imply a 57.55% chance of Minnesota winning more than 44.5 games and 42.45% of winning less than 44.5 games. So the betting market is really implying that Minnesota is projected to win slightly more than 44.5 games (probably around 47 if I had to guess).

http://sportsbettingsites.org/betting-t ... alculator/

When Pinnacle first opens the lines for this type of market, they'll list both sides with the same odds (-109 in this case) and then let the betting activity drive the movement of the odds. The amount the line moves after a bet is based on both the size of the bet and user profiling (essentially, winning players move the line more than losing players).

So the fact that Minnesota's odds are up to -150 on the over is an indication that Pinnacle has taken a fair amount more money on the over (likely from winning players whose opinions they respect) than the under.

I figured out roughly how many cents a win was worth last year so that I could convert the Pinnacle lines to the true 50/50 midpoint. I'll have to track that down or figure it out again. For evaluating how accurate the projections are from Pinnacle (or just about any sportsbook since most of them just shift the odds instead of moving the win total itself), it's not fair to just use the value of the line. For teams with significantly better odds on the over you'd be under-representing their win total as measured by the betting market. For those with better odds on the under you'd be over-representing their win total.

Hopefully that answered your question. If anything wasn't clear let me know.
Last edited by shadow on Fri Sep 07, 2018 6:04 pm, edited 1 time in total.

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Fri Sep 07, 2018 5:47 pm

Thanks Shadow, think that covered everything I was wondering about.

eminence
Posts: 138
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018-19 team win projection contest

Post by eminence » Mon Sep 10, 2018 12:16 am

Hmm, Booker out for a bit, any significant change in Suns predictions?

Post Reply