Way-too-early 2017-18 Forecasts

Home for all your discussion of basketball statistical analysis.
bigcymbal
Posts: 3
Joined: Tue Jun 13, 2017 9:05 pm

Re: Way-too-early 2017-18 Forecasts

Post by bigcymbal » Fri Jul 14, 2017 9:13 pm

How are those using BPM/RPM blends (or just RPM or just BPM) projecting rookie production?

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

Re: Way-too-early 2017-18 Forecasts

Post by jgoldstein34 » Sat Jul 15, 2017 4:45 am

I have a regression model I use to project rookie performance, than adjust based on Summer League. Anyone is free to use it if they'd like: Link. It'll change until the end of Summer League, but easy enough to keep it up to date.

JoshEngleman
Posts: 127
Joined: Thu Oct 23, 2014 12:13 pm
Location: Wilmington, NC
Contact:

Re: Way-too-early 2017-18 Forecasts

Post by JoshEngleman » Sat Jul 15, 2017 12:34 pm

bigcymbal wrote:How are those using BPM/RPM blends (or just RPM or just BPM) projecting rookie production?
I did a regression based on draft slot and Y1 performance. It shakes out so that the #1 pick in Y1 is roughly league average, with a sliding scale all the way to .025 at the end of the first round. I give all 2nd round picks .02, but that doesn't really matter too much. It's not as if I'm projecting any second round picks to play giant minutes right now.

Nathan
Posts: 137
Joined: Sat Jun 22, 2013 4:30 pm

Re: Way-too-early 2017-18 Forecasts

Post by Nathan » Sat Jul 15, 2017 3:13 pm

JoshEngleman wrote:Here's where I stand right now in raw projections. I made a small tweak to force 1230 wins for right now, since there isn't a schedule to sim. "Min?" is the number of minutes that haven't been allocated to a specific player.

*I just used the data below to sim the season 5000 times, using last year's schedule as a proxy. Each team in the East gains just under 1 win from playing in the East, opposite for the West. Basically, if you are a generic 45 win team in the East, bump that to 46. If you're in the West, drop it to 44. Looks OK to me.

*In one of the sims, the Knicks made the playoffs as an 8-seed with 31 wins, which is hilarious.

Image

Also, here are the median wins across 10,000 sims for each seed.

Image
I'm surprised Portland projects so badly at center. Nurkic was fantastic for them in the 20 games he played at the end of last season, leading them to a 14-6 record while averaging 15.2 points, 10.4 rebounds, 3.2 assists, 1.3 steals, and 1.9 blocks in just 29 minutes per game. He's just 22 years old, so it's not crazy to think he could be as good or better next season if he gets major minutes.

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

Re: Way-too-early 2017-18 Forecasts

Post by jgoldstein34 » Sat Jul 15, 2017 4:26 pm

Nurkic projects pretty poorly because he was terrible in his games in Denver last year and not great the year before that. He needs more than 20 decent games to have most models believe he's really that guy.

Nathan
Posts: 137
Joined: Sat Jun 22, 2013 4:30 pm

Re: Way-too-early 2017-18 Forecasts

Post by Nathan » Mon Jul 17, 2017 7:04 pm

Updated my post to reflect win total lines rather than championship odds.

Biggest discrepancies between Vegas lines and Josh's projections include:

-Boston and Utah. Oddsmakers (or at least one oddsmaker) seem to be treating Gordon Hayward as a bona-fide superstar. Does that theory hold any water?

-Oddsmakers actually like the East's middle class of Milwaukee, Detroit, and Philly (and even Atlanta for some reason) a lot. As I understand it Josh's projections don't account for age-related improvements; could that explain the discrepancy?

-On the extreme lower end, oddsmakers have Chicago, Brooklyn, and Phoenix winning very few games, possibly in anticipation of tanking and/or trading away talent to contenders

-Oddsmakers are very bearish on the Wolves, the Clippers, and the Jazz, making the West as a whole look much more tame than a lot of people (myself included) initially thought.

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

Re: Way-too-early 2017-18 Forecasts

Post by jgoldstein34 » Mon Jul 17, 2017 7:17 pm

Here's how my current projections compare to the Vegas lines listed in the first post:

Image

No clue why an oddsmaker, even a bad one, would have Boston valued so highly unless they're trying to sucker people into betting the over or something.

I actually agree that the middle class of the east is fine, it's the high end that is relatively lacking.

Nathan
Posts: 137
Joined: Sat Jun 22, 2013 4:30 pm

Re: Way-too-early 2017-18 Forecasts

Post by Nathan » Mon Jul 17, 2017 7:27 pm

The line on Boston is Over 54.5 (5/7) Under 54.5 (11/10), which I estimated to be about 56 expected wins. If I understand correctly, I think this actually means that the bookmaker's initial prediction was 54.5, and the odds shifted in favor of the Over as money came in.

Crow
Posts: 6188
Joined: Thu Apr 14, 2011 11:10 pm

Re: Way-too-early 2017-18 Forecasts

Post by Crow » Mon Jul 17, 2017 8:11 pm

On first pass, I generally favor either party that pulls teams down compared to the other. Some may need to go up, but that will take a second pass.



On that second pass I'll go with the higher figure (or higher) on Utah, Washington, Memphis Sactown, Miami, Atlanta. Others may fall in the middle.

shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: Way-too-early 2017-18 Forecasts

Post by shadow » Mon Jul 17, 2017 9:27 pm

jgoldstein34 wrote:
No clue why an oddsmaker, even a bad one, would have Boston valued so highly unless they're trying to sucker people into betting the over or something.

The max bets at BetOnline for win totals are only $250 and they are dealing 30 cent lines. I'm not exactly sure how many cents 1 win is worth, but needless to say BetOnline has a pretty damn wide margin for error and minimal risk exposure with this setup. BetOnline is notorious for posting lines as soon as possible, partially because they know they'll get free PR from people who are desperate for attention like RJ Bell to tweet about it. Their initial lines always have very low limits (often $250 max) such that if anyone who's remotely serious about betting had an opinion on the line, they would wait until books like Pinnacle & CRIS (and the myriad offscreen books used by local bookies that clone those sites) post their lines so they can get more than $250 down. As soon as a more serious book (Pinnacle or CRIS usually) posts their line, BetOnline doesn't even care about any movement that happened in their "World Opener" market and will move 'on air' (adjusting the line manually without actually taking any bets to drive the line movement) to get in line with the true market. Their whole 'World Opener" market is nonsense and whoever is responsible for posting the openers at BetOnline has a pretty well established track record of not being very good at it. They simply use wider than usual lines (30 to 40 cents) and extremely low limits to minimize any exposure from people with a clue who are content betting at very low limits.

All of that being said, if you really think the line is way off and you're content betting $250 a pop, then bet the under for the max as many times as it takes to get the line in tune with reality. Remember though that the line can be off by a sizable amount (probably 2-3 wins at least) due to the 30 cent lines and still not necessarily present a +EV opportunity.

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

Re: Way-too-early 2017-18 Forecasts

Post by jgoldstein34 » Mon Jul 17, 2017 11:01 pm

Oh I know, these aren't really a great indicator of where Vegas will end up. Suppose it was more the shock of seeing a line that high for Boston despite BetOnline's weird setup that got to me.

JoshEngleman
Posts: 127
Joined: Thu Oct 23, 2014 12:13 pm
Location: Wilmington, NC
Contact:

Re: Way-too-early 2017-18 Forecasts

Post by JoshEngleman » Tue Jul 18, 2017 11:58 am

Nathan wrote:Updated my post to reflect win total lines rather than championship odds.

Biggest discrepancies between Vegas lines and Josh's projections include:

-Boston and Utah. Oddsmakers (or at least one oddsmaker) seem to be treating Gordon Hayward as a bona-fide superstar. Does that theory hold any water?

-Oddsmakers actually like the East's middle class of Milwaukee, Detroit, and Philly (and even Atlanta for some reason) a lot. As I understand it Josh's projections don't account for age-related improvements; could that explain the discrepancy?

-On the extreme lower end, oddsmakers have Chicago, Brooklyn, and Phoenix winning very few games, possibly in anticipation of tanking and/or trading away talent to contenders

-Oddsmakers are very bearish on the Wolves, the Clippers, and the Jazz, making the West as a whole look much more tame than a lot of people (myself included) initially thought.
Just as a note, I do have some generic aging in place.

JoshEngleman
Posts: 127
Joined: Thu Oct 23, 2014 12:13 pm
Location: Wilmington, NC
Contact:

Re: Way-too-early 2017-18 Forecasts

Post by JoshEngleman » Tue Jul 18, 2017 12:23 pm

The following table is update with transactions through yesterday. I also made tweaks to rookie projections and second round guys since my last post. For the delta column, I've added an additional win to each team in the East and took one from each team in the West, to better account for schedule difficulty. I'm going to talk about a few of the big "misses".

Image


  • I'm a little surprised at how low Brooklyn's line is, but I really wouldn't move it too far.
  • If the Bulls sell during the season (as they should), then the line makes more sense. I didn't really account for that, so a Wade buyout would drop my projection. A Robin Lopez trade would drop my projection.
  • Boston at 56 just seems crazy to me, but they are a team set-up perfectly for the regular season. It seems like a high starting line, but it's not a shocking outcome at all.
  • Utah is the first one I can't wrap my head around. This is the first discrepancy where I would actually bet on the over.
  • Atlanta stinks. I would be comfortable betting the under.
  • The Knicks line is likely factoring in a Carmelo trade, which is not what I have done.
  • My conference balance seems pretty spot on. After making the +1/-1 schedule adjustment, I have 585 wins in the East. Lines total 590. The totals in the East go from 20.5 to 56, while mine go from 28 to 51. The West totals are 26.5 to 68, while mine are 28 to 63. Basically, I think there will be more natural pull towards the mean in the East.

shadow
Posts: 217
Joined: Fri May 29, 2015 3:38 am

Re: Way-too-early 2017-18 Forecasts

Post by shadow » Tue Jul 18, 2017 3:36 pm

To get an idea of how much uncertainty there is with win totals, I simulated this past season's regular season 10,000 times using SRS differential with a 3 point HCA to project a spread and converted that to a win probability for each game using normdist(spread,0,12,true) in Excel. For each game I generated a random number between 0 & 1 and compared it to the win probability to determine the winner. Here are the win ranges for each team after 10,000 runs, with low and high representing 2 standard deviations below and above the mean wins for each team:

Code: Select all

Team  Low     High
ATL	30.22	47.46
BOS	39.18	56.30
BRK	16.69	32.73
CHA	32.80	50.16
CHI	33.43	50.79
CLE	40.47	57.47
DAL	25.33	42.65
DEN	33.63	50.83
DET	29.73	46.81
GSW	60.12	73.68
HOU	47.00	63.04
IND	31.94	49.14
LAC	43.30	59.98
LAL	16.73	32.61
MEM	33.79	50.87
MIA	35.15	52.35
MIL	31.99	49.27
MIN	29.81	46.81
NOP	27.34	44.26
NYK	23.53	40.41
OKC	34.27	51.75
ORL	17.01	33.09
PHI	18.89	35.25
PHX	19.09	35.41
POR	31.20	48.32
SAC	23.21	39.97
SAS	50.53	66.29
TOR	43.19	59.79
UTA	42.45	59.09
WAS	37.13	54.29
So there is roughly a 16 win window each team could realistically fall within just by random chance if we assume team strength is constant throughout the season. Ideally I'd use the closing no-vig moneyline odds from Pinnacle for calculating win probabilities, but I didn't have those readily available so SRS is probably a good enough proxy for team strength.

sndesai1
Posts: 133
Joined: Fri Mar 08, 2013 10:00 pm

Re: Way-too-early 2017-18 Forecasts

Post by sndesai1 » Wed Jul 19, 2017 7:10 pm

if you still have the results, what is the average across all sims of the rmse compared to the true win total? thanks!

Post Reply