Playoff predictions?

[Crow]

If anyone wants a new test / game, there could be a playoff contest. This season data.

How to score? 1 point for first round series called correctly, 2 for second round, then 3 and 4? Or 1 point for each game called correctly?

I am unlikely to do this but throw it out as suggestion if any are interested, especially if anyone wants to test a system.

[Nathan]

For reference, here are the Vegas series odds (via oddschecker):

Celtics 1/5, Bulls 9/2 (~82% for Celtics to win)
Cavs 1/10, Pacers 15/2 (~90% for Cavs to win)
Raptors 2/7, Bucks 3/1 (~76% for Raps to win)
Wizards 5/11, Hawks 21/10 (~68% for Wiz to win)

No odds found for Warriors/Blazers (lol)
Spurs 1/11, Grizz 15/2 (~90% for Spurs to win)
Rockets 2/7, Thunder 31/10 (~77% for Rockets to win)
Clippers 1/2, Jazz 21/10 (~67% for Clippers to win)

In all cases, it would probably be pretty hard to argue statistically that the underdog should be the favorite. But which teams may be over/under rated here?

For instance, I suspect the Hawks may be underrated because their only players in the top 100 of RPM, Paul Millsap and Ersan Ilyasova, played only a handful of games together in the regular season. With both of them available for the playoffs, the Hawks may be substantially better than they were in the regular season. Because few people believe that Millsap is a star and Ilyasova is a substantially above-average player, vegas may not be adequately accounting for this.

The same may be true for the Bulls. Mirotic and Grant, their second and third best players by RPM, saw very inconsistent minutes throughout the season due to poor coaching. By the last few games, however, both had carved out large roles and were playing well. The Bulls could outperform expectations if Hoiberg continues to play them big minutes in the playoffs.

[shadow]

Based on team ratings I generated from 2016-17 RPM ratings I came up with the following win probabilities for each series:

GSW 98.10%
SAS 80.43%
LAC 73.30%
HOU 78.42%

BOS 71.72%
CLE 82.30%
TOR 68.67%
WAS 73.87%

I would probably discount the Clippers win probability a bit because the series calculator I used didn't let me specify HCA for each team individually and the Jazz have a slightly higher than usual HCA due to altitude.

Grizzlies, Bulls, Bucks, Pacers, Wizards appear a bit undervalued based on this model. However, it is a very simple model and I wouldn't anticipate something like this would beat Vegas.

[J.E.]

I get

GSW: 92%
SAS: 71%
HOU: 71%
LAC: 66%

BOS: 65%
CLE: 71%
TOR: 64%
WAS: 69%

using the Pythagorean formula with an exponent of 10.5 - that's what I found to give best prediction results when I studied it years ago
Seems like my version expects more upsets. Admittedly, 92% for the Warriors seems low. One rather terrible german betting site offers 1.01:1 on the Warriors, 15:1 on the Blazers

[Crow]

By these probabilities, about 2 first round upsets would be expected overall. Last year had 1, year before 2. 4-5 matchup is most likely to give of course.

By performance against top 10, Boston and Washington are considerably weaker than opponent. Probably at least one falls. By performance against top 10 they are not top 16, much less higher seeds. By performance against top 16, Washington is still clearly weaker; Boston with the slimmest of edges.

Houston, San Antonio and Toronto may have tougher time than overall record suggests.

[colts18]

I get

GSW: 92%
SAS: 71%
HOU: 71%
LAC: 66%

BOS: 65%
CLE: 71%
TOR: 64%
WAS: 69%

using the Pythagorean formula with an exponent of 10.5 - that's what I found to give best prediction results when I studied it years ago
Seems like my version expects more upsets. Admittedly, 92% for the Warriors seems low. One rather terrible german betting site offers 1.01:1 on the Warriors, 15:1 on the Blazers

Do you have any championship odds? How do they compare the Vegas odds:

http://www.espn.com/chalk/story/_/id/16 ... -superbook

[shadow]

Cavs give up 122 pp/100 to a team with league average OE on the year. More of the same Cleveland "defense" we saw in the regular season.

[Crow]

Saw a post elsewhere that said for the first 6 games finished the winner had 3 guys with 16 plus points and none of the losers had 3.

[shadow]

Game 2 predictions for spreads/totals:

GSW -14.32 / 219.16
LAC -8.03 / 196.36
HOU -7.29 / 220.46
SAS -8.24 / 187.60

BOS -5.06 / 205.92
WAS -7.05 / 209.19
TOR -3.82 / 197.20
CLE -6.97 / 209.71

For comparison here are the current spreads/totals at Pinnacle:

GSW -15.5 / 220.5
LAC -8 / 195.5
HOU -7.5 / 224
SAS -11 / 188.5

BOS -6 / 205.5
WAS -5.5 / 210.5
TOR -7.5 / 193
CLE -7.5 / 210

Hypothetical picks would be MEM +11, MIL +7.5, WAS -5.5, POR +15.5, HOU/OKC U224, GSW/POR U221

[Crow]

Of last two games yesterday, one followed the 3 players at 15 pts or above for winner but not loser and the other game missed by 1 pt for one player on the winner but followed the pattern for the loser. May update later in playoffs.

[shadow]

Switched to using multi-year RPM JE posted on twitter. Was previously using single year RPM from ESPN.

Game 3 projections:

GSW: -6.97 / 216.44
UTA: -1.56 / 195.47
OKC: -1.66 / 221.69
SAS: -3.7 / 181.28

CHI: -3.20 / 202.17
ATL: -0.67 / 209.99
MIL: -1.84 / 198.71
IND: -1.06 / 205.30

Hypothetical plays:

CLE/IND U210.5
SAS/MEM U187.5
BOS/CHI U207
HOU/OKC U224

IND +2
CHI -2
HOU +3
WAS +3
UTA +1.5

Here's how much the line swung in each series from Game 2 to Game 3:

CLE/IND: 7.5
TOR/MIL: 10
SAS/MEM: 6.5
BOS/CHI: 8.5
HOU/OKC: 10.5
LAC/UTA: 7
WAS/ATL: 9
GSW/POR: 5.5

So on average, the odds are implying HCA is worth around 4.03 points. Although some of the adjustment is likely stemming from underdogs like Milwaukee (2-0 ATS), CHI (2-0 SU & ATS), and OKC (1-1 ATS, led for 3 qtrs in G2) beating expectations against the spread in one or both games. Atlanta didn't cover the spread in either game, but only lost by 7 and 8 points. So the real HCA value is probably around 3.75.

[Crow]

In a post at cleaningthe glass.com Ben Falk said it was 5.5 pts in playoffs... and then said it was worth a lot less at game to game level. Twitter query. Twitter reply? Nope. New post but no reply.

[shadow]

There's no way it's 5.5 in today's league, even if that's per 100 possessions. He might be including all the playoff games in each series in his analysis. That inflates the HCA value because the higher seed gets 1 more home game than the lower seed when the series goes 5 or 7 games. His own post claims that HCA is even higher in Game 7's and that the affect is statistically significant. So if he's including Game 7's in his analysis for overall HCA, then that will inflate the typical HCA seen in most playoff games. Another possibility is he's using playoff data from before the mid 90's when HCA was much higher. HCA dropped off quite a bit in the mid 90's and has been as low as 2.5 points in the regular season in recent years. Including more seasons will give you a better idea of what HCA has been in the past, but it likely skews the results when trying to predict playoff HCA in today's league.

Ed Feng did an analysis from 2003 to 2012 and found HCA in the playoffs to be 4.25. He only used an even number of games for each series to address the first issue mentioned above.

I ran the numbers for the last four years of playoff games using Ed's method and HCA was 3.78 points on average.

http://wagesofwins.com/2013/04/19/home- ... -playoffs/

[Mike G]

Since 2000, the Spurs have had 120 home playoff games, which they have won .725 of the time. They've averaged +8.7 MOV
in these contests.
In 112 away games, they've won .482 with an avg MOV of -1.5
The differences are .243 in Win% and 10.1 PPG
http://bkref.com/tiny/6hlFM

The Lakers have played 197 games, in which their home games have them winning .344 more often and 10.0 ppg better.
These aren't anomalies. 30 teams in 2828 games (1414 actually) are 8.94 ppg better at home than on the road.
The home team wins 65% of the time, on avg.

The avg improvement of .295 in Win% is equivalent to a MOV improvement of 11 ppg at home vs away.

[shadow]

So I referenced how Ed Feng explained that using all playoff games ends up inflating the true value of HCA and you go ahead and use all the playoff games as a way of refuting my claim that HCA isn't really 5.5 points in the playoffs. Nice troll attempt, pal.