APBRmetrics

Curios as to what other people expect from the finals. I know most writers are hoping for a close finals, but I'm not sure I see it. I have 2 different playoff models, 1 which uses all season/playoff games with an equal weight and a 2nd which weights more heavily toward recent games to give more of a current view of the teams.

The equal weights model has Warriors at 90.5% to win with the following distribution:
Warriors in 4 -- 22.2%
Warriors in 5 -- 35.0%
Warriors in 6 -- 16.7%
Warriors in 7 -- 16.6%
Cavaliers in 7 -- 3.3%
Cavaliers in 6 -- 4.6%
Cavaliers in 5 -- 1.1%
Cavaliers in 4 -- 0.5%

The recency model has Warriors at 87.9% to win with the following distribution:
Warriors in 4 -- 18.5%
Warriors in 5 -- 35.2%
Warriors in 6 -- 17.1%
Warriors in 7 -- 17.0%
Cavaliers in 7 -- 3.6%
Cavaliers in 6 -- 5.7%
Cavaliers in 5 -- 2.1%
Cavaliers in 4 -- 0.7%

Neither view has it as a particularly close series. I know most models underrate the Cavs because of how much better they've been in the playoffs than the regular season. Even though the recency model sees them as the second best team in the NBA at +7.2 points per 100 possessions, that's still far behind the Warriors as the best at +13.9. I'd expect the Warriors to win in 5 with the 2 games in Cleveland being marginally close and one going to each team.

Vegas has the Warriors at -240 (implied odds of 70.6%) and the Cavs at +200 (implied odds of 33.3%). Cavs are really yet to face an elite offensive player and a great defense. I think they could be in for a harsh awakening. As good as they've looked, there's a real possibility that this Warriors team is the best we've ever seen.

Are your models based on player or team level data?

Also, with 10K limits at Bookmaker on the series line, if the true odds were 85-90% in favor of the Warriors, there would be a slew of max bets from sharps driving the line towards that range. Those bets would easily outweigh any bets from recreational bettors wagering $20-$100 on the series.

I don't think it's fair to say the Cavs haven't faced a great defense. The Celtics were excellent defending the 3 in particular both in the regular season and in the playoffs, and elite defensively overall with Thomas on the bench (which was true for most of the series against the Cavs). After the fact it feels like they just rolled over and gave up, but in reality they had a lot of very solid defensive players giving all they had and an savvy (if mildly overrated) coach, and they just couldn't find any answers.

My ratings are team level, but Kevin Pelton did similar analysis using RPM and found the Cavs to be +7.8 and the Warriors to be +14. It's scary to bet and pick against LeBron, he's amazing and I'm sure will find some way to make a few games interesting, but I don't see it being very close. The Warriors won 73 games last year with an all time great MOV and added Kevin Durant. As much as the Cavs have been playing their best basketball in the playoffs, so have the Warriors and the Warriors just have so much more talent.

The Celtics issue was they didn't have any semblance of rim protector. LeBron and Kyrie would get past the initial defense and there would be no one there to meet them. With the Warriors, you have to get past Klay or Iggy and then score over/around KD or Draymond. It's just another level of difficulty that the Celtics couldn't offer.

I don't personally think any team in the NBA can play at a +14 level for a sustained period of time. The '96 Bulls at +13.01 have the highest adjusted net rating since at least 1986. I ran an SRS calculation for the entire season (reg + playoffs) and the Warriors were rated +12.39. When I ran it again using a weighting scheme similar to what KenPom uses, they were rated +12.88. Granted those metrics are in points per game, but the Warriors average around 100 possessions in their games, so the difference is likely negligible. You can usually improve the predictive accuracy of ratings by using ridge regression instead of linear regression as is used with SRS. In my ridge regression ratings the Warriors are a +11.5 and +11.6 in the unweighted and weighted versions respectively.

If you run ratings for just the playoffs you get the following ratings for each team:

Warriors: +17.49 (SRS) ; +11.72 (Ridge)
Cavs: +15.84 (SRS) ; +10.24 (Ridge)

You can't necessarily compare the ratings at face value though because that would assume the average East playoff team is equivalent to the average West playoff team, which I think we can all agree isn't true. The average West playoff team (excluding GSW) in my unweighted ridge ratings is a +2.89, the average East playoff team (excluding CLE) is a -0.47. So if I wanted to come up with a hypothetical spread using just the playoff ridge ratings, it would be: 11.72 - 10.24 + 2.42 + 3.5 = 7.4, in favor of GSW. The 2.42 is the difference in the average rating among playoff teams in each conference, excluding CLE & GSW, the 3.5 is for HCA. The current line is GSW -7, so that seems like a good approximation.

Not saying that a playoff-only rating system is fair, but are the Blazers better than the Pacers? The Jazz better than the Raptors? The Kawhi-less Spurs better than the Thomas-less Celtics? Reasonable people could disagree on this.

Nathan wrote:Not saying that a playoff-only rating system is fair, but are the Blazers better than the Pacers? The Jazz better than the Raptors? The Kawhi-less Spurs better than the Thomas-less Celtics? Reasonable people could disagree on this.

I think I would say yes to all of those, especially considering that the Raptors didn't have Kyle Lowry. That's the only close one to me, and it's not that close sans Lowry.

jgoldstein34 wrote:

Nathan wrote:Not saying that a playoff-only rating system is fair, but are the Blazers better than the Pacers? The Jazz better than the Raptors? The Kawhi-less Spurs better than the Thomas-less Celtics? Reasonable people could disagree on this.

I think I would say yes to all of those, especially considering that the Raptors didn't have Kyle Lowry. That's the only close one to me, and it's not that close sans Lowry.

Really? Don't forget injuries to Nurkic (Blazers' second best player by BPM and only real frontcourt player) and Hill (Jazz's third best player by BPM and only real point guard). I'm probably with you on the Spurs though, I think they'd handle the Celtics even without Kawhi, though Aldridge looks awfully washed up these days.

Is Tyronn Lue a genius coach? He was last year.

I thought the Blazers, even without Nurkic, were a bit better than the Pacers. Utah without Hill much better than Raptors without Lowry, best player missing vs 3rd best.

I can't wait to find out how they compare tonight, hoping it's at least competitive for a bit.

Haven't seen anything out of Lue to think he's anything besides an above average coach this season. Hard to see where LeBron's leadership ends and Lue's coaching begins.

Series betting line moved to Warriors -700 (87.5%) and Cavs +450 (18.2%). Anyone who had Warriors at -240 or better has to feel pretty good right now. Game 2 line moved from Warriors -7 to Warriors -9 and Game 3 moved from Cavs -2 to Cavs -1. If the Warriors dominate game 2 again, that game 3 line could move to the Warriors being favorites, even on the road in the finals.

I think as much as projection systems underrated the Cavs because of their meh regular season, those same systems may have underrated the Warriors too because they have so much freaking talent. Having KD on top of a team that won 73 games last year matters so much more in Finals games than in regular season games. I think that betting markets may have forgotten that a bit because of how much better the Cavs have looked in the playoffs than the regular season, but the Warriors are legit the best team ever put together and should waltz through these finals even if LeBron goes crazy.

After 1 game:

The equal weights model has Warriors at 93.8% to win with the following distribution:
Warriors in 4 -- 26.4%
Warriors in 5 -- 39.2%
Warriors in 6 -- 14.1%
Warriors in 7 -- 14.1%
Cavaliers in 7 -- 3.1%
Cavaliers in 6 -- 2.9%
Cavaliers in 5 -- 0.2%

The recency model has Warriors at 91.1% to win with the following distribution:
Warriors in 4 -- 23.4%
Warriors in 5 -- 36.7%
Warriors in 6 -- 16.0%
Warriors in 7 -- 15.0%
Cavaliers in 7 -- 4.0%
Cavaliers in 6 -- 3.9%
Cavaliers in 5 -- 1.1%

Average shifts from Warriors at 89.2% to win before game 1 to 92.45% to win after game 1. Models only shift 3.25% to Warriors while Vegas shifts 16-17%, makes me think the betting public underrated Warriors. Only 1 game, but looked very promising for Warriors. Cavs need lots of great series from multiple guys to win.

Needing to win 4 of 6 games now, the Cavs' chances are now 1/10 or 1/11.
When they had 7 chances to win 4, their odds were some 1/7 to 1/8.

The 'recency' model drops from .148 to .108 -- they've lost 27% of their former chance, after 1 game.
Equal weight has them going from .129 to .097 -- a dropoff of 25% from the opening estimate.

Nice work, sir.

I would say 4-0 if I saw this thread before the first game and I'm saying 4-0 again after the first game. Last year I thought it was the common non-empirical 51%-49% ratio in favour of Cavs for me.

Durant is better than people and +/- based metrics give credit for. Durant-Curry-Thompson-Green would be stylistic nightmare for almost all teams in the history. I've been watching the game since the mid 90s and the 01 Lakers along with the 96 Bulls are the only two teams I would comfortably pick over this Warriors.

It wouldn't be appropiate to only talk about subjective arguments in this community so, are there any sophisticated team-based models out there for this matchup? I expect "4-0 Warriors" to be the highest probabilty of all outcomes .

After game 2:

The equal weights model has Warriors at 96.8% to win with the following distribution:
Warriors in 4 -- 29.2%
Warriors in 5 -- 42.6%
Warriors in 6 -- 14.2%
Warriors in 7 -- 10.8%
Cavaliers in 7 -- 1.6%
Cavaliers in 6 -- 1.6%

The recency model has Warriors at 95.3% to win with the following distribution:
Warriors in 4 -- 27.9%
Warriors in 5 -- 41.8%
Warriors in 6 -- 13.5%
Warriors in 7 -- 12.1%
Cavaliers in 7 -- 3.2%
Cavaliers in 6 -- 1.6%

Both models largely singing the same tune at this point with small differences in distribution (Recency gives a slightly larger chance to a sweep).

Vegas currently has Warriors -2000 (95.2%) vs Cavs +1000 (9.1%). The no-vig odds come out to 91.3% for the Warriors (Thanks to Shadow for the correction).

Average shifts from Warriors at 92.45% to win before game 2 to 96.05 to win after game 2, a shift of 3.6%. Vegas once again shifting a bit more towards the Warriors than my models (8.5% from post-game 1 to post-game 2), now nearly in line with what my system is saying.