Throwing out a 16 game winning streak as an outlier only makes sense if you're looking at "winning versus losing" as the only measure of performance and you're treating it as a totally probabilistic thing, i.e. getting 16 "wins" in a row is completely unlikely and shouldn't be seen as having an effect on future probabilities of getting a "win." The same way flipping a coin and having it land heads 5 times in a row doesn't change the probability of getting heads or tails.
The problem is that the variables that go into that winning streak, i.e. shooting efficiency, defense, rebounding, etc, are all part of the equation in every game, win or lose. The team can play well enough one night to win against most opponents and still lose. A team can play poorly and yet still win.
My recollection of that win streak is not that the team was blowing teams out night in and night out. They were playing well, certainly, but it's not like the whole team was having an extended hot shooting streak or something. Maybe they were playing above themselves a little bit in terms of clutch performance. Or maybe they just had a run of good luck.
I think it is well established that good teams are more likely to go on extended win streaks than bad ones. A team going on a win streak seems indicative of the ability of the team to play well for extended stretches. Perhaps it suggests some variability in the level of play. But on its own a W-L record doesn't provide you with a lot of useful information.
This is why over the course of a season, or a sample of multiple seasons, analysts tend to focus on broader indicators of team performance, e.g. point differential, +/-, ORTG and DRTG, because these tend to do a better job of capturing how the team is actually performing.
The problem with using the simple aggregate offensive & defensive ratings across the entire span is that, clearly, the ratings for some if not all of the players varied tremendously based on the change in rotations caused by starting or not starting Kyrie.
The "w/o Kyrie" minutes in a game that Kyrie started were performed at a totally different (much worse) level than the "w/o Kyrie" minutes in games he didn't play in.
If you roll those minutes all up together, you get a misleading on/off differential for with/without Kyrie that isn't informative at all. In fact, the simple aggregate numbers suggest that the team without Kyrie has been barely better than mediocre (a scant +1.29 Net Rating). That's the net rating of a 42-43 win club. But the team without Kyrie has won games at the pace of a 50-win club.
So, while yes, tools like ORtg & DRtg are very useful, like with all statistical tools you need to be very careful with how you interpret them.
On whether it is better to include the 16-game winning streak from the start of the 2018 season: While it is generally good to start with the desire to include all data in a data set, it is important to also understand and recognize outliers.
Inclusion of the 16 game winning streak in a simple average leads to the conclusion that the Celtics have been a 64% winning-percentage team, with or without Kyrie.
But in the 122 games _since_ the winning streak ended, they have most definitely NOT been a 64% winning team, with or without Kyrie. That percentage implies that they should have won 78 of those games. But they did not. They won 73 of them (60%).
Dropping that outlier data isn't about the comparison of with or without Kyrie because it doesn't change that comparison. The winning percentage comes out essentially the same with/without Kyrie whether you include the 18 games or not. But when you include it you get a number (for both with AND without Kyrie) that simply does not describe how the team has performed since the outlier. So dropping it is about getting a more accurate winning percentage (for both with AND without Kyrie) that more accurately describes how the team has performed since the outlier and right up to the present.
The team won 60% of it's games last year
after the hot start. They have won 60.5% of it's games so far THIS year. Those numbers are true regardless of the Kyrie question. This is not a 64% winning team right now. Maybe that will change later in the year. Maybe they will go on another hot tear. But it is not who this team has been for well over a year and who they are right now.
Topic: Celtics are unpredictable without Irving (Read 10668 times)