Rockets Strength of Schedule 1st vs. 2nd Half of Season

CH, please revisit this after the trade deadline. It should be interesting to see how trades impact the SoS for the Rockets and their 2nd half opponents.

 

Just show me where they are. Don't think I only deal with androids!

 

We need a moving average or an exponentially weighted moving average to get a better feel for the SOS trend. The graph is nice but it looks too noisy.

 

I like your dedication, and it's impressive work. I do agree though that you didn't have to go through all of that to say this

You could also say that since we have less "superstar" power than most top ten teams, and since we have a certain strength of depth, that we might fare better as the season goes on than other teams who ride certain horses harder.

 

Not to be picky... but as an engineer I'd be careful with some of these stats You should "zero" the graph at the SRS of the Rockets, and make the proper adjustments ([cell #] - [Rockets SRS]). Not sure if you've done this, but that'd make the graph a bit easier for people to understand if you haven't (above the line = better, below the line = worse). The bar graph was a good addition as well, because line graphs are only generally good for showing your dumbass CEO the profit over time.

SRS is nice and all, but it, like all advanced stats, can be used to its own demise. Taking in to account margin at times will skew your data, just look at Hollinger's power rankings, where he weights that stat heavily... You get people with lower records skewed high, and people with better records skewed low due to the fact that when good teams are losing big to bad teams, they tend to throw in the towel and play the bench. Similarly, when bad or even good teams are losing big to good teams, they often make late runs because the good team is either packing it in, or playing its bench (see the Portland/Lakers game last night). Good teams also have a propensity for not showing up against bad teams (while the bad team is hyped to play them) and playing only a quarter or two of basketball and still winning the game, albeit by a smaller margin.

Using a WMA or EMA makes an assumption on the pertinence of successive and previous data. One that may not necessarily be true (should the past data be weighted arithmetically or exponentially?). It'll pop out something, but not sure if it's statistically relevant. Especially because the way you should weigh previous or future data is predicated off your play at that time, which is why Hollinger weighs the last 10 games in his rating systems.

I like the work, but I just don't know if it tells us anything more than what an eyeball test of the schedule would tell us. Probability of going certain records off of weighted factors would be interesting though. Or making a guess on the future record off the data you gathered would also be interesting.

 

My stretch run "analysis"

Team ADJ
TOR 0.4 (Very Likely Win)
@ UTA 0.0 (Coin Flip)
DEN 0.1 (Likely Win)
LAC 0.0 (Coin Flip)
@ BOS 0.0 (Coin Flip)
@ TOR 0.0 (Coin Flip)
@ NJ 0.2 (Likely Win)
@ CLE 0.0 (Coin Flip)
@ OKC -0.3 (Very Likely Loss)
CHA 0.4 (Very Likely Win)
@ LAC -0.1
@ PHO 0.0 (Coin Flip)
LAL 0.0 (Coin Flip)
GSW 0.3 (Very Likely Win)
DAL 0.2 (Likely Win)
SAC 0.4 (Very Likely Win)
@ DAL -0.3 (Very Likely Loss)
MEM 0.1 (Likely Win)
IND 0.0 (Coin Flip)
@ CHI -0.4 (Very Likely Loss)
@ LAL 0.0 (Coin Flip)
@ SAC 0.3 (Very Likely Win)
@ POR -0.1 (Likely Loss)
UTA 0.1 (Likely Win)
PHO 0.3 (Very Likely Win)
@ DEN 0.0 (Coin Flip)
DEN 0.2 (Likely Win)
@ DAL -0.1 (Likely Loss)
@ NO 0.3 (Very Likely Win)
GSW 0.4 (Very Likely Win)
@ MIA -0.4 (Very Likely Loss)
NO 0.4 (Very Likely Win)

Suggesting we finish with 38 or 39 wins.

Took into account our record, opponent's record, home or away, back-to-backs, back-to-back-to-backs, 3 games in 4 nights, 4 games in 5 nights, etc (both us and opponent).

For example, the second game we play against Toronto:

@TOR (0.303)
Rockets record - Tor record = +0.285
Road game -0.1
Back to Back -0.1
3rd game in 4 nights -0.05
Toronto's 3rd game in 4 nights +0.05
Sum total (.285-.25 + .5 = .085)
Simplifying --> (-)0.99 to 0.99 ==> 0.0 (coin flip)
1-2 = Likely
3-4 = Very Likely