re: SB Nation's NBA divison realignment

quote:

---

You did a lot of work for nothing. You can't multiply their percentage by the new opposing SOS. That dog don't hunt. That's lazy math.

---

How else would you predict outcomes? You gotta use simple stupid math because I don't have an algorithm off the top of my finger to run 100 sample sizes of games to get a more accurate percentage or depiction if said teams were supposed to play 100 times to reduce the SD and p-value. I have 2-4 games from last year to re-predict the outcomes.

I gave you two methods:
1. Lazy math which multiplied their cumulative win % vs. certain batches of opponents and grouped them together to reduce the standard deviation; higher sample size, lower SD.

i.e. Grouping said opponents OKC and Denver, who they played 7 total times last year and went 6-1 cumulatively, got the win % (.857) and multiplied that by 10 = 8.57 wins.

2. Actually broke down individual win %'s of Phoenix vs. their opponents which created a slightly more precise prediction, but not as accurate.

i.e. They went 3-1 vs an Portland last year. In the new system, they only play Portland 3 times. It'd be dumb to wipe out game 4 off the map, so using win % off those 4 games and converting it to 3 games would give me a better prediction.

By calculations, due to the way SBN had zoned it out, a lot of the east games stayed the same and didn't alter much (i.e. played milwaukee twice both years) and zones that were altered where they went 2->3, the win distribution was barely changed. Why? Because teams like orlando they went 2-0, whereas miami they went 0-2. That means between those 4 games, it was a .500 record, adding a game vs. each opponent (2 more), odds are, it yields similar results and it i

The SOS was changed 1.3% and the mathematical odds balanced itself out. It isn't selection bias by taking out teams they did poorly against (i.e. Miami) and converting those games to teams they did well against (i.e. Orlando) or vise versa to make them do well. If it was, the numbers would be skewed and the p-value would be greater than 0.05 (which correlates to a change that is greater than +/- 4.1 wins).

All because the SOS increased by 1.2% it doesn't mean they're going to drop from 48 wins to 40 wins or something because the schedule is harder. Your theory doesn't take into account or factor in how well phoenix did against the sample sizes.