Punk-101 wrote:I think the argument is that the eyeball test of "good defense" doesn't necessarily jive with DRtg ranking of "good defense", right?
Maybe middle ground or clarity can be found by giving examples of great, good, average, and poor eyeball test teams past or current and see if that matches with what the DRtg says about them.
Not siding with either argument, BTW.
Nice try counselor......
Statistics can be indicators of a lot of things... that's my point. To grab a stat that was designed way after the fact in a lot of cases and then "backtested" and tweaked to seem meaningful isn't a true indicator.
What if we had a stat about championship teams that showed that teams with 3 of their starting 5 having been on that team under the same coach and system for at least 3 years and compare that to Championships? I'll bet most Championship teams would return out of the data base. What if I entered teams under new coaches and systems with only 2 of the starting 5 together longer than 1 year? Not many would be returned.
O.K.... now one could take that and say that if we can keep the same coach and 3 of our 5 together we've got a great chance at winning a championship.... but why would 3 of 5 stay together under the same coach?..... oops... we don't get that from the stat do we? Contract length, success, playing out the string.... a lot of variables not in the original data set....
I do a lot of work with statistics for a living. Statistics are rear looking indicators.... to project them forward one needs to know why the data sets returned the statistics in question. At best they can be a directional indicator used exclusive of secondary data inputs.... when I use statistics to generate forward looking statements I generally get between 5 and 10 data sets and run 8 to 10 differential statistical variations through the sets and graph the results and look for trends. Trending data is at best a starting point because then you have to assume that the future conditions meet the past criteria of the data and the method of variation used initially.
Will the next team to win the championship be in the top 10 in that stat..... 33% chance that ANY team will be to start with so statistically you're 1/3 there...... to jump into the second third isn't that hard based on the stats that make up that "rating". The variance is paper thin in a lot of those data points.
The real question here is..... are Championship teams good defensively? A second question is does that stat indicate that beyond a doubt as to believe in it's validity in predicting a future event?
My eyes tell me that Championship team are good defensively but some give up a lot of points and score a lot of points.... some don't score that much and don't give up that much.... some get blown out a couple games and win 12 close ones.... some blow teams out and lose a few close ones..... that stat cannot tell me which ones are which....
I could go on.... but suffice to say I'm very skeptical that these new "Fantasy Designed" statistics are in any way "predictive".