Bio not provided
@UnkleDrew Thanks for reading and commenting! You're absolutely right that rule changes play a big factor here, something I completely ignored. I will admit to taking some narrative liberties in attributing some much of this change to analytics. Perhaps knowledge would have been a better way of identifying the driving force. I agree that the rule changes opened the doors for a different style of play, but that style has become the norm not just because of opportunity but because some teams identified that opportunity, adjusted for it and found success. Other teams watched and imitated. I agree none of this happens without the rule changes, but I also don't think it happens without teams recognizing the opportunity and taking advantage. Whether that recognition came purely, or even mostly, from analytics may have been an unfair assumption on my part.
9 months, 1 week ago on The Dissection of Shot Selection: Historical Trends
@ou_sas The one I use is, I believe closer to the Basketball Prospectus uses. I felt that looking at averages and standard deviations would be more accurate than assigning points for reaching certain thresholds. Both have merit, I just felt like this system was more scalpel than baseball bat.
1 year, 8 months ago on Clusters of Scarcity
@ou_sas Ayer does not provide data for any other player combinations, but does say that their results were not statistically significant. I agree that looking at championships team would be pretty interesting.
Durant was not the only one who was right on the edge of two different clusters. I tried to stay consistent and just mark them wherever they were closest, but it would add a whole other layer of complexity if we look at players that were close to multiple clusters and what that meant for the overall trends. I'm not sure how Ayer clustered players so his method may have put guys into clusters in a more definitive way than my Similarity Scores. It was just the best way for me to work backwards and try to replicate. If you're interested I also used this similarity score method to find some statistical comparables for draft prospects - http://www.hickory-high.com/?page_id=1740 The numbers for this season should be done in the next few weeks.
@ou_sas I used a modified similarity score. Each player was compared to each statistical cluster by how many standard deviations away they were in each category. The number of Standard deviations is multiplied by 50 and subtracted from 1000. In the end you have a score between 0-1000 showing how similar the two profiles are. Whichever cluster had the highest similarity score for each player was where I placed them. If you follow the link to the Google Spreadsheet you can click through and see the similarity score pages for each player.
@SirThursday I completely agree that there may be some different statistics to make those clusters more exact and meaningful. I used a modified Similarity Score to put the players into each category, so the players are being compared to the cluster template not too each other. Kevin Durant was closest to the Cluster 2 average than any other cluster but he still could and does compare to that average in a very different way than Wade does. My method leaves plenty to be desired but it was the best approximation I could come up with.