New Site - CraftedNBA

[dtkavana]

I plan to add the rolling 250 day version of RAPM that can be found here ->

http://canzhiye.com/rapm/

[dtkavana]

I have made some changes to the Team pages, specifically to the 'Young Core' tab.

The idea was to figure out how good a team's young core would be at their peak if surrounded by average NBA talent. There is a more detailed description of the method on each of the Team pages. The win totals are a bit on the lower side, but this is bc while we have surrounded the Young Core with a few decent players, the rosters are built with the idea that the players listed are the best players on the team, i.e. Boston no longer has a player as good as Al Horford surrounding the Young Core players.


Holler if you have any thoughts/questions,

Thanks

[dtkavana]

Side Note:

The Wins Above Replacement player projections come from 538's CARMELO, and are from preseason. Therefore, they will underestimate players who have surprised/had great season's (like a D'Aaron Fox) and overrated other players.

Also, While a player like Jayson Tatum is listed at an 8.1 peak WAR, this is what they consider his most likely outcome. It is just as possible that he goes over that 8.1 then he goes under. CARMELO lists a broader range for each player, so if you think he is likely to be better than that 8.1, you can use the upper range of the confidence interval which is around a 14.

https://projects.fivethirtyeight.com/ca ... son-tatum/

[dtkavana]

Hey guys,

I have made some changes to the way we calculate our Combo +- formula.

I am now standardizing each of the variables (RPM, PIPM, BPM, etc) to convert them to the same scale before combining them and then dividing by the number of variables. In this case, the total number of variables is 10 as we weight RPM and PIPM 3 times as much as we weight BPM, CARMELO, MYPIPM and MYRPM. We then take this number and convert it to the MYRPM scale. You can find this formula as well as definitions for many other advanced stats on our http://craftednba.com/glossary page.

I have also added percentiles next to any mentions of Combo +- on the site to give users a better idea of exactly what that number says about a player. You will find these percentiles on Team pages and any individual player page.

As always, do let me know if there is anything you would like to see added to the site.


Thanks

[Crow]

It sounds good.

I visit but I should visit more. I am often on mobile, so I guess I haven't gotten into the habit. But I should go to your site when I can, when I want to look at metrics.

[Mike G]

That's great, but how about updating; it's been a few weeks?

[dtkavana]

All plus minus and basketball reference stats have been updated.

[Mike G]

Thanks, man.
You still have no my-RPM showing for some veteran players: Cousins, Augustin, Markkanen, Nene`, Ilyasova, Caboclo, Brewer, Lee, Rose, Waiters, Nowitzki, Canaan. Some of these might be zeroes; but Dirk, Nene, Cousins?
http://craftednba.com/player-cats/king
Also, still can't imagine why rookies' PIPM and my-PIPM would be different values. Most are close, but some are off by 1.0 or more.

[dtkavana]

The players who have missing MYRPM data are due to them not being on the spreadsheet that was shared by Englemann just prior to the season. I may have missed someone, but that is the case for most of them.

As for the PIPM vs MYPIPM for rookies, I believe that all rookies start with an MYRPM before they have even played a game which is based upon Jacob's predictive PIPM. The two numbers merge more quickly for rookies and I would imagine by the end of the season, or once a certain minutes threshold is hit they would even out.

[Mike G]

The reason I suspect some of those missing values are really zero, is that there aren't any players with 0.00 myRPM.

Code: Select all

0.08   Justin Holiday
0.08   Aron Baynes
0.05   Stanley Johnson
0.04   Shaquille Harrison
   
-0.03   Jonas Jerebko
-0.03   John Collins
-0.05   Channing Frye
-0.06   Taj Gibson
-0.06   Ed Davis
-0.09   Luke Kennard

That gap of .07 is unusual, considering the density of players in the middle of the spectrum.
Here's something: Cousins, Dirk, and Nene` did not appear in any games before Dec. 1. I have seen other RPM lists in which late arrivals never make it into the list at espn.

[Crow]

The ESPN pages for RPM have had repeated issues with dropped players, questionable position assignments, lack of sort by team and last name. Appeals from readers and the RPM creators go unanswered. Apparently the analytic leads there recently have either not had the power or interest to enhance these pages or put RPM on player profile pages. Oh well. But the data itself is a welcome asset.

[Crow]

Hey guys,

I have made some changes to the way we calculate our Combo +- formula.

I am now standardizing each of the variables (RPM, PIPM, BPM, etc) to convert them to the same scale before combining them and then dividing by the number of variables. In this case, the total number of variables is 10 as we weight RPM and PIPM 3 times as much as we weight BPM, CARMELO, MYPIPM and MYRPM. We then take this number and convert it to the MYRPM scale. You can find this formula as well as definitions for many other advanced stats on our http://craftednba.com/glossary page.

Is the weighting still as shown in this glossary link or has it been simplified to fewer metrics? If as shown at glossary, is the less common metric data available at your site anywhere?

Have you contacted analytic oriented media / Twitter accounts to try to get them to use your blended PM? I probably should use but sometimes had difficulty getting to the player data on mobile and it dissuaded me. Got to it ok today so might use more in future.

[dtkavana]

Hi Crow,

Yes, the player pages are wonky at times as well as some other issues with the site that I am not happy about. To remedy these issues, I am switching technologies and will be using Nuxt.js for the site in the future. The shift to the new site should be completed before the NBA resumes in late July and will also include a "Hall of Fame" section with stats on ~100-200 historical players.

The weighting is still as shown in the glossary.