Alternative Season 25 Clustering

So, I've been doing these historical analyses since season 16, but season 25 has seen a big jump in views.  For that reason, I want to reemphasize that these are not meant to be predictions, and to share a bit more analysis.

One challenge in these data analyses is identifying a comparable group.  Back when I started these analyses, I chose to split combat, basketball, and tennis athletes ("CBT") from other athletes, because they had done poorly through the first 15 seasons.  That means that Monica Seles is still in Nikki Bella's comparison group.  It was tempting to drop Monica from that group.  Instead, I broadened the age range to include Paige VanZant, despite her being over a decade younger than Nikki.  Had I dropped Monica instead, the final clusters would have been:

Notice that Nikki's historical average would have filled a gap between Nick and Terrell, pushing both Nikki and Terrell up into Cluster 2, and collapsing Clusters 3 and 4 into one Cluster 3.  A "within-one" interpretation of these clusters would yield more conservative results, broadening the "likely" range for all couples.  However, three clusters are a lot less interesting than four clusters, because the middle cluster can literally place anywhere and stay within one cluster.  That was partly why I went with the more interesting four clusters.

My plan was to see how this season played out before deciding whether the "athlete" category should be re-categorized for future analyses.  Given the unexpectedly high view total, I wanted to share this information now.  Again, these results aren't meant to be predictive, but they can be interesting and insightful for what's historically more or less likely.

DWTS Season 25 - Historical Comparisons

First, a preview / graphical summary of the ensuing DWTS Season 25 historically-based analysis:

Note: See link for an alternative clustering.

This Season 25 analysis is similar to the Season 24 analysis.  These analyses are not meant to be predictive.  They are merely an assessment of relative strength based on historical comparisons.  Job-age comparison groups for each star are listed below.  The "job" categories come from the defunct "Cast DWTS" game formerly on ABC's website.  The age ranges are typically the star's age +/- 5 years (with exceptions noted).  Past season results are scaled from 1-12 (representing 1st place through 12th place) so that results are comparable regardless of field size.  Withdrawals and All-Star season results are excluded.

The comparison groups are listed from strongest average to weakest average.  (Scroll down to view all groups):

Notes: Age range was adjusted for Lindsey, Jordan, Frankie, Nikki, Nick, Barbara, and Debbie because of excessive or insufficient comparison. Job categories were broadened for Drew and Barbara because of inadequate comparison.  "Singer" was the closest available category to "musician" for Lindsay, but that's admittedly a debatable point.

Here's a summary of the above age-job group averages, arranged in ascending clusters of average placement:

Here are the pro averages, from strongest to weakest.  (The average placement for new pros from seasons 4-25 is indicated in red):

Next are the height group averages, for same gender height +/- 1 inch.  These are especially interesting for season 25, imho:

Next are the overall weighted averages*.  They are arranged in clusters of average placement, from strongest to weakest.  Wildcards (highlighted in blue) are Victoria (physically challenged, but with an inspirational story) and Debbie (new pro).  EDIT: Marking Nikki as a wildcard, due to problematic comparison group (explained in addendum post).

*Technical Note: Averages are weighted as 50% Age-Job average, 33% Pro average, 17% Height average, based on correlation analysis of weighted historical averages versus actual historical results.

Comments:

First, a reminder that these are not predictions.  Inevitably, some stars will beat their historical averages and some will fall short.  That said, based on historical comparisons, stars within the same cluster are likely most competitive with each other relative to the field.  Also, stars are more likely to finish within one cluster of their historically-based clustering than to finish further away.  (This was true for the Season 24 analysis.)

Based on historical comparisons, it would be less likely that stars in the top cluster finish lower than 7th; stars in the 2nd cluster finish lower than 10th; stars in the 3rd cluster finish higher than 4th; and stars in the bottom cluster finish higher than 8th.  (These results are summarized graphically in the figure topping this post.)  I'll leave it to the reader to predict their most likely "less likely" results.

DanceSport Records - Recent TV Dancers

(for recent SYTYCD / WOD ballroom and DWTS troupe dancers)

Summaries of Dancesport competitions and U.S. Nationals results, per DanceSportInfo.net.  The database focus is International Latin/Standard.  Dancers' names link to profile pages and partner histories.  These profiles may include duplicate entries.  

Database results are incomplete for some competitions (i.e. partial results), missing for others (e.g. 2013-14 US Amateur Nationals), or excluded (e.g. Pro-Ams, non-DanceSport circuits like American Rhythm/Smooth, or "other" social dance circuits like Cabaret, Salsa, or Swing).

Juv-Jr combines juvenile and junior levels.  Youth+ combines youth and adult amateur levels.  (And pro levels where noted, but most recent TV dancers retired from Dancesport soon after aging out of Youth events.)  Ballroom/10 includes Ten Dance.  Abroad reflects international experience, including Blackpool and World Championships.

Note: These are manual compilations as of August 2018.  Please contact me or comment below regarding errors or corrections you might find.

DanceSportInfo.net Competition Summaries

U.S. National Amateur Championships Summary

Win = 1st place; Top 3 = 2nd or 3rd place; Fnl = made Finals;
SF = semi-finals; QF = quarterfinals; R1 = Round 1 elimination

U.S. National Amateur Titles - Latin / Standard

>

DWTS Pro & Troupe Current Ages*

*Notes
  Current ages and birth dates based on available Internet information
  Uncertain birth dates or ages in lighter print
  Current pros in bold
  Troupe in italics
  Info missing for Fabian Sanchez, Jesse DeSoto, Charlotte Jørgensen, Andrea Hale

DWTS Season 24 - Historical Strength

This Season 24 analysis is similar to last Season 23 analysis.  As those results demonstrate, these analyses are not meant to be predictive.  It's merely an assessment of historical strength.  (And there was plenty of defying history last season.

Job-age comparison groups are composed for each star.  The "job" categories come from the defunct "Cast DWTS" game formerly on ABC's website.  The age range is usually the star's age +/- 5 years (with exceptions noted).  Each season's results are scaled from 1-12 (representing 1st place through 12th place) to make results comparable regardless of field size.  Withdrawals and All-Star season are excluded.

The comparison groups are listed from strongest average to weakest average:


Notes: Age range was broadened for Nancy, Charo, and Mr. T because of insufficient comparison. Job categories were broadened for Heather, David, Charo, Erika, and Mr. T because of inadequate comparison.

Here's a tabular summary of the age-job group averages from the above spreadsheet, arranged in ascending clusters of comparable average placement:

Here are the pro averages, from strongest to weakest:

Finally, here are the overall weighted averages*, including heights.  They are arranged from strongest to weakest average placement, in clusters of comparable strength.  Wildcards (highlighted in blue) include David (first MLB player) and Nancy (first mid-aged female athlete). Also noteworthy is Bonner, this season's most inspirational story.  Averages based on just one or two comparisons are highlighted in red.

*Technical Note:  The factors were weighted as follows: 50% Age-Job average, 33% Pro average, 17% Height average.  These weights were based on correlation analysis between weighted historical averages and actual historical results.

Comments:
First, a reminder that these are not predictions.  Inevitably, some stars will beat their historical averages and some will fall short.  Second, the averages worked out such that there are three evenly defined clusters.  I wouldn't be surprised to see any celebrity finish one cluster higher or lower than history suggests, but I would be surprised were one of the top four to land in the bottom four, or vise versa.  Third, I'm looking forward to how Rashad, Nancy, and Charo fare, given their pro partners' low historical averages versus similar celebrity types having fared better (see spreadsheet and first table above)..