Age-Job Assessment of DWTS Season 17

After 15 non-allstar seasons, there is enough DWTS data to compare new contestants to past contestants with regard to job and age.  A previous blog entry broke down the performance of nine job categories into young, middle, and senior age groups.  Here, we look at the Season 17 field of contestants, first by the age-job groups, and then by their overall projected average of historical age-job, height, and pro performance.  The goal here is not prediction, but assessment of relative strength and weakness.

Historical Age-Job Groups

The Excel table below shows the historical averages of Age-Job groups for the twelve Season 17 DWTS contestants.  ("Historical average" refers to the average final placement after adjusting results to a standard field of 12 contestants.)  Note: Scroll down to see more data.



Overall Projections for Season 17 Contestants

Combining the Age-Job historical averages from above with historical height and pro averages, we can project an overall average for each current contestant.  For this assessment, height averages are based on the historical results of previous contestants of the same gender who were within one inch of the current contestant's height.  Pro averages are based on the average historical result for each pro.  (Historical results are adjusted to a standard field size of 12 contestants.)

Here are the results, clustered into three groups from relatively strongest to relatively weakest contestants (in a historical sense).  **Additional notes are explained at the end of this blog.

Observations

Emphasizing again that this is not a prediction, but an assessment of relative strength and weakness, here are a few personal observations.  I find the lone senior male Other-TV comparison for Bill Nye, namely Jerry Springer, to look very appropriate.  I don't think Nye will last as long as Springer, but there are definite fanbase similarities there.

Looking at the 40-somethings, we see both Berkley's and Remini's projected average being seriously depressed by how poorly other 40's actresses have done.  From the Age-Job groups, Remini appears more like her cohorts in shared background than does Berkley, imho.  The general expectation seems to be that Berkley's dance experience, combined with partner Val's recent success, will serve her well.  However, both Remini and Berkley--and their partners--may face the hidden challenges that have hurt other females of their age group.  I've often wondered if "busy Mom" syndrome hasn't hurt this group.  If so, then the new practice schedule may prove beneficial.

For K. Johnson, being an athlete is still an advantage in his 40's, but he's not as advantaged as younger male athletes have been.  Here, his average projects into the middle cluster, rather than the top cluster.

Finally, I'm heartened by the strong, middle averages for Valerie Harper and Jack Osbourne.  Obviously, these historical averages are blind to their medical histories, but I do hope both can stick around awhile.

**Additional Notes
a) The overall averages are weighted so that Age-Job is 50% more important than Pro average, while Height average is 50% less important.  The formula, based on historical correlations, is:
    Overall weighted average = (1.5*Age-Job Average + 0.5*Height Average + Pro Average) / 3
b) The average for new pros is based on the average of the six 1st year pros in seasons 13 and 16.
c) The height average for Nicole is based on all females under 5', because of a lack of data for females her size.
d) The pro average for Val includes All-Stars, which is otherwise excluded from the data.  This was a subjective call, based on post hoc results.
e) Where needed, age-job average is the first tie-breaker, and pro average is the second tie-breaker.
f) Withdrawals are not included in the historical averages, nor in the Age-Job groups.

This is not a prediction, but...

a look at the historical advantages and disadvantages for the stars on the upcoming 16th season of DWTS.  This is based on past season results, with the raw data accessible here.

Age

Top-to-bottom, this chart shows the most statistically age-advantaged stars to the most statistically age-disadvantaged stars:

The average placement (Avg Place) is the averaged result of all previous stars who fit the age group of the current star when they competed.  Those results are adjusted to a scale of 1st through 12th place (a.k.a. "normalized") prior to averaging.

For most stars, the comparative age group consists of same gender stars within 2 years of their current age.  For some stars, the comparative group was too small to calculate a reliable average, so a bigger comparative group was improvised.  These improvised groups are in italics, and their averages are in red.

Height

Following the same idea described for age, this chart lists the most statistically height-advantaged stars down to the most statistically height-disadvantaged stars:

Heights are shown in inches, and height groups are defined as males or females within 1 inch of the current star's height.

Job

This chart shows the most job-advantaged stars down to the most job-disadvantaged stars:

Job groups are based on the categories defined on ABC's fantasy DWTS casting site, and are separated by gender.  Also, a subgroup of Basketball, Boxing, & Tennis (BBT) Athletes was distinguished from the other Athletes.  For those groups with too few stars to yield a reliable average, males & females were combined to create a larger comparison group.  These combined groups and averages are shown in italics and red type.

Pros

This chart shows the most statistically pro-advantaged stars down to the potentially pro-disadvantaged stars:

The averages are most meaningful and stable for the pros who have had the most star partners.  Only Derek and Cheryl are notably advantageous.  Despite last season's win, Tony is historically somewhat disadvantageous, but that's only relative to the other established pros.  (His average placement of about 7th place isn't bad, but it's about a full place below the 4 pros above him.)  The potentially disadvantaged group is more a wild card group that includes three first-time new pros, and two still relatively new pros.

For the first-timers, an improvised average was based on the first-year results of the 7 most recent new pros.  (Namely, Peta, Tristan, Val, Damian, Anna Demidova, Chelsie, and Dmitry.)  Aside from being recent enough to be relevant to the current show's dynamics, this group includes pros with similar backgrounds as the new pros.  (i.e. SYTYD crossovers Chelsie & Lindsay, Burn the Floor alums Peta & Sharna, pro competitors Val & Gleb.)

Technically, Val and Tristan rank below the improvised new pro average, but their averages (and Peta's) are still volatile.  Meaningful and stable averages for the first-year and fourth-year pros won't be available until they have competed with at least 5-7 partners.

Alternate New Pro Average:

Another way to improvise an average for the new pros is to use the average of the Season 13 new pros (Peta, Tristan, Val).  Although a smaller group, their situation may be more comparable to the new pros', likewise being a recent set of three new pros.  The new improvised average falls below Val & Tristan:

Overall Averages

This chart indicates the overall average of the historical age, height, job, and pro group averages for each current star.  They are listed from the most overall statistically advantaged star down to the most overall statistically disadvantaged star:

I used job average as a tie-breaker where needed.  The reasoning is that job is the nearest available proxy for fanbase and skills.

Weighted Average

The simple overall averages showed little separation of the middle group.  I tried recalculating the numbers with a weighted average.  Based on both correlations (in the spreadsheet linked above), and the idea of job as a fanbase and skill proxy, I increased the importance of job by 50%, and decreased the importance of height by 50%.  Using job average as a tie-breaker, here are the weighted overall averages, listed from the most overall advantaged star to most overall disadvantaged star:

The results are clustered by three, because they break down somewhat nicely that way.

Alternate New Pro Average:

The bottom 2 clusters are slightly different if the alternate new pro average is used, with D.L. & Victor switching positions:

Comments

Again, this is not a prediction post, but rather an assessment of historical advantages and disadvantages.  That said, the weighted averages appear consistent with general predictions that fans and media pundits are making.  (The one surprise, personally, is Sean's better-than-expected average, resulting primarily from his youth.)  Otherwise, I generally think that the stars' final placements will transition within these clusters, or possibly transition to within one cluster above or one cluster below.

That is, I imagine that the three most advantaged stars (Aly, Jacoby, Kellie) will each finish somewhere in the top 6; that the three most disadvantaged stars (D.L./Victor, Lisa, and Andy) will each finish somewhere in the bottom 6; that the top middle cluster (Dorothy, Ingo, & Zendaya) will each finish somewhere 1-9; and that the bottom middle cluster (Sean, Winona, Victor/D.L.) will each finish somewhere 4-12.  (But again, that's not a prediction.  It's just my personal interpretation of the numbers.)

Forecasting Star Age for Dancing's Newest Pros

Peta, Tristan, and Val have partnered three stars each since becoming pros on Season 13 of Dancing with the Stars.  Knowing the significance of star age, we can compare those triads to previous pros' and try to guess their future stars' age range.

Following are brief presentations of star age percentages, triad averages, and triad extremes, followed by "best guess" forecasts.

Overall Percentages of Younger and Older Stars

Here is how the new pros' current percentages of under- and over-40 stars rank against other pros who have partnered at least three stars:

For reference, the pros' average competition age is listed along with their average stars' age.  The bright yellow zone indicates pros with 80% or more stars under 40.  The dim gray zone indicates pros with 50% or less stars under 40.  The mid-range neutral zone falls between.  It appears that the bright zone is dominated by younger pros, while the gray zone is dominated by older and/or married pros.

Peta, who at 25-26 has been the second youngest pro in her three seasons, has consistently partnered under 40 stars.  Val, also 25-26, falls into the mid-range percentage of under 40 stars.  Tristan, a little older at 29-30, has yet to partner an under 40 star.

The numbers suggest that they will tend to stay in or near their current zones.  Eventually, Tristan should attain the 40%-50% range shown by other pros in the gray zone, but it may take awhile.  His closest comparison, Kym, stood at 25% after 8 partners.  Only recently has she gotten mostly under 40 partners.

Triad Comparisons

Overall percentages are useful for long-term projections, but we need more information to forecast star ages based on the first three, or triad, of stars.  To guess what may be ahead for the new pros, we can consider the progression of the veteran pros' triads in regard to average, minimum, and maximum star age.

Here are the pros' average star age across triads of stars.  They are listed in order of ascending 1st triad average:

The new pros are highlighted in blue and are listed among the veteran pros with multiple triads.  Additional pros with are listed separately.  Bright yellow indicates triad averages that were younger than the 1st triad by 5+ years.  Gray indicates triad averages that were older than the 1st triad by 5+ years.  Generally, pros either stay close to their 1st triad average, or get older star averages as they themselves get older.

This can be relative.  Mark's stars have gotten younger in later triads, probably because through 10 seasons, he remains the youngest male pro.  And pro roles can change.  Kym and Karina's bright "younger" triads coincide (season-wise) with Anna's gray "older" triad, suggesting that roles have shifted.

To date, Kym has the "brightest" decrease from a 1st triad average, her 4th triad being 10 years younger.  Maks has the "grayest" such increase, his 4th triad average being 15 years older than his 1st.  Generally, the brightest decreases from a 1st triad average have been about 5 years younger, and the grayest increases have been about 10 years older.

DWTS fans often are more interested in extremes like youngest and oldest than in averages.  Here are the pros' youngest star ages across triads, listed in order of ascending 1st triad minimum:

Bright yellow indicates a triad minimum 5+ years younger than the 1st triad minimum; and gray indicates a triad minimum 5+ years older than the 1st triad minimum.  Although the focus is on completed triads, significant minimums from incomplete triads are shown in italics.

Generally, pros' youngest stars remain consistent across triads.  Maks and Kym are misleading cases, who appear to have grayed across triads, because they had unusually young (for them) stars in their 1st triads.  Otherwise, Maks consistently has partnered the 30-and-over crowd, and Kym the 35-and-over crowd.  

Typically, the biggest changes are plus / minus 10 years from the 1st triad minimum, although there are exceptions.  For Derek, the 14 year decrease from 1st triad Shannon Elizabeth to would-be-4th triad Shawn Johnson represents the biggest drop in 1st-to-subsequent triad minimum, possibly due to the atypical All-Star season.

The biggest increase is 19 years for Anna:  Her 1st triad minimum was 22-year-old Albert Reed, and her 4th triad minimum was 41-year-old Carson Kressley, who was a replacement for 70-year-old Ryan O'Neal.  If not for that change, Anna's youngest 4th triad star would be have been 52-year-old Jack Wagner for an even more exceptional 30 year increase.

Now, here are the pros' oldest star ages across triads, listed in order of ascending 1st triad maximum:

Bright yellow and dim gray indicate maximums that respectively are 5+ years younger and older than their 1st triad maximums.  A pattern emerges here.  Pros with relatively younger "oldest" stars in their 1st triad tend tot get older "oldest" stars in their subsequent triads.  And pros with relatively older "oldest" stars in their 1st triad tend to get significantly younger "oldest" stars in their subsequent triads.

Typically, pros with 1st triad maximums under age 50 had either similar maximums in subsequent triads, or older maximums by around 15-20 years.  (Mark continues to be the exception, with a significantly younger 3rd triad maximum.)  Pros with 1st triad maximums over age 45 tend to have either similar maximums, or about 10 years younger maximums in subsequent triads.  (Note the overlap in Karina's and Anna's triads.)

The biggest decreases from 1st triad maximums are 16 years for Edyta, and 22 years for Kym.  They (and Tristan) had the oldest 1st triad maximums, so were likely to get "more typical" maximums in later triads.  The biggest increase on record is 26 years for Ashly, who returned in season 10 to partner 80-year-old Buzz Aldrin in an unusual circumstance.  Among the regular pros, the biggest increase was 23 years for Cheryl (from Drew Lachey to Wayne Newton).

Best Guess Forecasts

So what can we guess for the newest pros?

For Peta:  Her 1st triad numbers are most similar to veteran Derek's 1st triad numbers.  On that comparison, we can expect Peta to get mostly under 40 stars for as long as she is one of the youngest female pros.  Her future triad averages are likely to fall in the early 30's to early 40's range.  Like Derek (and Chelsie), an atypical triad maximum of mid 40's to mid 50's is possible.

For Val:  The closest 1st triad comparisons are Cheryl and Lacey.  If Val follows a similar trajectory, then his average star ages likely will fall into the mid 30's to mid 40's range, and his oldest stars in the late 50's to early 60's range.  Minimums are a little harder to forecast.  The Cheryl / Lacey pattern would suggest late 20's to early 30's.  But because Val and Mark currently are the youngest male pros, Val could be paired with someone younger.

For Tristan:  His closest comparisons are Kym and Edyta who likewise had 1st triad stars over age 60.  Unlike them, Tristan did not get an atypically young 30-year-old in his 1st triad.  We can guess that his future triads will be similar to Kym's and Edyta's, with the possibility of an atypically younger partner along the way.  That translates into 40-something averages, mid-to-late 30's minimums, and early 50's to early 60's maximums.  His 1st triad maximum is exceptional, however, and a drop of 20+ years off the maximum, like Kym's, is possible.

We should note that those atypical age pairings mentioned in the preceding paragraphs can result from celebrity requests (e.g. Maks-Kirstie, Cheryl-Rob), physical needs (e.g. Season 4 Kym-Joey), or thematic purpose (e.g. Anna-Evan right after his Olympic gold that was criticized by Russian officials).  Also, last minute partner scrambling has been rumored to result in atypical age pairings like Chelsie-Michael Bolton and the aforementioned Ashly-Buzz Aldrin.

Regarding any potential new pros introduced on Season 16:  The safest blind bet is that they will fall into the "neutral" zone of 50% to 80% stars under 40, with 1st triad profiles comparable to neutral zone pros like Maks / Karina / Lacey / Val / Cheryl / Louis.  That bet will be heavily affected by their relative age and height, however.  (I may post a follow-up after the pro lineup has been announced.)

Finally, a personal aside:  I have argued in various TWOP discussions that Val and Tristan are similar to Cheryl and Kym, respectively.  Numbers like the ones presented here are partly why I think that.  (Along with their personalities and their stars' personalities.)  However you feel about those comparisons, remember that both Cheryl and Kym have won MBT's.  :-)

How old is "old" for Dancing with the Stars?

Fans of Dancing with the Stars have become adept at handicapping a star's chances to win the Mirror Ball Trophy. Older celebrities are presumed to be disadvantaged, but how old is "older"? Would you believe 38? Several celebrities have won DWTS in their mid-thirties, but there's been a notable gap after 37. From youngest to oldest, the winners' ages have been:

It appears that mostly very young athletes, young performers, mid-30's athletes, and early 50's pop culture icons have dominated the winner's list. The age 37 cut-off could be a fluke. Possibly, a Kristi Yamaguchi or Donald Driver could have won in their 40's . However, it's not just winners, but finalists who have been dominated by the 37 and under crowd. The historical percentages through 15 seasons are:

The disproportionate success of young stars is apparent, comprising over 80% of all finalists. The struggles of older stars also is apparent, with a finalists rate (16%) about one-third their overall participation rate (46%).  While performance generally declines with age, further analysis shows that males and females decline differently. Rounded to the nearest full place, their average placements (normalized on a scale of 1st place to 12th place) are:

The similar 5th place average of young males and young females suggests that they can compete directly with each other. Males' results decline progressively with age. Surprisingly, middle-aged females struggle for placement as much as senior males, but senior females compare to middle-aged males. Conceivably, these middle- and senior-aged groups could benefit from competing with each other, under age-adjusted rules.

Why middle-aged females have struggled so dramatically is unclear. As stated previously, young Athletes in their mid-thirties have done well. That raises the question whether not just age, but an interaction of job and age affects performance. Considering the seven job categories of stars that ABC identifies in their online fantasy casting forum, here are the averages of the various groups' normalized (1 to 12) placements, with the number of corresponding stars in parentheses:

Averages have been rounded to the nearest half place. If we assume that young stars have the highest potential placement, then their performance sets a "baseline" for each job category. The jobs are listed from the best to worst average placement for young stars, with bright yellow indicating an average of 5th place or better, and light gray indicating an average of 8th place or worse. (Shading indicates a weak average based on just 1 or 2 stars, while dark gray indicates an absence of stars.)  Overall, Actors, Athletes, and Singers have the highest baseline potential.  Because we know that males and females age differently, we can parse the numbers by gender:

Several things stand out. First, there has been a lack of middle-aged and senior Reality stars. The youth baseline indicates that Reality stars do "OK", so maybe over-37 Reality stars would do "OK".

Second, as poorly as middle-aged actresses have generally placed, senior actresses have placed much better. The senior actresses include winner Jennifer Grey and one-time finalist Kirstie Alley, as well as Susan Lucci, Jane Seymour, Florence Henderson, and Cloris Leachman . Nostalgia may have helped these women, along with a bit of train-wreck humor in Leachman's case. Among the middle-aged actresses, Lisa Rinna, Marlee Matlin, and Melissa Gilbert are among those who have done OK. Tatum O'Neal, Denise Richards, and Shannen Doherty are among those who haven't. (And there's Pamela Anderson, who has and who hasn't.)  Why middle-aged actresses fare so much worse than young actresses--a group that includes Kelly Monaco, Jennie Garth, Marissa Jaret Winokur, Shannon Elizabeth, and Melissa Joan Hart--is unclear, especially given the relatively greater fame of the middle-aged group. Nevertheless, their struggles contribute to the overall lack of success for middle-aged females.

Third, there have been many over-37 Others, but not so many over-37 Athletes. Given the high baseline average of young Athletes, this factor could explain the drop-off in performance by over-37 stars, especially female stars. Both the Other and Athlete job categories have notable sub-groups worth separating. It's general knowledge that Basketball, Boxing, and Tennis (BBT) athletes have done notoriously worse than other athletes. (This could be due to center of balance issues related to height or defensive crouch positions.)  And while Others are generally not performers, there is a distinction between Others who work in front of TV cameras like hosts and reporters, and Others who don't. Here are the averages for these subgroups:

The overall strength of non-BBT Athletes becomes clear: non-BBT Athletes have the highest youth baseline average of any group. We further see that male athletes continue to do well in their 40's, though none have won and only Jerry Rice has made the Finals. The absence of middle-aged female Athletes could partially explain why their placements drop off significantly. Or like female Actors, they too might struggle in middle-age and recover as seniors. Hopefully, we'll have the opportunity to learn which is the case. For the Others job category, the influence of TV exposure appears to make a difference. Certainly, seniors Jerry Springer and Nancy Grace appeared to benefit from their fanbases. However, the super-high success of very young Other-TV females could result from Derek and Maks having partnered all three (Brooke Burke, Maria Menounos, Erin Andrews). Additional data is needed to see whether this group reliably yields exceptional results

Comments

Having seen the disparity between young stars versus middle-aged and senior stars, we can ask, "Can the playing field be leveled?" Drawing attention to the problem can start conversation about why the disparity exists. Campaigning for more middle-aged and senior Athletes might help, especially female Athletes. The same might go for Reality stars and possibly Others. Suggesting reasonable adjustments to scoring criteria for various age groups can address the root of the problem:  That of having such an age-varied mix of stars compete directly against each other in the first place. Ultimately, advocating for a separation of age groups might be the best solution. Defining individual seasons as Under-40 Stars or Over-40 Stars should create fairer competitions. (Or Under-/Over-39 , if the semantics make the demo-sensitive producers happier.)

Simply drawing attention to the problem may not be enough. At best, it might spur them to cast more Jennifer Grey, Osmond, Kirstie Alley types with mega-fanbases, leaving most over 37 stars at a continuing disadvantage.  At worst, it might unfortunately lead the producers to manipulate a PC winner to prove a point, and then revert to the usual youth-dominated format.  Lasting format changes are needed. If it turns out that there's not enough concern about age disparity on "Dancing with the Stars", then anyone who does care can simply choose to support the older stars they like and feel deserve their vote. After all, the only criteria for "best dancer" or "deserving winner" that any viewer is obligated to vote for is their own.

Update

The new Season 16 cast addresses some of the holes mentioned above.  Namely, we have a senior Reality star (Lisa Vanderpump) and a senior Female Athlete (Dorothy Hamill).  There also are no 30-something stars, which sets the season up as something of a lab experiment in terms of how very young teens and 20-something dancers are perceived by fans and judges, versus how middle aged and senior dancers are perceived.  Should be interesting.