20 replies to this topic

[katdlf]

A while ago, I suggested starting a numbers thread and got a positive response. Now's a good time because there's sufficient data from 14 seasons, and Season 15 could be very different with its All Star format.

To start us off, I've posted a spreadsheet at the blog linked below. It's interactive, and hopefully fun. There may be errors or omissions, and I welcome all help and suggestions with those. The focus is celebrity "types" that have done well, poorly, and been assigned to each pro. It's a kitchen sink of stuff, and you'll have to scroll around to see everything. Let me know if there are problems accessing it.

DWTS Stats

There are several possible topics for discussion in these numbers. Hopefully there's enough interest to keep this thread fun and active for awhile, even for the math-phobic. (About that, I'm comfortable enough with the basics, but if there are statistics experts here, then please contribute to this thread. :) )

BTW, if you're interested in scoring and prediction, the PureDWTS blog does regular analyses on those topics. It was their predictions series that prompted this spreadsheet to better understand past results, so the breakdown of factors is somewhat similar.

**UPDATE** Feb 5, 2013: A friendlier version of the spreadsheet is now up.

Edited by katdlf, Feb 5, 2013 @ 6:17 PM.

[katdlf]

OK, there's been over 100 views of this thread. I'll take that as a sign that there's enough interest and curiosity to go ahead and start a topic. Like I said in my opening post, the linked spreadsheet is a bit of a kitchen sink, so I'll post the numbers most relevant to the topic below. Please reference the "AgeStat" worksheet in the linked Excel workbook for the full analysis.

The topic is: Performance of 40-Something Females. The numbers are:

AgeGrp__N12Avg__#Stars

<40 F_____6.0_____48
40s F_____8.2_____20
50+ F_____6.7_____11

<40 M_____5.1_____46
40s M_____7.3_____19
50+ M_____8.5_____18

"N12" means "normalized to 12", 12 being the typical DWTS field size. This allows us to compare placement across seasons of varying field size. "N12Avg" (or N12 Average) means the average N12 placement of all Seasons 1-14 stars who fit the Age Groups (AgeGrp) of under-40 Females/Males, 40-something F/M, and 50-and-over F/M. Think of this as the average 1-12 (or first to last) placement of all stars who have fit that age category.

The expected N12 Average for a group of stars performing equally well across 1-12 placements would be 6.5, because that's the average of all placements 1-12. A large group with an N12 average below 6.5 would be placing above (better than) statistical expectations. A large group with an N12 average above 6.5 would be placing below (worse than) statistical expectations.

The numbers suggest that under-40 Males on DWTS have performed better than statistically expected as a group. Under-40 Females, 50+ Females, and 40-something Males have performed close to expectations (within 1.0 the expected N12 Average of 6.5). Performing below expectations are 40-something Females and 50+ Males. (Please note that these results are most meaningful for the larger groups, i.e. under-40 Males and Females, and relatively less meaningful for smaller groups, i.e. 50+ Females.)

So that's what the numbers show. Now, what do the numbers mean? It makes sense that Male stars' placements would get progressively worse as they age. It doesn't make immediate sense that Female stars' placements would suddenly drop off in their 40's, then seem to recover after 50. I have some ideas about this, but I'd like to hear your thoughts about this surprising result. Any theories on why 40-something Females have struggled to date??

Edited by katdlf, Jul 26, 2012 @ 1:31 AM.

[Elsa]

I guess I have a controversial opinion about all this. I don't think the scores reflect reality. I think they're manipulated. I think the voting audience is largely female, and women like to vote for footballers (Emmitt, Donald, Hines - no way were any of them the best dancers of their seasons) and sweet "lovable" and/or cute guys (Apolo, Helio, Drew, Donny was pretty sweet and a past heart-throb for the older female viewer, who's the average viewer and probably voter - some great, like Drew, some just cute, like Helio), etc.

I think with females, the show usually props them to victory (or to the finals) - and Derek and Mark use razzmatazz choreography that makes them seem better than they are which helps the show/judges prop them with higher than deserved scores and praise. So since Derek and Mark are so young, they give them young girls to win with - and they don't bother putting the 40's women with their pet pros, chosen to prop females to victory. Why women do a bit better when older may be compassion or nostalgia - not sure as I haven't studied who they are, but Jennifer Grey is the only older female winner, right?

I may be way off here but I don't think the scores reflect "true" scoring. They might reflect popularity - like after the first week, the show starts giving higher scores to the higher vote getters because they realize they're stuck with them no matter what.

IF I thought the scores were fair, I guess I could get into the analysis better, but I can't because I don't. It's just a popularity contest led by the show (who wants some females to not make it an all-male-winner thing) and by the middle-aged women (and fangirl blocs) who bother to pick up that phone (or get online.) I don't think men vote in big enough numbers to make any difference.

I could be totally wrong on this, but I just don't trust the reliability of the numbers at ALL.

What I really wish is they'd do what So You Think You Can Dance is starting this season - have two winners every season - a male and a female. Why not? How expensive can those stupid mirrorballs be anyway? And it really is apples vs oranges in so many ways -- the dances themselves are different dances, the female's part of a rumba is WAY different than the man's part and they have to show so much more skin -- and then, the voters are not equally represented from the genders -- so it doesn't make sense to me that they're compared as if they're doing the same things and voted on by objective viewers and judged objectively - they're not!

Edited by Elsa, Jul 27, 2012 @ 12:56 AM.

[Serendi]

Thanks for the topic!

What immediately occurs to me is, are the 40-50 and 50+ women mothers, and if so, were their kids adult as of, say, three years before they did the show? 'Cause I think a lot of women get a new direction once the birds have flown the nest, and this could definitely include getting into better shape.

Had thoughts regarding forceful personalities keeping your fame viable (mostly because Cloris Leachman, Kirstie Alley, and Gladys Knight all seemed pretty definite personalities; they didn't all get that far, though). I suspect that idea won't pan out for this question. Still, would like to (sometime when my brain is more functional) look at that for both men and women and see if it has any effect on anything at all, pro or con.

Not that I know much about most of the people before Season 11, when I started watching. Hmm... (Need to follow QueenL's DWTS Through the Years YouTube links.)

[katdlf]

Thanks Elsa and Serendi for your posts. Sorry for these belated follow-ups.

----------------------------------------------------------------------------

IF I thought the scores were fair, I guess I could get into the analysis better, but I can't because I don't. ...

FWIW, the analysis here is placement-based rather than score-based. You may not realize it, but you've suggested several (strongly-felt) hypotheses for why 40's Females have placed below statistical average. :-)

WRT your post: It's possible that 40's Females are scored unfairly. That's difficult to assess without a parameter of "correct" scoring. It's also possible that female voters favor male stars, particularly young ones. That can explain <40 Males doing so well, but doesn't address 40's Females versus other Females.

since Derek and Mark are so young, they give them young girls to win with - and they don't bother putting the 40's women with their pet pros, chosen to prop females to victory. Why women do a bit better when older may be compassion or nostalgia - not sure as I haven't studied who they are, but Jennifer Grey is the only older female winner, right?

Jennifer Grey and Donny Osmond are the only 50+ winners. Nostalgia could explain them and 50+ Females doing OK, but apparently not other 50+ Males? Just a thought, but that could reflect differential casting. IMHO, whether by chance or by choice, 50+ Females tend to present as "icons", whereas 50+ Males tend to present as "characters". Or it could be that 50+ Females retain more dance-related skills than 50+ Males, perhaps due to females generally having more dance background across all age categories.

WRT Derek and Mark: Derek's partners have consistently placed in the top half, including his two not-so-young partners (Grey and Lake), so it's true that his results have been exceptional, and arguably favored. OTOH, Mark hasn't had an over-40 star, but his <40 results have been mixed: 5 reached the Finals, the other 5 placed in the bottom half including 3 early eliminees.

Broken into Top (1-4), Middle (5-8) and Bottom (9-12) groups of normalized placement, Tony's and Louis' <40 partners have placed comparably to Mark's, while Maks' have placed *better* than Mark's (higher % of Top placements, lower % of Bottom placements). So overall, Mark's results with this age group really haven't been exceptional. Perhaps more than youth, Mark has been lucky in receiving several partners with consummate skills, personality, and fanbase, especially his winners Shawn and Kristi.

I could be totally wrong on this, but I just don't trust the reliability of the numbers at ALL.

They're reliable within their context, ie. it's true that "40's Females have collectively placed below statistical average through 14 seasons". They're not reliable as broad generalizations nor as future indicators, so skepticism is good. Still, the numbers can be worth exploring, whether to suggest new perspectives, or just to support what you "already know". :-)

----------------------------------------------------------------------------

What immediately occurs to me is, are the 40-50 and 50+ women mothers, and if so, were their kids adult as of, say, three years before they did the show? 'Cause I think a lot of women get a new direction once the birds have flown the nest, and this could definitely include getting into better shape.

Possibly, although being initially out-of-shape didn't prevent 40's Marie Osmond and Ricki Lake from making the Finals, and Kirstie Alley wasn't initially in better shape at 60 than in her 40's (I don't think). Along with Jennifer Grey, they're the only over-40 females to make the Finals. It could be that fitness makes a difference between the middle and bottom placements for these age categories (haven't looked closely), but maybe isn't critical to reaching the top places?

Had thoughts regarding forceful personalities keeping your fame viable (mostly because Cloris Leachman, Kirstie Alley, and Gladys Knight all seemed pretty definite personalities; they didn't all get that far, though). I suspect that idea won't pan out for this question.

It may pan out OK. Here's the list of 50+ males and females and 40's females in order of normalized placement 1-12, and grouped into {top 1-4}, {middle 5-8}, and {bottom 9-12} placements:

50+ Males: {Donny Osmond, John O'Hurley}, {Jerry Springer, George Hamilton, Lawrence Taylor, John Ratzenberger, David Alan Grier, Steve Wozniak}, {Harry Hamlin, Sugar Ray Leonard, Tom DeLay, Wayne Newton, Buzz Aldrin, Jack Wagner, Michael Bolton, Ted McGinley, Penn Jillette, David Hasselhoff}

50+ Females: {Jennifer Grey, Kirstie Alley}, {Nancy Grace, Susan Lucci, Jane Seymour, Cloris Leachman, Priscilla Presley, Gladys Knight, Florence Henderson}, {Belinda Carlisle, Martina Navratilova}

40's Females: {Marie Osmond, Ricki Lake}, {Lisa Rinna, Melissa Gilbert, Niecy Nash, Pamela Anderson, Marlee Matlin, Toni Braxton}, {Vivica A. Fox, Chynna Phillips, Debi Mazar, Giselle Fernandez, Leeza Gibbons, Margaret Cho, Sherri Shepherd, Kathy Ireland, Tatum O'Neal, Wendy Williams, Macy Gray, Paulina Porizkova}

It's subjective, so you can decide for yourself: Which group has the most "forceful" (yet likable) personalities, and whether that can explain why 50+ Females have the best average placement among these three groups.

----------------------------------------------------------------------------

Here's another idea I'll offer for consideration: Perhaps 40's Females are a difficult group for pros to "present" well, especially if they feel pressure to compete directly with younger stars. Consequently, they may attempt programs that are too difficult (e.g. Melissa Gilbert), too "young" (e.g. Sherri Shepherd), or too "forced" to make a comfortable impression, and that can hurt judges' scores and voter response. By contrast, 50+ Females generally face low expectations: from themselves, their pros, the fans, the judges. (Exception: Jennifer Grey) That could make it easier for 50+ Females to enjoy themselves, to create a positive impression, and to sway judges and fans by "surpassing expectations".

Presumably, 40's Males also feel pressured to compete with younger stars, but this may be where female pros have the advantage in choreographing around their partner's limitations (according to some of the male pros, anyway). Female pros choreographing around their stars' abilities would be consistent with results reflecting males' physical skills: best for <40's, average for 40's, and decling for 50+.

Maybe DWTS needs pros who can help 40's Females enjoy themselves and improve their results (at least to statistical average). To date, Louis arguably has done the best job with multiple 40's stars: Lisa Rinna and Niecy Nash achieved middle placements, with only Margaret Cho leaving early. OTOH, all four of Tony's 40's stars finished in the bottom 9-12 places, including Chynna Phillips who many thought would go further. Neither Louis nor Tony seems likely to continue on the show much longer, so 40's Females may be a niche opportunity for newer pros and troupe members. At least, that's my optimistic hope. :)

Comments are welcome. Later this week, I'll post a simple algorithm for partnering Pros and All-Stars. (i.e. How might a blind computer do it?)

Edited by katdlf, Aug 8, 2012 @ 1:23 AM.

[Elsa]

Oooh I didn't realize how rude I sounded when I said the numbers weren't "reliable." What I meant, and ONLY meant, was that the numbers the judges award (or the "placement" of each dancer within the group of an evening's dancers) are NOT remotely objective, imo. I think they sit around a table, watching the dress rehearsal, having previously watched the camera blocking AND some of the rehearsal videos -- and they decide on scores and comments designed to "lead" the voting part of the viewership.

They want it to relate *somewhat* to the quality of the dance, to keep most people in the dark (which is why Derek's flashy choreography works so well for them as it confuses the viewers). But not always. If they have a motive for down-scoring someone, like J.R.'s Pink Panther Jive just to take one of hundreds of examples, they'll look for an excuse (like it's not supposed to be funny ??? or it's not supposed to have any "Lindy" in it - huh? Ricki's didn't even have proper jive kicks in hers and that's exactly what they bragged on!) or a "bad guy" judge (Len might claim lack of heel leads when he often ignores the heels; Carrie Ann might pull out her 'lift' card when other times she doesn't notice; Bruno might inexplicably 'love' somebody's emotions even if their dancing is horrid.)

I don't think these judges are that flaky. What they are is coached, or even more precisely, they're "scripted." I think that's what's on those notes in front of them.

Now does that mean they don't take the home viewer vote in account? Of course not. It's up there on the storyboard. But as QueenL pointed out so eloquently (and I tend to want to discount), they also know that the vast majority of viewers are neither voters nor do they go online to read about the show, and those viewers must stay on board or the ratings will fall and the show will be canceled.

So they have a huge incentive to be as fake as they wanna be to get the results they think will keep the show going. I don't like it (unless it favors my obsessees - and I'm really livid when I think they're making the wrong judgment about what's best for the longevity of the show AND hurting my obsessees) but, well, I'm an obsesser and not a "typical viewer" nor am I an expert in what this "general" audience wants to see, even if I sometimes think I know better than ABC. I don't know the voter percentage of viewers - and I don't know the fangirl percentage (me - well I'm in both categories) -- but I betcha they do!

But there's a LOT to be gleaned from your analysis, for many reasons. One of many reasons is they DO pay attention to the votecount because they've promised they'll abide by it. That's why they go to such great lengths to manipulate it -- with Len's "it's time for you to go home, Cinderella" type comments - and with 15 point bonus cha cha's - and marathon dances that are scored ahead of time and have absolutely nothing to do with the dancing in front of them - remember William Levy jumping around like a kid after school in the gym who's bored and decides to just slide on his knees and play? And he was one of the last standing?? Yeah right. Somebody should have told him to at least make it look vaguely "honest" - but then, I'm glad I see it for what it is because I hate being duped.

But most people don't take this seriously and don't give a hoot - and if you explained to them how you know something is fake - they look at you like you're an idiot. Well, ok, but it takes all kinds I guess. And I'm 99.999999% I'm right and this thing is MASSIVELY micro-managed and fake. From what little I've seen of Project Runway, SYTYCD and a little of AI, I think they're micro-managed as well.

Would they really need very many writers on staff to write Tom Bergeron's and Brooke Burke's parts? Tom ad-libs and anybody on this board could write Brooke's questions.

Another multi-thousand word essay would be needed to address what they think the general viewer needs to see to keep watching and keep this show on the air (they know the obsessers are hooked so they don't matter as much, though they can get upset and flee, like the "base" of a political party when their desires are totally ignored.)

ETA And I am so totally off your point and will try to address a couple aspects, anyway. About women in their 40's, one theory: they're young and being compared to women in their 20's and 30's, so they don't come out looking as good - and women voters might even be jealous or resent women their age looking so good or thinking they can look sexy like younger women can -- but 50+ women are given a break and not compared as critically -- with the guys, women are just turned on by certain things, maybe it relates to age groups or "cuddliness" or good-sportsmanship (Apolo on both; Helio/Drew on cuddly and somewhat on niceness too); overcoming a difficult life (J.R., Donald - both had the niceness thing) - well, it would take pages to analyze and I suck at it anyway. But ABC knows a lot about this sort of thing, I bet.

Edited by Elsa, Aug 8, 2012 @ 2:14 AM.

[katdlf]

Quick response to Elsa: I agree that 40's females as a group may be judged more critially on multiple fronts, unfortunately. And it's OK to believe that the show is scripted or rigged--whether it is or isn't--but that doesn't invalidate numbers analysis. (Although it probably limits your interpretation.)
-----------------------------------------

Getting back to numbers: There's a lot of speculation about All-Star pairings based on height, age, personality, etc. We know personality is an important though subjective consideration, but what if we let a computer choose the pairings objectively based on height and age? We can devise a simple algorithm to see how that might turn out.

One numeric approach would be to minimize age differences and simultaneously to optimize height differences. For this algorithm, the most "typical" (i.e. median or middle) height difference among the possible pairings is considered the "optimal" height difference. The algorithm is blind to everything else, including personality, celebrity requests, prior partnerships, etc.

Step-by-step, here's how a computer program or spreadsheet might calculate the "logical" lineup based on age and height.

-----------------------------------------
*Algortithm for All-Star Pairings*
-----------------------------------------

Step 1: Calculate the age difference in years (star age minus pro age):

Age Diff..................Mark.....Val......Derek......Tris.......Maks.....Tony
________________26.3____26.5____27.4____30.2____32.7____39.2_
Shawn....20.7...|...-5.7......-5.8......-6.7.......-9.5.....-12.0......-18.5
Bristol.....21.9...|...-4.4......-4.6.....-5.4........-8.2.....-10.8......-17.3
Melissa...29.5...|.....3.2........3.0.......2.2........-0.6......-3.1......-9.6
Kelly.......36.3...|...10.0.......9.8.......9.0.........6.2.......3.7.......-2.9
Pamela....45.2..|....18.9......18.7......17.9.......15.1........12.........6.0
Kirstie.....61.7...|...35.4.....35.2......34.3.......31.5......29.0.......22.5

Age Diff..............Chelsie.....Peta......Cheryl...Anna......Karina.....Kym
________________23.2____26.2____28.4____31.7____34.7____36.1_
Apolo.....30.3...|.....7.2........4.1.......1.9.......-1.4.......-4.4......-5.8
Joey......35.7...|....12.5.......9.5........7.3........3.9........0.9......-0.5
Drew......36.1...|....13.0.......9.9.......7.7........4.4........1.4........0.0
Gilles.....36.7...|....13.5......10.5........8.3........4.9........1.9.......0.5
Hélio......37.4...|....14.2......11.2.......9.0........5.6........2.7.......1.2
Emmitt....43.4..|....20.2......17.2......15.0......11.6........8.6........7.2


Step 2: Calculate the height differences in inches (male height minus female height):


Height Diff............Mark........Val........Derek.......Tris........Maks......Tony
________________68______69______70______71______74______73_
Shawn......58..|.....10.........11..........12.........13..........16.........15
Bristol.......65..|......3...........4...........5...........6............9..........8
Melissa......64..|......4..........5............6...........7..........10..........9
Kelly.........63..|......5..........6............7...........8..........11.........10
Pamela......67..|......1..........2............3...........4...........7...........6
Kirstie.......67..|......1..........2............3...........4...........7...........6

Height Diff...........Chelsie......Peta......Cheryl.....Anna......Karina......Kym
_________________62______67______64______64______65______67_
Apolo......68..|.........6...........1...........4...........4...........3..........1
Joey.......72..|........10...........5...........8...........8...........7..........5
Drew......66..|..........4.........-1...........2...........2...........1.........-1
Gilles.......73..|........11..........6...........9...........9...........8..........6
Hélio.......68..|.........6...........1...........4...........4...........3..........1
Emmitt....69..|.........7...........2...........5...........5...........4..........2

Step 3: Find the median (middle) height differences. Here, they are 4 inches between male stars & female pros, and 6 inches

between female stars & male pros. This serves as the "optimal" height difference for each set.

Step 4: Subtract the median height differences from the actual height differences. The adjusted height differences indicate

how close each pairing is to the optimal height difference.

Step 5: Adjust the age differences to be of comparable "size" as the height differences. (i.e. "Normalize" them.) To do this, we can find divisors that make the median age difference equal the median height difference. (Sounds difficult, but easy for a computer or spreadsheet. The divisors were 3.8 for female stars and 3.6 for male stars.)

Step 6: Now that we have comparable measures of age difference and optimal height difference, we sum their absolute values to get a combined measure of age and optimal height differences.

ETA:
Post is too long. Will add the rest later.

Edited by katdlf, Aug 12, 2012 @ 6:30 PM.

[QueenL]

Here's a seperator post. Enjoying this, though I don't have much to say at the moment.

I appreciate all of the hard work.

4" between the male stars and female pros isn't much, especially when you add in the heel factor. It is amazing to see so many short male celebs. And I do think that Apolo and Helio are both shorter than their listed heights.

[katdlf]

Thanks QueenL...

4" between the male stars and female pros isn't much, especially when you add in the heel factor.

Someday I'll post the details, but historically male stars with smaller height differences tend to place better than male stars with greater height differences. I suspect it's easier for female pros to back-lead their stars with smaller height differences, but that's just a guess. Back to the algorithm...


The results of steps 3-6 are:

____________Mark____Val____Derek___Tris____Maks____Tony_
Shawn...........5.5........6.5........7.8.......9.5.......13.2.......13.9
Bristol............4.2........3.2........2.4.......2.2........5.8.........6.5
Melissa..........2.8........1.8.........0.6.......1.2........4.8.........5.5
Kelly..............3.6........2.6........3.4.......3.6........6.0.........4.8
Pamela.........10.0........8.9........7.7.......6.0........4.3.........1.6
Kirstie..........14.3.......13.3......12.0......10.3........8.6.........5.9

___________Chelsie___Peta___Cheryl___Anna___Karina___Kym_
Apolo.............4.0.........4.2........0.5.......0.4........2.2........4.6
Joey...............9.5........3.6........6.0.......5.1........3.3........1.1
Drew..............3.6........7.8........4.1........3.2.......3.4.........5.0
Gilles............10.7.........4.9........7.3.......6.4........4.5........2.1
Hélio..............5.9.........6.1........2.5.......1.6........1.7........3.3
Emmitt...........8.6.........6.8........5.2.......4.2........2.4........4.0

Step 7: Find the pairings that minimize overall age differences while optimizing overall height differences for the field.

For each star or pro in the above table, their closest physical "match" is indicated by the minimum difference value in the corresponding row or column. (Their furthest match is indicated by the maximum difference value.) Interesting that either Melissa or Pam is every male pro's best match.

To determine the lineup, we find the set of star-pro pairings with the minumum overall sum of combined age-height differences.

The results turn out to be:

Shawn-Mark
Bristol-Tristan
Melissa-Derek
Kelly-Val
Pamela-Maks
Kirstie-Tony

Drew-Chelsie
Gilles-Kym
Joey-Peta
Apolo-Cheryl
Helio-Anna
Emmitt-Karina

It happens to work out well for all new pairings. For an optimized lineup of entirely new pairings, we need only swap the Shawn and Kelly pairings:

Shawn-Val
Kelly-Mark

So that's an example of a computer-generated pairings based solely on the physical factors of age and height. In actuality, we know that's not how pairings are chosen, and that human factors like personality play a big role, as they should. We'll find out tomorrow the producers' view of "logical" where all factors are considered.

Geeky footnote: Based on historical rank analysis, age or height "logically" match about 30% of the time, age *and* height about 15% of the time, but that's a post for another time.

Edited by katdlf, Aug 12, 2012 @ 7:29 PM.

[QueenL]

Interesting, the male pro-female celeb pairings are close to the ones I figured TPTB would put together.

It does make sense that they are only predictors 15% of the time, though. There's a lot more to consider than just that for good pairings.

Once again, enjoyed reading this!

[katdlf]

With old and new teams in the mix this season, managing expectations will be a bigger challenge than usual, imho. The old teams have their own standards to live up to or to surpass. The new teams begin with a cleaner slate, but may face expectations that are too high, too low, or misplaced. For consideration, here are comparisons of new teams with past teams, based on heights and ages. Some couples have more and better comparison points, but these comparisons might help calibrate expectations for skills and relationships a bit, bearing in mind that personalities are always the immeasurable X factor. :D

Star....................Age.....Height.....Pro.....Age.....Height....AgeDf....HtDf

Shawn Johnson.......20.7.....58.....Derek.....27.4......70.......-7.......-12
----------------
Shawn Johnson........17.......58.....Mark.......22.......68.......-5.......-10
Lil' Kim...................34.......59.....Derek......23.......70.......11.......-11
Marissa Jaret Winokur..35.....60.....Tony......34.......73........1........-13

Kelly Monaco........36.3.......63.......Val.......26.5.......69.......10.......-6
----------------
Melissa Joan Hart.....33.......62.......Mark.......23.......68.......10.......-6
Denise Richards.......38.......66.......Maksim.....29.......74.......9........-8
Tia Carrere............39.......68.......Maksim.....25.......74.......14.......-6
Kristi Yamaguchi......36.......61.......Mark.......21.......68.......15.......-7
Jennie Garth...........35.......65.......Derek......22.......70.......13.......-5

Pamela Anderson.....45.2.....67.....Tristan.....30.2.....71.......15.......-4
----------------
Debi Mazar............45.......65.......Maksim.....29.......74.......16.......-9
Marie Osmond........47.......65.....Jonathan.....33.......72.......14.......-7
Leeza Gibbons.......49.......68........Tony........33.......73.......16.......-5
Melissa Gilbert.......47.......63.......Maksim......32.......74.......15......-11
Pamela Anderson...42.......67........Damian......32.......68.......10.......-1
Ricki Lake.............42.......64.......Derek........26.......70.......16.......-6

Star....................Age.....Height.....Pro.....Age.....Height....AgeDf....HtDf

Apolo Anton Ohno...30.3.....68.......Karina.....34.7......65.......-4.......3
----------------
J.R. Martinez..........28.......70.......Karina.......33.......65.......-5.......5
Mike Sorrentino.......28.......70.......Karina.......32.......65.......-4.......5

Drew Lachey...........36.1.....66.......Anna.......31.7.....64.......4.......2
----------------
Drew Lachey..........29.......66.......Cheryl.......21.......64.......8.......2
Joey Lawrence........30.......68.......Edyta.......25.......66.......5.......2
Warren Sapp..........35.......74.......Kym..........32.......67.......3.......7

Gilles Marini...........36.7......73.......Peta.......26.2.....67.......10.......6
----------------
Gilles Marini.............33.......73.......Cheryl.......24.......64.......9.......9
Cameron Mathison....38.......73.......Edyta.......26.......66.......12.......7
Cristián delaFuente..34.......74.......Cheryl.......23.......64.......11.......10
Donald Driver..........37.......72.......Peta.........25.......67.......12.......5

Helio Castroneves....37.4.....68.......Chelsie.....23.2.....62.......14.......6
----------------
Emmitt Smith............37.......69.......Cheryl......22.......64.......15.......5
Hélio Castroneves.....32.......68.......Julianne....19.......64.......13.......4
Ty Murray...............39.......68.......Chelsie.....19.......62.......20.......6
Jake Pavelka............32.......68.......Chelsie.....20.......62.......12.......6

Thanks for asking QueenL, but the post is complete.

Edited by katdlf, Aug 17, 2012 @ 2:23 PM.

[QueenL]

Do you need a seperator post?

And yeah, I never understood why people were making a big deal of the height differential between Shawn and Derek. Derek's already worked with someone around Shawn's height (Lil' Kim) and it worked out well.

[katdlf]

Historical averages may have little relevance this All-Star season, but maybe there's still interest in checking them out? Here's a breakdown of major factors, starting with age and height. (Raw data from previously linked Excel sheet.)

Age Breakdown

The age average is the average normalized final placement (N12) of all males or females within a -2 to +2 age range. e.g. For Shawn, age 20, it's the average of all females 18-22. For Joey, it's the average of all males 33-37. Based on data through 14 seasons, the Season 15 age-advantaged, age-neutral, and age-disadvantaged stars can be grouped (relatively and somewhat arbitrarily) as:

Star............................Age...........Age Avg

*Age-Advantaged*
Shawn Johnson..........20...............3.1
Bristol Palin...............21...............4.4
Joey Fatone...............35...............4.9

*Age-Neutral*
Kirstie Alley................61..............5.2
Hélio Castroneves.....37...............5.3
Apolo Anton Ohno.....30..............5.3
Drew Lachey...............36..............5.6
Gilles Marini...............36..............5.6

*Age-Disadvantaged*
Melissa Rycroft..........29..............6.1
Sabrina Bryan............28...............6.1
Kelly Monaco.............36..............6.5
Emmitt Smith.............43..............7.0
Pamela Anderson......45..............8.4

Note: Nearly all the stars placed better than their current age averages in their previous seasons.

Height Breakdown

In addition to a) star height, we can account for pro height by calculating b) male minus female height difference, and c) female-to-male height %. (These measures are related in that knowing two, we can calculate the third.) Seasons 1-14 normalized placements are averaged across -1 to +1 inch height, height difference, and height % for each star wrt to their gender. Rather than choose a "best" average, an overall average is calculated for simplicity. With arbitrary groupings, the results are:

Star......................Ht (in)........Ht Avg......Ht Df......Df Avg.......Ht %.......% Avg......Overall Avg

*Height-Advantaged*
Shawn Johnson..........59...........3.2.........11...........3.5...........84.3.........5.1..........3.9
Drew Lachey.............66............4.0.........2.............4.6...........97.0........3.1...........3.9
Apolo Anton Ohno....68...........4.6.........3.............4.3...........95.6.........6.0..........5.0

*Height-Neutral*
Emmitt Smith............69..........5.7...........5..............5.1..........92.8.........5.0..........5.3
Joey Fatone...............72..........6.4..........5..............5.1...........93.1........5.0...........5.5
Hélio Castroneves....68..........4.6...........6.............6.2...........91.2........6.0...........5.6
Gilles Marini..............73..........5.7..........6..............6.2...........91.8........5.4...........5.8
Melissa Rycroft.........64..........5.7...........9.............5.9...........87.7........5.9...........5.8

*Height-Disadvantaged*
Sabrina Bryan............64.........5.7...........3..............6.6...........95.5........6.0...........6.1
Bristol Palin...............65..........5.9..........3..............6.6...........95.6........6.0...........6.2
Kelly Monaco............61..........6.4..........9..............5.9...........87.1........6.4...........6.2
Pamela Anderson.....67..........7.3...........4..............6.8..........94.4.........5.8..........6.6
Kirstie Alley..............67..........7.3...........7..............7.0..........90.5.........5.9..........6.7

You'll notice a divergence in male and female height patterns: Smaller height differences / bigger height ratios are better for male stars, but worse for female stars. This could mean, in general, that smaller males and females move better than taller males and females. Or it could mean that it's easier for male pros to manouver smaller females, and easier for female pros to backlead smaller male stars. (Or both, or something else?) There have been many individual exceptions, but it's an interesting general result.

More later.

Edited by katdlf, Sep 20, 2012 @ 8:59 PM.

[Elsa]

Kelly's 5'1". Someone recently tweeted her the question and she admitted she was 5'1" (lots of online sources say 5'3" but she's not that tall.) I don't think Apolo is quite 5'8" either, probably 5'7". I'm also doubtful that Pamela is 5'7" - looking at pictures of her in heels standing next to 5'11" Tristan, she looks shorter than 5'7" to me.

[katdlf]

Thanks for the update on Kelly. I also saw where Shawn told a reporter that she's 4'11" now. Actually, she claimed 5', but the reporter called her on it, lol. Derek has also said that she's similar to Lil'Kim, who is a reported 4'11". I've updated heights for Kelly and Shawn in the last post. There are quite a few suspect star (and pro) reported heights, but other than Shawn, they probably are the same actual heights as in previous seasons. The historical averages are based on reported height, so for now, other heights remain unchanged.

Continuing with historical averages:

Job Breakdown

"Job" is somewhat a proxy for fanbase and physical skill. There's a problem in that classifications greatly affect outcomes. In setting up this analysis, I used the 7 categories (Actor, Athlete, Comedian, Model, Reality, Singer, Other) that ABC recognizes in their online "casting" game, plus a few warranted by historical numbers. Like separating Basketball, Boxing, and Tennis (BBT) athletes from other athletes. Also, Entertainer for performers with dancing, choreography, or "show" experience; and Host for celebrities with an interactive TV audience. Averages are gender-based, so in effect there are 20 categories. The All-Stars can be grouped, relatively and somewhat arbitrarily, as:

Star...............................Job................Job Avg

*Job-Advantaged*
Sabrina Bryan............Entertainer........3.1
Apolo Anton Ohno.......Athlete............3.3
Emmitt Smith..............Athlete............3.3
Hélio Castroneves........Athlete............3.3

*Job-Neutral*
Drew Lachey..............Entertainer........4.0
Joey Fatone...............Entertainer........4.0
Shawn Johnson...........Athlete.............4.4
Gilles Marini................Actor...............5.7

*Job-Disadvantaged*
Kelly Monaco..............Actor...............6.3
Kirstie Alley................Actor...............6.3
Pamela Anderson........Actor................6.3
Melissa Rycroft...........Reality..............8.1
Bristol Palin................Other...............8.1

Pro Breakdown

Getting right to it, relatively and somewhat arbitrarily:

Star............................Pro............Pro Avg

*Pro-Advantaged*
Shawn Johnson...........Derek........3.1
Emmitt Smith..............Cheryl.......4.5
Bristol Palin................Mark..........5.6
Kirstie Alley................Maks.........5.6
Joey Fatone...............Kym..........5.9
Apolo Anton Ohno.......Karina........6.0

*Pro-Neutral*
Gilles Marini...............Peta...........6.5
Pamela Anderson.......Tristan........6.5
Hélio Castroneves......Chelsie........6.6
Sabrina Bryan............Louis..........6.7

*Pro-Disadvantaged*
Melissa Rycroft..........Tony..........7.4
Drew Lachey.............Anna..........7.5
Kelly Monaco.............Val...........10.5

The more stars a pro has partnered, the more likely their normalized average will converge to 6.5. Pros with averages a half-place above this midpoint can be considered relatively advantageous, imho, if they have partnered more than a handful of stars. A cute example of regressing to the mean: Both Peta and Tristan are at the expected 6.5 average after two partners. They did it in very different ways, with Tristan hitting middle-placements of 5 and 8, and Peta hitting the extremes of last and first. With just two partners, this is very changeable (for Val also), but it's still cute. :-)

Hope you're enjoying this if you're reading it. Summary post to follow.

[katdlf]

Thanks, anothermi. Regarding analysis details: Part of the reason for creating a separate numbers thread was to allow for more technical discussion from time to time. I'd be happy to answer your questions to the best of my ability, but perhaps we could put some kind of *Technical Stuff* header in those posts so that they can be more easily skipped or sought. :)

Finishing up the preseason numbers:
Here are the overall averages of historical performance (again, relatively and somewhat arbitrarily grouped):

Star.....................Age Avg....Ht Avg*....Job Avg....Pro Avg....Overall

*Overall Advantaged*
Shawn Johnson...........3.1........3.9..........4.4...........3.1..........3.6
Apolo Anton Ohno.......5.3.......5.0...........3.3...........6.0..........4.9
Emmitt Smith..............7.0.......5.3...........3.3..........4.5...........5.0
Joey Fatone...............4.9........5.5..........4.0...........5.9...........5.1
Hélio Castroneves.......5.3........5.6...........3.3..........6.6...........5.2
Drew Lachey..............5.6........3.9...........4.0..........7.5...........5.3

*Overall Neutral*
Sabrina Bryan.............6.1.......6.1...........3.1...........6.7..........5.5
Gilles Marini................5.6........5.8..........5.7...........6.5..........5.9
Kirstie Alley................5.2.......6.7...........6.3...........5.6..........6.0
Bristol Palin................4.4........6.2..........8.1...........5.6...........6.1

*Overall Disadvantaged*
Melissa Rycroft..........6.1........5.8..........8.1...........7.4...........6.8
Pamela Anderson........8.4........6.6...........6.3..........6.5...........6.9
Kelly Monaco.............6.5........6.2..........6.3..........10.5..........7.4

A few random notes:
-Sabrina's on the cusp of advantaged, but she is only advantaged in one area (job) whereas the stars ahead of her are advantaged in multiple areas.
-Just an observation: Sabrina's the only All Star to not meet or beat their current historical average in their previous season.
-Anna has her first ever height-advantaged partner in Drew. It could be interesting to see it there is an attributable difference in choreography or backleading.
-Louis becomes the third 40-something pro, joining Nick and Corky.

And a few comments:
I'm not one for bold predictions. In a regular season, I'd very broadly guess:

-The advantaged group likely won't place in the bottom 4.
-The disadvantaged group likely won't place in the top 4.
-The neutral group could place anywhere, but probably not first or last.
-It's more likely that an advantaged star falls to the bottom 4 than that a disadvantaged star rises to the top 4. (That's a hedge, based on 6 advantaged stars vs 3 disadvantaged stars.)

And that's the extent of my comfort zone for guessing. In a regular season. Which this isn't. There are so many new factors for All Stars, e.g. winners vs non-winners, finalists vs non-finalists, new couples vs old couples, etc. that could be viewed or scored on different scales. And there's greater knowldege of the individual stars as dancers and personalities. I don't think anyone would be surprised should Gilles win, for example, or should one of the 3 male athletes or 2 boy band members be hurt early by vote splitting.

One thing that hasn't changed: We'll all feel a little bolder about making predictions after seeing the dancing and scoring on opening night. Good luck to all your personal favorites.

[ImSoDizzy]

katdlf - Thanks for starting this topic and thanks for the Straw Man. I like to try to put number-explanations behind phenomena as well, and DWTS is an interesting playground for this type of analysis. I downloaded your worksheet, and congratulate you on the thought and work you've put into your predictions.

I take your point that Season 15 will be very different than past seasons, but that just makes interpolation all the more fun. Some attribution-based advantages should hold true regardless though, IMO.

I agree the variables you’ve chosen: Age, Height, Job (M/F), and Pro Experience are good ones. I have a few more bouncing around my head that I might add. I see you already looked at the Star’s gender and Pro’s DWTS tenure. (I also thought that a Pro’s DWTS experience would be important, but looking at the record it seems to be a fairly short learning curve for some.)

I do think there is a link as to how the Pro placed the previous season. There are several instances where a pro has won back-to-back. But a 2nd place finish seems to be a jinx spot for the following season, and 3rd place finish is not a great precursor either. Whether this is caused by a seasonal carry-over un/favorables, the Pro’s motivation, or ?? - I don’t know. I do think this will lower Bristol and Emmitt’s chances somewhat this season.

I also wish there was a way to predict the elusive ‘chemistry’, where I believe ‘romance’ (real or believable) plays some part. I see you looked at a Pro’s ‘type’ of Star – was this your attempt at capturing this? I don’t see where you reached any conclusions however.

While I haven’t run the numbers (read: I’m basically talking out of my hat here ;), I don’t think married pros do as well as single ones - although this is already picked up in the Pro-Advantage you’ve calculated. I also think that pairings of: a younger-male (< 30) pro/ older-female Star, a younger-female (< 25) pro/ older-male Star, or a female pro/ comparable-aged Star; do better. Older-female pros/younger male Stars, older-male pros/ older female Stars, and younger male pros/ comparable-aged Stars are not as successful.

My ‘couple chemistry’ theory would somewhat lower the chances of: Shawn, Apolo, Kirsti, Bristol, and maybe Pam (Tristan turned 30 in July :). It could also raise the chances of Kelly, although given Val's very-low Pro Advantage, this will likely be wiped out by other factors.

Like you, I’m looking forward to this season. It will be especially interesting if early cast-offs (Sabrina & Pam) are repeated - and we will know this fairly soon.

[katdlf]

katdlf - Thanks for starting this topic and thanks for the Straw Man. I like to try to put number-explanations behind phenomena as well, and DWTS is an interesting playground for this type of analysis. I downloaded your worksheet, and congratulate you on the thought and work you've put into your predictions.

ImSoDizzy - Thank you. Sorry for this season-later response. It's difficult to maintain stats during any season, and All-Stars was a very draining season.

I agree the variables you’ve chosen: Age, Height, Job (M/F), and Pro Experience are good ones.

They're similar to the ones used by PureDWTS, plus height. The main difference is separation by gender.

I have a few more bouncing around my head that I might add. I see you already looked at the Star’s gender and Pro’s DWTS tenure. (I also thought that a Pro’s DWTS experience would be important, but looking at the record it seems to be a fairly short learning curve for some.)

I need to credit someone (QueenL?) for looking into the pros' Finals records. Reaching a first Finals has historically been quick, typically within 2-3 seasons. Reaching a 2nd Finals can take longer.

I do think there is a link as to how the Pro placed the previous season. There are several instances where a pro has won back-to-back. But a 2nd place finish seems to be a jinx spot for the following season, and 3rd place finish is not a great precursor either. Whether this is caused by a seasonal carry-over un/favorables, the Pro’s motivation, or ?? - I don’t know.

Very interesting, and worth looking into at some point. :-)

I also wish there was a way to predict the elusive ‘chemistry’, where I believe ‘romance’ (real or believable) plays some part. I see you looked at a Pro’s ‘type’ of Star – was this your attempt at capturing this? I don’t see where you reached any conclusions however.

Not really. That was an attempt to assess whether certain pros tend to get "better" stars, a question of ongoing interest to DWTS fans. :-) I'm debating whether All-Stars season should count in the overall analysis. The whole season was an anomoly, but every season has had atypical aspects. Because of that, I'm inclined to include All-Stars. The averages already were based on a hodgepodge of Reality TV semi-randomness.

The closest consideration of "chemistry" was a pre-season comparison of the All-Star couples' heights and ages to previous couples' heights and ages. For me, the biggest "aha" was realizing that Chelsie-Helio might not be the awkward pairing many predicted, based on previous examples like Cheryl-Emmitt. IMHO, Chelsie has had good "chemistry" with athletes like Ty and Louie, and now Helio. Maybe growing up with older brothers means she's at ease in their "jock" world. Or maybe athletes respect her more as a teacher/coach even if they're older. Whatever the reason, her example suggests "job" as somewhat a predictor of chemistry.

It's hard to assess "chemistry". We could come up with a formula that includes age, height, and job as measures of "chemistry", but still miss the subjective nature of chemistry. We could have viewers rate the "chemistry" of previous partnerships, look for correlates, and still miss interactions. Some pros might do well chemistry-wise with a particular age group, while others do well with a particular job group, and still others with a specific interaction of the two (e.g. good with younger actors and with older athletes). And we might still miss capturing the crucial, immeasurable "chemistry" factor entirely.

In an objective way, I suppose the chemistry question is captured indirectly within the analyses in the spreadsheet. It's just a question of how one chooses to interpret the data. (Sorry for the long-winded answer.)

While I haven’t run the numbers (read: I’m basically talking out of my hat here ;), I don’t think married pros do as well as single ones - although this is already picked up in the Pro-Advantage you’ve calculated. I also think that pairings of: a younger-male (< 30) pro/ older-female Star, a younger-female (< 25) pro/ older-male Star, or a female pro/ comparable-aged Star; do better. Older-female pros/younger male Stars, older-male pros/ older female Stars, and younger male pros/ comparable-aged Stars are not as successful.

This is heavily conflated with age. Placements drop off significantly with Star age, especially after age 40, so older celebrity / younger pro teams will struggle regardless of gender. OTOH, given the general youth of the pros, the advantages you're describing could be the result of typical Star-Pro pairings for stars in their 20's and 30's. These age-advantaged stars tend to get Pros of similar age, because the Pros also are in their 20's and 30's.

-----------------------------

We should look into marriage. I've seen discussions about why married pros may be less successful than unmarried pros, a frequent hypothesis being a lack of sexual tension, but I don't think I've seen it quantified in any way. And we do now have one married Pro in the mix. I've added an interactive Groups page to the spreadsheet. The pre-defined groups are Male and Female, but this can be changed easily in the Group column. I changed the groups to S for single pros and M for married pros (Anna, Jonathan, Tony, Edyta & Alec starting in Season 5, Ashly in Season 10, Nick, Corky, and Fabian) and came up with this result:

Single Pros average (normalized) placement: 6.0
Married Pros average (normalized) placement: 7.9

So yes, there appears to be a difference with married pros finishing, on average, two places lower than single pros. That includes the All-Star results. More could be done here, like comparing married female pros, married male pros, and single male and female pros. Changing the group codes to MF, SF, MM, SM yields:

SF........6.0
MF.......8.0
SM.......6.0
MM......7.8

Pretty much the same as singles versus marrieds, so which gender doesn't appear to matter as much as "married" here. Marriage historically makes a difference for pros.

My ‘couple chemistry’ theory would somewhat lower the chances of: Shawn, Apolo, Kirsti, Bristol, and maybe Pam (Tristan turned 30 in July :). It could also raise the chances of Kelly, although given Val's very-low Pro Advantage, this will likely be wiped out by other factors.

Val's average was based on just 2 partners prior to Season 15, so it was less reliable and bound to move closer to the 6.5 expected average with additional partners. He and Kelly had great chemistry, but so did Pam and Tristan off the dance floor, imho. On the dance floor? I think we saw some of my previously stated hypothesis on the struggles of 40-something women play out, namely that they're difficult to package. In general, I feel like 40's women are expected to compete directly with younger females in dance-ability and sexuality, but then are harshly criticized for the awkward results when they actually can't compete directly with younger females. I also think the awkwardness is increased because of mid-life transitions that make 40's women particularly vulnerable or sensitive to criticism. It's not until they turn late-40's or 50's that they start catching a break in public perception, imho. But that's just a hypothesis. :-)

I'm glad you like numbers, because I really like numbers. (Yay!) If you or anyone come up with further ideas for measuring or predicting chemistry, please share them.

---------------------------------------------------------------------

In the meantime, I've updated the Excel Spreadsheet to include Season 15. I've also simplified the worksheets. Please check it out if you're interested in numbers. The new Groups sheet might be helpful in running your own analyses:

DWTS spreadsheet

Also, I have a blog entry based on data from the spreadsheet, HOW OLD IS OLD FOR DANCING WITH THE STARS? This one's worth a bit of discussion, hopefully.

Edited by katdlf, Feb 6, 2013 @ 10:11 PM.

[Serendi]

Corky's married? I know he and Shirley are divorced, but I realize I don't know how long ago, nor whether he's remarried. Tried some Googling and didn't get very far.

[katdlf]

Corky's married?

I think he was still married when he competed on DWTS, no?? I'll redo those numbers if he wasn't.

I've just posted another article called Forecasting Star Age for Dancing's Newest Pros that should be of particular interest to Peta, Val, and Tristan fans. (It was intended to be posted last week, but blogger ate my initial draft. Errghh...)

Thank you to everyone who checked out the previous post (How Old is "Old"). I'm curious if there's any reaction to it. In particular, is age disparity in DWTS results something worth addressing? Or is it an unavoidable reality of the show? If it's worth addressing, then is it better to make changes that allow young and old stars to still compete together? Or is it better to separate the competition by age group? And lastly, was there anything in the numbers that really surprised you?

[katdlf]

For those who like numbers, I have a new blog entry that compares the statistical advantages and disadvantages of the Season 16 stars:

This is not a prediction, but...

Comments are welcomed.

ETA a summary of the final clustered groups of statistically most advantaged to statistically most disadvantaged stars. Details and some personal comments are on the blog.

STAR..................AGE Avg......HT Avg....JOB Avg...PRO Avg...OVERALL Weighted Avg**

Aly Raisman............2.9..........6.2...........4.0..........5.9........... 4.5
Jacoby Jones.........4.3.......... 6.1.......... 3.7.......... 5.9.......... 4.7
Kellie Pickler.......... 6.7.......... 6.2.......... 5.9.......... 3.0.......... 5.4

Dorothy Hamill...... 6.8.......... 6.4.......... 4.0.......... 8.3.......... 6.1
Ingo Rademacher.. 6.7.......... 6.1.......... 5.7.......... 6.3.......... 6.1
Zendaya Coleman.. 2.9.......... 8.2.......... 6.2.......... 7.9.......... 6.1

Sean Lowe............ 4.9.......... 6.8.......... 7.2.......... 6.2.......... 6.3
Wynonna Judd....... 7.5.......... 5.9.......... 5.9.......... 6.9.......... 6.6
Victor Ortiz............ 3.0.......... 5.7.......... 9.1...........7.7.......... 6.8

D.L. Hughley.......... 6.9.......... 4.9.......... 9.5.......... 4.5.......... 7.0
Lisa Vanderpump..... 6.8.......... 6.6.......... 7.4.......... 7.7......... 7.2
Andy Dick.............. 7.4.......... 6.7.......... 9.5.......... 7.7.......... 8.2

ETA: Alternate bottom cluster. Details can be found in the blog.

Victor Ortiz............ 3.0.......... 5.7.......... 9.1...........9.3.......... 7.2
Lisa Vanderpump..... 6.8.......... 6.6.......... 7.4.......... 9.3.......... 7.6
Andy Dick.............. 7.4.......... 6.7.......... 9.5.......... 9.3.......... 8.6

Edited by katdlf, Mar 18, 2013 @ 10:16 AM.