An interesting read(why okcupid match system is flawed)

Page 1 of 1 [ 16 posts ] 

spongy
Veteran
Veteran

User avatar

Joined: 17 Jul 2010
Age: 33
Gender: Male
Posts: 8,055
Location: Patiently waiting for the seventh wave

04 Apr 2012, 4:58 am

http://isomorphismes.tumblr.com/post/13 ... -algorithm

This is an article from a mathematician and its based on mathematics, it explains in an excellent manner why okcupid match system is flawed.


TLDR version: there are plenty of questions that most people agree in(would you go out with someone that was already in a relationship?/ how would you react to cheating...).
Each of this questions has a value of 250 and several of this questions lead to a extremely high score. Take it that you answer 10 of this questions in the same way 25000/25000 if you get somewhat similar results in other questions you are going to be around the 80% or higher and if you happen to have opposite views you would still get a 50% match based on questions that according to the article shouldnt even be there because they cant be used to determine wether you´d be a good match or not.



nat4200
Veteran
Veteran

User avatar

Joined: 10 Jan 2011
Gender: Male
Posts: 704
Location: BANNED

04 Apr 2012, 5:37 am

Redacted



Last edited by nat4200 on 21 Apr 2012, 2:03 am, edited 1 time in total.

beers
Snowy Owl
Snowy Owl

User avatar

Joined: 29 Jan 2012
Age: 37
Gender: Male
Posts: 128

04 Apr 2012, 6:28 am

Any system to objectively assess a subjective behavior will be flawed to some degree.


_________________
Your Aspie score: 131 of 200
Your neurotypical (non-autistic) score: 62 of 200
You are very likely an Aspie


nat4200
Veteran
Veteran

User avatar

Joined: 10 Jan 2011
Gender: Male
Posts: 704
Location: BANNED

04 Apr 2012, 6:56 am

Redacted



Last edited by nat4200 on 21 Apr 2012, 2:03 am, edited 1 time in total.

minervx
Veteran
Veteran

User avatar

Joined: 13 Apr 2011
Age: 33
Gender: Male
Posts: 1,155
Location: United States

04 Apr 2012, 7:55 am

I agree with the OP, but this is not a huge problem. Generally, people on OkC don't take match percentages very seriously. I'm not going to see a 98% as better than a 95%, but I would look into things further if we had less than a 70% match.



Maerlyn138
Velociraptor
Velociraptor

User avatar

Joined: 2 Nov 2005
Age: 52
Gender: Male
Posts: 499
Location: The Island of Misfit Toys

04 Apr 2012, 8:12 am

Ya, some of those questions are real butt-nutters with no possible dating relevance. I really wonder what the people were thinking when they wrote them.


_________________
We are the music makers, and we are the dreamers of dreams.

Aspie score: 159 of 200 NT score: 64 of 200


rabbittss
Veteran
Veteran

User avatar

Joined: 29 Dec 2011
Age: 39
Gender: Male
Posts: 1,348

04 Apr 2012, 9:22 am

I think the biggest flaw is that you cannot filter messages by length. I think it would greatly improve the chances of people meeting if the women didn't have to sort through tons and tons of "Lets hook up" and the ilk.


The Questions do sometimes come across as very strange, and they aren't gender specific, so you get some really odd ones occasionally.



ValentineWiggin
Veteran
Veteran

User avatar

Joined: 15 May 2011
Age: 36
Gender: Female
Posts: 4,907
Location: Beneath my cat's paw

04 Apr 2012, 9:35 am

This is embarrassing...
could someone greatly-stupefy this article for me?

:oops:

I like OKC's questions because you can mark their importance.
What part of the algorithm is being analyzed here, without getting too far into the author's methodology?


_________________
"Such is the Frailty
of the human Heart, that very few Men, who have no Property, have any Judgment of their own.
They talk and vote as they are directed by Some Man of Property, who has attached their Minds
to his Interest."


TechnoDog
Veteran
Veteran

User avatar

Joined: 16 Feb 2012
Age: 40
Gender: Male
Posts: 869
Location: Thornaby, UK

04 Apr 2012, 11:04 am

Don't see what he is going on about, his system can just be spammed too. Any system can be spammed, once you know how it works & his hotness scale is ilogical, as that would be thinking that everyone uses it the same.

You got to analyse the person yourself. I just go on Location, gender, age & read the profile. If I am interested I will send a email. Then I will start to analyse.

Basic human instincts denotes, stay away from any negative answering. Why would they want to decrease they chances. Unless they honest.


_________________
INTJ, Type5 Observer, Ecologists,
?When you make a mistake, don't look back at it long. Take the reason of the thing into your mind and then look forward. Mistakes are lessons of wisdom. The past cannot be changed. The future is yet in your power.?


01001011
Veteran
Veteran

User avatar

Joined: 3 Mar 2010
Age: 43
Gender: Male
Posts: 991

05 Apr 2012, 11:51 am

I suspect this is by design so that everyone get some matches.



Girljinxed
Butterfly
Butterfly

User avatar

Joined: 3 Apr 2012
Age: 40
Gender: Female
Posts: 9
Location: Washington, DC

06 Apr 2012, 6:57 pm

I've been on the site for years. I've certainly found that the more questions I answer the better my matches. I also highlight (as a link) the important interests of mine. I've come to really trust OkCupid's suggestions for in the percentage rankings.



ToadOfSteel
Veteran
Veteran

User avatar

Joined: 23 Sep 2007
Age: 36
Gender: Male
Posts: 6,157
Location: New Jersey

07 Apr 2012, 8:43 am

ValentineWiggin wrote:
This is embarrassing...
could someone greatly-stupefy this article for me?

:oops:

I like OKC's questions because you can mark their importance.
What part of the algorithm is being analyzed here, without getting too far into the author's methodology?


He's analyzing how "Mandatory" questions are worth waaaay more in terms of weight when computing match %. He says that by answering some "easy" questions like that marked as mandatory, it makes the match % artificially high, even if all the lesser questions (which actually may be of great importance) are answered differently. You could be completely incompatible with someone, but because you both answered questions like "Is the Earth bigger than the Sun" the same way and have it marked "Mandatory", you could still have 80% match or higher.



rabbittss
Veteran
Veteran

User avatar

Joined: 29 Dec 2011
Age: 39
Gender: Male
Posts: 1,348

07 Apr 2012, 9:16 am

ToadOfSteel wrote:
ValentineWiggin wrote:
This is embarrassing...
could someone greatly-stupefy this article for me?

:oops:

I like OKC's questions because you can mark their importance.
What part of the algorithm is being analyzed here, without getting too far into the author's methodology?


He's analyzing how "Mandatory" questions are worth waaaay more in terms of weight when computing match %. He says that by answering some "easy" questions like that marked as mandatory, it makes the match % artificially high, even if all the lesser questions (which actually may be of great importance) are answered differently. You could be completely incompatible with someone, but because you both answered questions like "Is the Earth bigger than the Sun" the same way and have it marked "Mandatory", you could still have 80% match or higher.


The problem is that I've found, at least in my area, is how many people skip those sorts of questions.. I can only hypothesis it's directly proportional to the number of people who are also "Christianity and very serious about it"



noname_ever
Veteran
Veteran

User avatar

Joined: 25 Dec 2011
Age: 49
Gender: Male
Posts: 500
Location: Indiana

07 Apr 2012, 10:34 am

ToadOfSteel wrote:
ValentineWiggin wrote:
This is embarrassing...
could someone greatly-stupefy this article for me?

:oops:

I like OKC's questions because you can mark their importance.
What part of the algorithm is being analyzed here, without getting too far into the author's methodology?


He's analyzing how "Mandatory" questions are worth waaaay more in terms of weight when computing match %. He says that by answering some "easy" questions like that marked as mandatory, it makes the match % artificially high, even if all the lesser questions (which actually may be of great importance) are answered differently. You could be completely incompatible with someone, but because you both answered questions like "Is the Earth bigger than the Sun" the same way and have it marked "Mandatory", you could still have 80% match or higher.

I thought the big problem was that answering the mandatory questions correctly was too many points for "showing up". Answering mandatory questions incorrectly should be reflected very negatively in the matching process.



starkid
Veteran
Veteran

User avatar

Joined: 9 Feb 2012
Gender: Female
Posts: 5,812
Location: California Bay Area

07 Apr 2012, 6:23 pm

rabbittss wrote:
The problem is that I've found, at least in my area, is how many people skip those sorts of questions.. I can only hypothesis it's directly proportional to the number of people who are also "Christianity and very serious about it"


I skip those sorts of questions because they are too obvious to be important to matching. Anyone who is at a high enough reading level to actually comprehend my profile knows the correct answer.



rabbittss
Veteran
Veteran

User avatar

Joined: 29 Dec 2011
Age: 39
Gender: Male
Posts: 1,348

07 Apr 2012, 6:32 pm

starkid wrote:
rabbittss wrote:
The problem is that I've found, at least in my area, is how many people skip those sorts of questions.. I can only hypothesis it's directly proportional to the number of people who are also "Christianity and very serious about it"


I skip those sorts of questions because they are too obvious to be important to matching. Anyone who is at a high enough reading level to actually comprehend my profile knows the correct answer.


You are in California, I, however, am in the bible belt. Those questions are very important here.