How the NRMP Match Works

[Lccjblu]

I've been trying to understand the whole ranking algorithm works by reading the NRMP website. To me, it appears as though a program's ranking of you is the determining factor aka programs have the upper hand in many instances.

For example, Andrew applies at Harvard, does well and is ranked #45 by the program. Ben also interviews and is ranked #44 by the program.

Andrew submits and ranks Harvard #1 on his list. Ben submits and ranks Harvard #29 on his list.

The NRMP algorithm starts and magically there is one slot left at Harvard. Ben, has not matched yet (he somehow did not perform well on any other of his interviews) and neither has Andrew, so theoretically the spot goes to one of them.

From what I read, Ben will get the spot because he was ranked higher by Harvard, even though, based on their individual ROL, Andrew clearly wants Harvard so much more since he ranked it number 1.


Is my interpretation of the algorithm correct?

---

Members don't see this ad.

 

[Raryn]

I've been trying to understand the whole ranking algorithm works by reading the NRMP website. To me, it appears as though a program's ranking of you is the determining factor aka programs have the upper hand in many instances.

For example, Andrew applies at Harvard, does well and is ranked #45 by the program. Ben also interviews and is ranked #44 by the program.

Andrew submits and ranks Harvard #1 on his list. Ben submits and ranks Harvard #29 on his list.

The NRMP algorithm starts and magically there is one slot left at Harvard. Ben, has not matched yet (he somehow did not perform well on any other of his interviews) and neither has Andrew, so theoretically the spot goes to one of them.

From what I read, Ben will get the spot because he was ranked higher by Harvard, even though, based on their individual ROL, Andrew clearly wants Harvard so much more since he ranked it number 1.


Is my interpretation of the algorithm correct?

Yes. Your scenario is absolutely correct.

The match takes into account everybodies preferences, but in general it uses the applicants list first. That is, Harvard ranking someone #1 cannot force them to come there... unless all of their other choices (that they ranked higher) didn't want them in the first place.

Basically, the way it prioritizes things is that applicants go to the absolute highest program on their list that they have a chance of matching at.

 

[organdonor]

For some reason sdn won't let me paste the address here, but search youtube for "how the nrmp match works" by mrbabymonkey.

 

[gutonc]

As mentioned, your analysis of the outcome is correct but your interpretation of the meaning of it is not.

 

[Lccjblu]

I Appreciate the responses guys. glad I was understanding the match process correctly. It's Disheartening that The applicant above who wanted the program less (ben) would be matched there despite only being ranked one tiny slot higher than Andrew especially since Andrew wanted it so much more. But hey I guess that's just how it works to make sure everyone matches SOMEWHERE.

 

[Abby_Normal]

Yes. This is only the case assuming Ben didn't make the list at his choices 1-28. Basically, the algorithm is designed to put the applicant at the highest program on his/her rank list that hasn't been filled. This does favor the program by giving them their preference, but it also favors the applicant by making it safe for the applicant to put his/her "reach" programs at the top of the list without screwing their chances with their safeties.

 

[Law2Doc]

I Appreciate the responses guys. glad I was understanding the match process correctly. It's Disheartening that The applicant above who wanted the program less (ben) would be matched there despite only being ranked one tiny slot higher than Andrew especially since Andrew wanted it so much more. But hey I guess that's just how it works to make sure everyone matches SOMEWHERE.

You are looking at it all wrong. Ben didn't want it less. For all you know he could have liked every place equally. He didn't get his first 28 places but the system is stacked in the APPLICANTS favor, to give them the best match they could get. So he was in free fall but still hit the jackpot. Hes probably thrilled. And maybe Andrew just threw Harvard on there as a lark and actually didnt expect it or maybe wanted it less. Hes probably thrilled with his second choice, which is what he was thinking more realistic anyhow. He's never going to know he came up just short for Harvard or that he lost out to Ben. For all he knows, he wasn't even on Harvards final list. And likely Harvard, which depending on the specialty, might rarely ever have to dip down to the forties actually wanted neither of these clowns, but saw them as slightly better than SOAP -- the program was the real loser here -- it's rank list meant very little as it ended up with it's 44th choice. All it got to decide was if it had to have one of these two it would prefer Ben over Andrew. But that still means there were 43 Steves and Marys and Joes it liked much better. So your analysis looks at it all wrong. Both of these lackluster applicants got the best matches they could, while a program got saddled with one of them, despite not ranking either of them particularly highly. Who is really the loser here? Score: Applicants 1, Crimson 0.

 

[dyeguy21]

As I posted in another thread:

Simple Example:
Ranks
Applicant A: 1. State, 2. County, 3. University
Applicant B: 1. State, 2. County, 3. University
Applicant C: 1. State, 2. County, 3. University
etc.

State - 2 spots: 1. App B, 2. App C, 3. App A
County - 1 spot: 1. App C, 2. App A, 3. App B
etc.

Match:
Round 1: App A goes first, wants state #1 and State ranked him and has a spot open. Applicant is matched.
Round 2: App B goes next, wants state #1 as well, State ranked App B higher than App A, App B is Matched, App A is still matched. All spots filled at program.
(This next step is what most people don't get, being "matched" is fluid until all possible match permutations are made).
Round 3: App C goes next, wants state #1 as well, State ranked C lower than B, B keeps spot, C is ranked higher than A, C takes A spot. A gets bumped to "rematch" list.
Round 4: App A goes again, wants state #1, all spots filled move to next rank, wants County #2, spot is open and applicant is ranked, A takes spot at County.
Repeat until all possible matches are made (no movement to rematch list or impossible match qualifications met)

This is why you rank programs the way YOU like them because the system is biased towards the applicant and their preferences. In the example above you see that County wanted App C as their number 1 pick, but because the applicant ranked them lower, the applicant already had a match for their #1 program. So if Applicant C thought that State was a "reach" but had a "guaranteed" spot at County and if they therefore ranked County higher they would have missed out on their chance to get into State. (i.e. Because County was always a sure thing as they ranked C highest, he would AT LEAST match there as long as he put them on his rank list ). As you can also see County "wasted" a rank (two actually) on someone that preferred another school, hence why the match is skewed towards the applicant as we have no "wasted" ranks.

 

[operaman]

It amazes me how much misunderstanding there is surround the match algorithm. It's so very simple yet seems to evoke much confusion.

Algorithms aside, I will argue that the candidates list "matters" more in some ways because a tiny change in order can completely change your outcome. If a program typically matches in their top 15, then much of their actual order doesn't matter until they get closer to 15. Any changes in order prior to that point would have no impact on the makeup of their incoming class (assuming it was an average year for them). Contrast this with applicants who can change the future of their training by switching their 1 and 2, or their 2 and 3, and so on.

 

[Lccjblu]

Why should an applicant's desires trump the program's? Why should a program be forced to accept an applicant that they liked less, simply because said applicant "wants" it more?

Your situation sets up something of a false scenario because the two applicants are adjacent to each other on the program's rank list. What about if Ben was #1 and Andrew #20? Should Andrew still get to match there because he "wants" it more?

And what about Ben? Is it fair for Ben to get passed over for a position when the program liked him, and he has already fallen far down his rank list. Why should he miss out on yet another spot (and possibly the last spot on his rank list) because the guy behind him in line wanted it more?

You bring up some interesting points. The reason I had the two ranked next to each other by the program (44 vs 45) was to insinuate that the program wanted the two applicants just about the same (especially in large IM programs where they interview 500+ applicants), and that is why I was concerned about an applicant's desire to match at a program based on their rank list.

Anyway, I totally agree that Ben should match since he was rightfully placed ahead on the rank list. My point was just to say that when two applicants are desired basically to the same degree by a program, it sucks for Alex in the situation I described.

 

[Winged Scapula]

Since this comes up every year, I'll Sticky this thread and the video .

 

[MEN2C]

Thanks for the video!

 

[wingshorns]

The part that I am still confused about is who is more likely to obtain a spot in a program out of the following two candidates:

1) The candidate who ranks the program #1 and, say, the program ranks him #90. If the algorithm is just surveying everyone's #1 choice at first, let's say the #90 ranking is sufficient to find him a tentatively open spot in the program since it is an IM program with a plethora of ranked candidates.
2) The candidate who ranks the program #2 and, say, the program ranks him #10. Will this person be able to displace the above candidate if this person does not obtain his #1?

Hopefully this made sense.

 

[rd31]

The part that I am still confused about is who is more likely to obtain a spot in a program out of the following two candidates:

1) The candidate who ranks the program #1 and, say, the program ranks him #90. If the algorithm is just surveying everyone's #1 choice at first, let's say the #90 ranking is sufficient to find him a tentatively open spot in the program since it is an IM program with a plethora of ranked candidates.
2) The candidate who ranks the program #2 and, say, the program ranks him #10. Will this person be able to displace the above candidate if this person does not obtain his #1?

Hopefully this made sense.

Yes.

 

[IMPD]

Yes.

The thing to remember is that once you've lost your #1 spot on your list, everything moves up and your #2 is now your #1, etc.

 

[labrat10]

Still confused about how the match algorithm works, especially regarding advanced programs that guarantee some but not all prelim spots.
For example, program X is an advanced neurology program that has 8 neurology spots and guarantees 4 internal medicine prelim positions. Assuming 6 of the applicants who matched into neurology at program X also ranked the prelim program there #1 in the supplemental list, who gets the 4 spots? I know the IM spots have nothing to do with the internal medicine interviews or whether or not you get one in the first place. Does the computer just give the 4 spots to the 4 applicants who match higher in the neurologist rank list for the program?

 

[Raryn]

Still confused about how the match algorithm works, especially regarding advanced programs that guarantee some but not all prelim spots.
For example, program X is an advanced neurology program that has 8 neurology spots and guarantees 4 internal medicine prelim positions. Assuming 6 of the applicants who matched into neurology at program X also ranked the prelim program there #1 in the supplemental list, who gets the 4 spots? I know the IM spots have nothing to do with the internal medicine interviews or whether or not you get one in the first place. Does the computer just give the 4 spots to the 4 applicants who match higher in the neurologist rank list for the program?

There's two ways that this can be done, and both ways it is a simple match.

1) Exactly as you said, X is an advanced neurology program that has 8 spots and guarantees 4 prelim positions in their corresponding IM program. What that means is that the IM program has a separate "track" for their prelims that is guaranteed to be filled with neurology. That track will have a rank-list. You will rank them, and they will rank you, and it will go through a match. The rank list may be similar (or identical) to the main neuro program rank list... but it could also be that the IM program makes the rank list. In this possibility, it is possible that none of the neuro residents want to do their prelim there (preferring a cush prelim elsewhere), so that could leave 1, 2, 3, or even all 4 of those prelim spots in that track empty. If thats the case, there's usually a rule set up to revert them to the IM programs general prelim pool and just match an extra person there.
2) The neuro program has 8 spots that are split between two tracks. 4 of them are "categorical" and 4 are advanced. The categorical program has the prelims built in. Thee are 2 separate program codes and you rank them both separately, and the program has two rank lists. You decide which one you rank higher.

 

[dyeguy21]

If you're STILL having trouble understanding this: http://fivethirtyeight.com/features...ut-their-future-in-the-hands-of-an-algorithm/

The algorithm they use for match has won a NOBEL PRIZE.

It favors you because you are the one proposing. i.e. The programs don't get to go out and recruit you, you have the choice in who's proposal you decide to initiate. (This change happened in 1997)

Although you have the advantage in the algorithm, the NRMP has said statistically it has made little difference, but (supposedly and not likely given everyone not understanding it) increases the public perception of the process.

 

[iniquus]

I understand how the match works--everyone should rank things in their order of preference, you'll be tentatively matched at a program highest on your ROL until you are preferred by another candidate when no spots are left, then the process repeats until there are all stable solutions. My understanding with most algorithms is that there should only be "one ideal answer" when it is run. However, I have been told by an NRMP member that in actuality, they do run the algorithms multiple times for precision and reliability since it's possible that different results can occur each time it is run (I'm assuming the differences are minimal on the grand scheme of things unless its computer error). That blew my mind and I'm trying to wrap my mind around it.

Also, for the couples match, it seems that really complicates the match algorithm even further. If we refer to our trusty resource wikipedia, it mentions that the couples match is "NP-complete" and that "the algorithm does depend on some arbitrary factors (e.g. the order in which applicants are processed), one or both of the individuals could end up in a better position by chance alone, although this is extremely unlikely." Can anyone translate this game theory mumbo jumbo for me?

 

[dabeags]

My understanding with most algorithms is that there should only be "one ideal answer" when it is run. However, I have been told by an NRMP member that in actuality, they do run the algorithms multiple times for precision and reliability since it's possible that different results can occur each time it is run (I'm assuming the differences are minimal on the grand scheme of things unless its computer error). That blew my mind and I'm trying to wrap my mind around it.

Unless the inputs change then the output will be the exact same every time. I'd imagine they run it multiple times just to be sure there is no programming error or the like.

 

[iniquus]

Unless the inputs change then the output will be the exact same every time. I'd imagine they run it multiple times just to be sure there is no programming error or the like.

I don't think it's as simple as that. Straight from an NRMP representative that results can vary when the algorithm is run each time--even depending on the order of how applicants are processed. If you also read up on the literature about the stable matching algorithm, there is not always solely one possible "stable matching" and that it's a spectrum dependent on which party is favored.

Just found this thread now, pretty useful: http://forums.studentdoctor.net/threads/nrmp-match-algorithm-foolproof.900665/page-2. Oh well, < 2 weeks, I'll just have faith in the process--I don't need to know what's going on behind the scenes. Good luck everyone.

 

[iniquus]

The algorithm and principles involved are pretty interesting at the more complex level.

But while this may come across as paternalistic, for all intents and purposes it doesn't matter to the applicant in its practical application, so the more simplified principle (rank all programs you want in the order you want them) is the important one to convey.

Yeah, it was more out of interest as all this complex stuff would only matter to <0.1% of applicants. I know there's no strategy to ranking other than to rank it as you prefer them, regardless of post-interview communication.

 

[NotAProgDirector]

I was going to post that thread/post for you. If you have any questions about the algorithm, feel free to post here or PM me.

 

[lego1234]

I was going to post that thread/post for you. If you have any questions about the algorithm, feel free to post here or PM me.

I applied to two specialties as I was equally interested in both. Instead of continuous ranks of each, I ranked programs based on quality of training and location. Hope that won't be bad.

 

[gutonc]

I applied to two specialties as I was equally interested in both. Instead of continuous ranks of each, I ranked programs based on quality of training and location. Hope that won't be bad.

It's neither good nor bad, it just is. You'll match where you were going to match in whichever individual specialty you ultimately wind up.

 

[1dayatatime]

i've read several sdn threads on this topic and what i don't get is how the programs are going to rank us applicants. it seems they obviously rank based on their perception of how much an applicant will want to go to their program otherwise things like number one emails, calls from mentors, and asking us q's like "where does your family live" wouldn't matter. if programs rank us based on their perception of how much we want them, instead of objectively who they want regardless of whether that applicant is likely to go to that program, wouldn't it make more sense for us applicants to rank based on how much we think the program liked us versus the other applicants, especially if we're applying to a competitive specialty with approx 3 spots per program?

 

[BigRedBeta]

@1dayatatime Your general assumption is wrong. Programs in competitive specialties are not in any risk of not filling (in a general sense, nothing is 100%), and they aren't basing their ROL on the assumptions of applicants...at least they shouldn't be as students are notoriously fickle.

Further, you are wrong in the primary intent of calls from mentors - particularly if people know each other...that phone call on your behalf is to explain your strengths as a candidate, not to say "this student I know loves your program".

Lastly, how are you going to make any sense of how a program likes you compared to other candidates? That's information you could never, ever, ever know.

Bottom line, you are best served by ranking programs in the order that you genuinely prefer them. Any deviation from that raises the likelihood that you end up at a program that you don't like as much...or in worst case scenarios, you misread the situation and end up unmatched when you could have been successfully matched.

 

[rokshana]

i've read several sdn threads on this topic and what i don't get is how the programs are going to rank us applicants. it seems they obviously rank based on their perception of how much an applicant will want to go to their program otherwise things like number one emails, calls from mentors, and asking us q's like "where does your family live" wouldn't matter. if programs rank us based on their perception of how much we want them, instead of objectively who they want regardless of whether that applicant is likely to go to that program, wouldn't it make more sense for us applicants to rank based on how much we think the program liked us versus the other applicants, especially if we're applying to a competitive specialty with approx 3 spots per program?

not sure where you got this notion that the programs are basing their ROL on the perception of how much an applicant wants to be there...the rank based on which applicants they feel are a good fit for their program...ones that will work hard, are eager to learn and will get along with others...and those that will be pass their board...if you have been reading on sdn, you would have also noted that it has been posted ad nauseum that #1 emails, thank you notes, DON'T matter and mentor calls only have some sway if the mentor is known (and respected/liked) to the PD.

they rank based on their preference for students...you should rank based on your preference for programs...don't overthink this.

 

[1dayatatime]

i guess i've become more cynical as i go along. people keep telling me to "play the game" and i feel that as an honest individual who is terrible at lying, i don't seem to be as good as others appear to be at maneuvering questions asking me whether i have ties to a certain region. I.e. I told the truth, and was later explicitly told by someone else that they lied because "they can't track down that information about a random family member". Why do they ask these questions then? I realize that someone with ties might be "a better fit" for the program, but I might also be "a better fit" without ties to the region but why do I get the feeling as soon as I said "no" they resigned me to below all the people who had ties to the region?

 

[jennashell]

I agree with you. During the interview - some wanted to know who you have in the area or if you are connected. Fortunately I did have a family member ( who I don't even talk to that often but I wasnt lying atleast) . I have been away from my parents since high school and living out of suitcases for as long as I remember. For someone like myself who doesnt even have a bf or kids or any sort of strings, these questions are worrisome. I dont want to be overlooked because I am not from that area. At one point, I said " I dont mind moving to a new city and explore my way through since after all it is a new chapter of my life."

 

[NotAProgDirector]

i guess i've become more cynical as i go along. people keep telling me to "play the game"

I think that SDN suggests that there's more "game" than there is. If you're applying to programs in NYC, or Boston, or some other popular metropolis, I doubt anyone is going to care if you have "roots" in the area. Because lots of people apply to Cali, some programs may try to sense out who is serious and who is just applying -- although it really should make no difference in their ranking. If you apply to a place that's less popular, then people may ask -- simply because their experience in the past may have included people who were unhappy and ultimately left. And often, the question is just to see if you either have some linkage to the location, or at least have done your homework and know something about the location.

 

[mvenus929]

You are looking at it all wrong. Ben didn't want it less. For all you know he could have liked every place equally. He didn't get his first 28 places but the system is stacked in the APPLICANTS favor, to give them the best match they could get. So he was in free fall but still hit the jackpot. Hes probably thrilled. And maybe Andrew just threw Harvard on there as a lark and actually didnt expect it or maybe wanted it less. Hes probably thrilled with his second choice, which is what he was thinking more realistic anyhow. He's never going to know he came up just short for Harvard or that he lost out to Ben. For all he knows, he wasn't even on Harvards final list. And likely Harvard, which depending on the specialty, might rarely ever have to dip down to the forties actually wanted neither of these clowns, but saw them as slightly better than SOAP -- the program was the real loser here -- it's rank list meant very little as it ended up with it's 44th choice. All it got to decide was if it had to have one of these two it would prefer Ben over Andrew. But that still means there were 43 Steves and Marys and Joes it liked much better. So your analysis looks at it all wrong. Both of these lackluster applicants got the best matches they could, while a program got saddled with one of them, despite not ranking either of them particularly highly. Who is really the loser here? Score: Applicants 1, Crimson 0.

Your interpretation of the numbers themselves is a very surgical one. In an IM program of 30 (or more) interns (I don't know how many Harvard has, but I assume it is many), falling to 44 on your rank list would be amazing, given that you had to interview 300+ applicants to get your list. Now, in a surgical program,of only a handful of interns, then yes, falling to a rank of 44 on a list of 75 might be perceived as bad. But presumably, the program is grateful in either case that they filled. We've seen a handful of times when a program thinks they don't need to interview as many applicants as others in the field because they have a great reputation and end up having open slots at the end of the Match.

So in the end, the applicants that matched and the program that filled are both probably happy that they don't have to endure the SOAP, even if there is some initial lamenting over the exact list when match results are first announced.

 

[Brucelaa67]

I've been trying to understand the whole ranking algorithm works by reading the NRMP website. To me, it appears as though a program's ranking of you is the determining factor aka programs have the upper hand in many instances.

For example, Andrew applies at Harvard, does well and is ranked #45 by the program. Ben also interviews and is ranked #44 by the program.

Andrew submits and ranks Harvard #1 on his list. Ben submits and ranks Harvard #29 on his list.

The NRMP algorithm starts and magically there is one slot left at Harvard. Ben, has not matched yet (he somehow did not perform well on any other of his interviews) and neither has Andrew, so theoretically the spot goes to one of them.

From what I read, Ben will get the spot because he was ranked higher by Harvard, even though, based on their individual ROL, Andrew clearly wants Harvard so much more since he ranked it number 1.


Is my interpretation of the algorithm correct

The interpretation ''a program's ranking of you is the determining factor aka programs have the upper hand in many instances'' is wrong.
Lets look at similar example.
Andrew applies at Harvard, does well and is ranked #1 by the program. Andrew also applies at UCLA and is ranked#29 by the program.
For some reasons, Andrew ranks Harvard at #45, and UCLA #44.
The NRMP algorithm starts and Andrew isn't to match at his first 43 programs in ROL. And there is still a vacancy at UCLA and Andrew fills it (tentavitely at first, but doesn't get displaced).
So even though Harvard strongly wanted Andrew but didn't get him. So whats the interpretation now?

 

[ThoracicGuy]

The interpretation ''a program's ranking of you is the determining factor aka programs have the upper hand in many instances'' is wrong.
Lets look at similar example.
Andrew applies at Harvard, does well and is ranked #1 by the program. Andrew also applies at UCLA and is ranked#29 by the program.
For some reasons, Andrew ranks Harvard at #45, and UCLA #44.
The NRMP algorithm starts and Andrew isn't to match at his first 43 programs in ROL. And there is still a vacancy at UCLA and Andrew fills it (tentavitely at first, but doesn't get displaced).
So even though Harvard strongly wanted Andrew but didn't get him. So whats the interpretation now?

I'm pretty sure the OP doesn't have to worry anymore since that was a 3 year old post...

 

[Brucelaa67]

I'm pretty sure the OP doesn't have to worry anymore since that was a 3 year old post...

Right. But, It isn't meant for OP only, but anyone who is viewing this page in the preset times, trying to understand the whole match concept.

 

[rhomboid]

There are still instabilities to the match when couples are included.
1. Consider a program with a ROL of A1 B1 B2 A2.
2. The program has two spots.
3. The program is not aware of couple matching status of A1-A2 and B1-B2.

If the algorithm starts from A1 to B2, it will match A1 A2.
If the algorithm starts from B2 to A1, it will match B1 B2.

I guess there is not a solution to this. But given most programs are much larger than 2 spots and the fraction of couples is small, this usually won't be a problem.

 

[Oso]

 

[NotAProgDirector]

There are still instabilities to the match when couples are included.
1. Consider a program with a ROL of A1 B1 B2 A2.
2. The program has two spots.
3. The program is not aware of couple matching status of A1-A2 and B1-B2.

If the algorithm starts from A1 to B2, it will match A1 A2.
If the algorithm starts from B2 to A1, it will match B1 B2.

I guess there is not a solution to this. But given most programs are much larger than 2 spots and the fraction of couples is small, this usually won't be a problem.

Although you're correct that couples create complexities for the match, your example isn't one of them. For this example, there's one program (let's call it MBH) with 2 spots, and 4 applicants. So no matter what, at least 2 people will be unmatched.

We also need the couple's rank lists. Let's say that they both rank only the option that they match together: they both rank MBH - MBH as their only option.

The couples submit a joint linked rank list. So they are processed together by the algorithm.

If we start with the A's:
1. A1 and A2 prelim match to MBH
2. B1 knocks out A2, which knocks out A1 also. B1 and B2 match to MBH

If we start with the B's, the same result. B1 and B2 prelim match, A1 and A2 don't bump them out.

However, the results are different if the couples include the no match options. Each couple submits an ROL with 3 options -- MBH/MBH, MBH/X, and X/MBH where X = no match.

Now:
A1 and A2 prelim match to MBH
B1 and B2 bump out A1 and A2, prelim match to MBH
A1 now bumps out the B1 and B2 couple's match (the option where A1 matches and A2 does not). So now A1 is prelim matched to MBH.
B1 prelim matches to MBH (without B2)
Net result: A1 B1 match. Happens no matter what order you go in.

 

[BorntobeDO?]

If you're STILL having trouble understanding this: Another 34,000 People Are About To Put Their Future In the Hands Of An Algorithm

The algorithm they use for match has won a NOBEL PRIZE.

It favors you because you are the one proposing. i.e. The programs don't get to go out and recruit you, you have the choice in who's proposal you decide to initiate. (This change happened in 1997)

Although you have the advantage in the algorithm, the NRMP has said statistically it has made little difference, but (supposedly and not likely given everyone not understanding it) increases the public perception of the process.

Because Programs can choose just to not rank a candidate at all. Which then throws it all back to them. Its the simple marriage algorithm in which the proposer is favored (i.e. the medical student). But if a program just doesn't rank people, then it doesn't work out the same as the classic scenario in which each bride has to rank each groom. Plus simple marriage doesn't start with the bride interviewing each groom before they propose. All said, it seems to me some programs actually prefer the SOAP and its shenanigans.