*A match-making metric* Getting a job in astronomy can be a fraught process. Institutions vie to select from a pool of candidates, who are in turn vying to select from a pool of institutions. *Phillip Helbig* has come up with a plan to make it all rather easier. Astronomy is a field in which there are few specialist jobs. Other branches of physics offer jobs outside of academia, including permanent jobs at a relatively early age, while essentially all astronomy jobs are within academia. If one considers the high degree of specialisation, it is clear that the number of potential jobs for a given astronomer is very low. Indeed, most astronomers with a doctorate in the field never get a permanent job and eventually leave academia. (To first order, each professor with students will, on average, have one student who goes on to become a professor with students.) There are more applicants than jobs on offer and a typical job might attract, say a few dozen applicants (though there is a wide range). The oversubscription is not that high, though, as applicants typically apply for several jobs. The usual scheme is that jobs are advertised, mostly in the autumn, people apply, applicants are ranked, and offers are made for each job until a candidate accepts. Good applicants will have several initial offers from desirable institutes. Lesser applicants could have offers from lesser insitutes or offers from better institutes after others have declined. Many will get no job offer at all. The scheme seems on the surface of it to work quite well: institutes rank applicants, good candidates might be able to negotiate more if they have more than one offer, and eventually all jobs are filled. Most jobs are advertised in the autumn so that competing offers can be compared without being forced to accept a less desirable job only because it has an earlier deadline. This can be avoided with a common deadline, as is exercised in high-energy physics, where most offerst adhere to a decision deadline of 7 January. However, there are many other things which could be improved. First, good applicants will have several offers. Although good for such candidates, that has disadvantages both for some other applicants and for some institutes. Consider an applicant who is very good, but not at the top of most lists, perhaps not at the top of any -- because an excellent, high-profile candidate occupies the top positions. The absolute difference in quality might be quite small. Nevertheless, in the first round the best applicant will get several offers and the second-best perhaps none, at least not from good institutes. Usually, because most jobs are advertised at approximately the same time, decisions must also be made within the same time frame. A good candidate, especially considering the possibility of negotiating on the strength of having more than one offer, might take essentially the entire time to accept. That means that the second-ranked applicant will not get an offer from any institute which ranked the high-profile candidate first, unless the high-profile candidate declined some offers before finally accepting one. That would not be in the interest of the high-profile candidate and probably rarely happens. The second-ranked will then, at best, have an offer from a lesser institute to which the high-profile candidate did not apply (perhaps one to which the second-ranked applicant applied for personal reasons). Being forced to accept such an offer, the second-ranked applicant is no longer available to all the other institutes who ranked the high-profile candidate first, and such institutes are forced to make an offer to a candidate lower down on the list. So while multiple offers are good for the top-ranked candidate, they are bad for good though lower-ranked candidates and for almost all institutes which ranked the high-profile candidate first. It is also arguably worse for the institute chosen by the high-profile candidate since that candidate might have negotiated some advantages which would otherwise not have been offered. *Ranking the Rankers* My suggestion to improve the situation would be that applicants rank jobs just as employers rank applicants. Such rankings could be given to a neutral, trusted third party in order to match the candidates: all candidates with an offer from their top-ranked institute are obliged to accept that offer and are removed from the pool. The lists are then updated, i.e. entries on lists from which a candidate or a job have been removed are moved up to fill the gaps, and the process repeated. Although not essential, candidates who have applied to a given institute could be informed when that position has been filled by another candidate, and institutes could be informed when one of their applicants has accepted a job elsewhere. The former usually happens eventually _via_ rejection notifications. The latter usually happens only when a candidate has more than one offer and declines all but one; applicants don't always notify institutes to which they have applied but from which they have no offer that they have accepted an offer elsewhere. The process is repeated until a round in which no applicant/job pair is removed, or until all jobs have been filled. *``Instead of the chaotic results of the current system, which might take several months to play out, a transparent algorithm maximises the overall good, optimally matching candidates to jobs.''* A deadlock situation could occur in which no pair is removed but there are still candidates without jobs and jobs that are still open. For example, candidate 1 might prefer institute A and candidate 2 institute B, but institute A has ranked candidate 2 the highest and institute B candidate 1. Obviously, some compromise must be made. My suggestion is that in such cases it favours the applicants, because in practice the difference between candidates will be small, perhaps even within the noise, and any insitute should have no problem accepting a candidate one position lower on the list. However, the difference between institutes, even of the same quality, could be much greater for candidates, primarily for personal reasons, perhaps because an institute is nearer, more familiar or colleagues work there. It would be obvious to the candidates if such a deadlock exists: they would have neither received an offer nor been informed that the position is filled. Thus, a match to a candidate's first-ranked institute would be welcomed, even if the candidate isn't the institute's first choice. An additional advantage of such a scheme is that it levels the playing field. With the current scheme, if the candidates involved in a deadlock are well connected, they could mutually agree to decline their offers, resulting in new offers from their preferred institutes. Those not well connected cannot take advantage of such an opportunity. In practice, there might be some risk, though well connected candidates might be aware of the shortlists involved. *Making metrics* An initial offer from a candidate's preferred institute is best for both the candidate and the institute although, as noted above, in the current system it could be associated with disadvantages for other applicants and other institutes. How should other combinations of candidate and institute be ranked? A simple function would be $C^{2}+J^{2}$, where $C$ is a candidate's rank for a certain job and $J$ an institute's rank for a certain candidate. In contrast to, say, $C+J$, such a function favours similar ranks. Good istitutes want good candidates, not just candidates who ranked the institute highly; good candidates want good jobs, not just jobs at institutes that rank them highly. So when a deadlock occurs, such functions could be calculated for all possible pairs of jobs and candidates and those with the best (lowest) value would be matched and removed from the pool. (The case discussed above with $C=1$ and $J=2$ is a special case.) The lists are then updated and the process then repeats. The entire process could be based entirely on that function, because a candidate being ranked first by the candidate's first-ranked institute -- $1^{2}+1^{2}$ -- is the minimum of such a quantity, $1^{2}+2^{2}$ describes the deadlock situation, and so on. In cases where the function value is the same, preference should be given to the cases with $C