This file is a work in progress. It is meant to provide voters with more detailed explanations of how the poll works, what the rules are and how they are enforced, and why it works the way it does. The ballot invite tries to keep these matters simple, and for most cases that is sufficient. But the real world doesn't neatly follow our conceptions, so discrepancies inevitably crop up. This file explains how we deal with them, and to some extent why. At least, that is the ambition. For now, this is just a rough edit of the previous year's file. You can help us improve it by asking questions.
Voting is by invitation. If you are interested, and think you are qualified, you are welcome to contact us and present your case, in which case we may send you an invitation. (Or if you actually send in a completed ballot, we might just say "thanks.")
We are looking for individuals who have a broad and/or deep exposure to jazz, and who regularly communicate to the public about jazz, in some media form or another: print and broadcasting are the well established forms, but internet media is common, especially blogs and newsletters. (I suppose podcasts may count, but that's not something I pay much attention to.) You don't have to make a living as a jazz critic (like, who does?), but you do have to work at it.
We generally do not seek out individuals who run or work for labels, do publicity, or are musicians, although we've made some exceptions, especially if they have solid credentials as critics, and have broad experience beyond their specific interests. We do ask voters "to refrain from voting for any album "in which they were directly involved as performer, producer, annotator, or in any capacity whatsoever."
We generally don't purge out invite lists of critics if they become inactive. As long as established voters feel confident in their ability to present credible ballots, they are welcome to continue voting. We will, of course, drop anyone from our mailing lists who asks. And we do drop critics known to have passed, although we don't assume that missing a poll means anything more than that you couldn't be bothered. (But way too often it means that the spam police intercepted our mail. It's very hard for us to ensure delivery, so it helps to be vigilant on your end.)
Every year we not only do our own research to find more jazz critics, we ask everyone we know to make recommendations. Most of our new voters come from these references. We welcome readers to send in the names (and references, and if possible, email addresses) of their favorite jazz critics.
For the first Jazz Critics Poll, Francis Davis focused on a small number of critics centered in New York City. Within a few years, he expanded his circle across the United States, and added a few European critics. The number of non-American jazz critics has expanded considerably in the last few years, but is still a distinct minority, and not something we [Americans] can easily remedy. Jazz is in many ways the real world music, so it would be nice to reflect that in the poll.
For Mid-Year Polls, we try to open up the website and send out invitations in early June, leaving a 2-3 week window for voting. The initial round of ballot invites will be send out to the jazzpoll email list sometime in the week of June 8-12. Additional invites will follow shortly. It's possible we may send out multiple invites via different routes in hopes that at least one will break through. (Didn't I mention spam police already?)
Deadline for voting will be Sunday, July 5, 2026. There is no need to specify a time of day. We're up late, and will count anything that shows up before we're finished counting (which will certainly slide well into July 6). The results will be published by ArtsFuse a week or so after deadline. Turnaround for publication should be quicker than for the end-of-year poll, as we'll have fewer ballots to count, fewer albums to footnote, and a shorter essay to write. The previous Mid-Year polls have had 90 and 113 voters, vs. 177 and 168 for the EOY polls.
The voting process is meant to be intuitive enough that you shouldn't have to read anything more than the Ballot Invitation itself. If you've voted before, the only thing different this time is that we've eliminated the "special category" sections for Vocal, Latin, and Debut, so this ballot is even easier. (Actually, previous Mid-Year Polls didn't have them either.)
This section offers a much more heavily annotated version of the Ballot Invitation. While the poll is intended to be as simple as possible, the world that we're trying to take a picture of is rather more complicated. Therefore, to see how the poll fits the world, and how it doesn't quite fit perfectly, we need to look at more details. This section will focus on how to fill out a ballot, and on what we do to check and count the ballot. Later sections will look even more closely at the album data and its presentation.
You can read the standard ballot invitation here. We keep a copy on the website for easy reference, especially in case voters lose their emailed copy. (If you didn't get your invite, and believe you should have been invited, you can respond to the one on the website. If you hadn't been invited, we will at that point consider inviting you.)
We've been trying to make the ballot as simple as possible. We've eliminated the "special categories" for the Mid-Year Poll, so no need to waste time on what is or is not Latin or Vocal or a Debut: all should be considered as appropriate for the two remaining categories:
In past years, we've tried to enforce category distinctions, but this year I would rather just let the chips fall where they may. The reason for having two separate categories for old and new music is that they're often incomparable. Today's latest music is difficult enough without having to compare it against a previously unreleased classic by Coltrane or Monk — as had happened in the early days of the poll, when only reissues were considered in the old music category. But the 10-year dividing line is just a recommendation. For instance, if an album less than 10 years old feels archival — perhaps the artist has died or become incapacitated, and the label is just putting up old tapes regardness of vintage — then it may make sense to vote it that way. If I had it to do all over again, I might have picked a more recent cutoff date, in the 6-8 year range, but rather than complicate the rule, I figure it's best to loosen up the enforcement.
It's less likely to find an old recording that feels so new voters insist on slotting it there, but there are other forms of ambiguity. I can't possibly list them all here, ranging from a bit of old music in an otherwise new album to a quick physical reissue of a self-released digital album on a real label. Then there are things like a 2-CD set with one old and one new set, where the old may be considered a bonus (perhaps in digital they're offered separately?). If you ask me, I'll come up with an opinion, and write it up in our Notes on Albums file, but if you just vote, I'll probably count it wherever you put it. If other voters disagree, I'll combine the old and new votes under whichever category gets the most votes (or, if tied, the one I rule for).
Two points here: (1) trust your instincts; (2) I'm not here to pick a fight.
One more concern here: when I said "released this year," that too is open to question. I meant released in 2026, but we've always allowed votes for albums released in the preceding year — sometimes qualified as "late last year," or as "new to me this year." The problem with "late last year" is that we'd have to set a boundary date, then we'd both have to figure out the exact release date to see whether it fits. So if we're allowing any 2025 albums, we might as well allow all 2025 albums. So where I say "this year" above, I'll accept any album released in 2026 or 2025. One of the few lines I will draw is that I'll reject any vote for any album first released in 2024 or earlier. (I will also reject any album scheduled for release in 2027. That's unlikely to be a Mid-Year problem, but it does happen every year in the EOY poll.)
"New for me" is obviously subjective, so I'm not going to question it. What I will say is that the main reason for picking a previous year release is to bring attention to something that was generally overlooked when it first came out. (That's most likely if it came out in December. Not many albums do, but unless a publicist is pushing out advance copies, hardly anyone can keep on top of such late releases.) So this isn't a rule, but just a bit of friendly advice: don't bother voting for any 2025 releases that got noticed in the 2025 EOY poll. It's not going to help them, and it's just going to make you look slow. The point of having a 2026 Mid-Year poll is to consider 2026 releases.
We ask you to list up to 10 New Jazz Albums and/or up to 5 Rara Avis. If you don't have the maximum, give us what you can. If you have 0 Rara Avis, leave it blank. No problem. If you only listen to reissues, you can leave New Jazz Albums blank. I've never seen that happen, but I'll accept your ballot. If you enter more than the maximum number of votes, the last ones on your list will be omitted. I will note that, and alert you. If you vote for albums that are not eligible, I will drop them, note them, and alert you.
If you wish to file an unranked ballot, say "unranked" somewhere I can find it, or alphabetize the list by artist name (surname for individuals; omit leading articles for groups; various artists may be sorted by title). I may fix the sort order of partially (or incorrectly) alphabetized unranked lists, but if you mark a list as "unranked" but do not make any effort to sort it, I will enter it in the order given.
If you don't do at least one of those things, I will treat it as a ranked list. Ranked lists will be formatted as numbered, regardless of whether you number them. If I wind up interpreting this wrong, you will have an opportunity to set me right. Ranked lists get extra points, but they only come into play when breaking a tie between albums with the same number of votes. Otherwise, the totals are sorted by the number of votes (voters), not points.
You don't have to understand the points system (explained further down. It is designed to be informative, fair, and balanced, without imposing any burden either on the voters or the counters. Unlike the Pazz & Jop system, you don't have to assign points, nor do you have to balance out your budget, nor can you use maximum votes to skew the results. (I've known voters there to give the maximum 30 points to their top 2 albums, then the minimum 5 votes each to the next 8.) Unlike the original Francis Davis system, nobody's saying that their best album is worth 10 times as many points as their 10th, or that unranked lists show a better understanding of game theory than ranked lists. And unlike my first stab at an alternative to the Davis system, there is effectively no penalty here for submitting an unranked list.
I'd like to reiterate this point from the invite file:
The point isn't to judge everything. You're just making recommendations, and in doing so, we will all benefit from your knowledge and taste. We don't give out plaques or ribbons. We don't throw banquets. All we are really doing is helping our readers to better understand the great bounty of jazz released each year.
I worry that lots of people put too much pressure on themselves when asked to fill out a simple list. You don't have to go back and relisten to everything, or anything. Just jot down some records you recall liking. No one can hear everything, or remember everything they've heard. But if a friend came over and asked to hear something new that you really like, just go with that. No one (at least no one who's ever had to do such a thing) expects that your list will be anything more than a transient opinion. But by combining your picks with those of many others, we get a broad (if not rigorously accurate) view of recorded jazz right now, as a living art form, and that turns out to be very useful, and helpful, to the many people who love jazz, but find its plenitude to be overwhelming.
Of course, you're welcome to use the days remaining before deadline to do your own further exploration. As we count ballots, the name of every album that receives a vote can be accessed with our albums list tool. (It isn't instantly updated, but will be every couple days or so.) Slightly more comprehensive is my tracking file (which I will add to as albums get votes, so should be a superset of the albums list). Other sources for 2026 jazz releases:
Just make sure to send your ballot in before the deadline. Actually, to make sure no one slips up, send it in as soon as possible. We allow you to change your ballot up to the deadline, if you come up with anything late-breaking or "new for you," or just want to fiddle with the order. And if you don't receive acknowledgment within two days, check in to make sure we got your ballot. (We've had people complain after publication that their ballots weren't counted. That's mostly because they didn't actually send them, but all kinds of things can go wrong with email.)
If you get an invite in email, just click on "Reply" or "Reply to Sender" (not "Reply to List"[1]), then type (or cut-and-paste) your list into the message body[2], under the bold "Your choices" lines. Before that, fill in a line below "Your name and primary affiliation(s)." Use whatever is current there: print magazines, websites (especially if you have a blog or newsletter; we'll include a link if we have one), radio station (usually call letters and location); you can mention a recent book if you have one; some people just say "freelance." We generally don't do links to the websites of multi-author publications (like AAJ, FJC, or NYCJR, let alone DownBeat or the New York Times) or radio stations, but we may make an exception for more obscure sources. We generally do not cite, much less link to, social media accounts.
[1] If you send email to the list, it will be trapped, instead of being sent to anyone on the list. As Administrator, I will get a notice, which I can easily miss before the message is automatically discarded. No real harm, as I usually get the reply directly, but it is annoying, extra work for me, and can lead to further errors.
[2] If you put your list into a separate file and attach it, it takes me an extra step (or several) to access it. (Worst case, which rarely happens, is I cannot.) All I ever do with it is cut-and-paste the text into another file (UTF-8 text, stripping markup). Even if you prefer using some other program to write, it should be as easy for you to cut-and-paste your text into the email as it is to attach the file, and that would be less work and trouble for me.
If you don't have an email invitation, you can copy the form from the website. Fill it out as above, and email it to Tom Hull, directly, or through this forwarder. If you send in a ballot without having been invited, we will at least consider your eligibility, and get back to you.
Don't bother trying to anticipate the formatting. Anything that's clear will be accepted (subject to rules explained below), and anything that isn't clear will be questioned. The procedure for processing your ballot is:
Changes are handled the same way, with edits to "log-ballots" and the data files as needed, regenerating new pages, and further email, until we are both satisfied.
If your list is unranked, please note that. If you alphabetize a list, it will be read as unranked. I may fix a poorly alphabetized unranked list (say, if one or two albums seem out of place), but I will leave an unalphabetized unranked list as you provided it. If you don't say "unranked," and don't alphabetize, I will count your list as ranked. If you say something like "in no particular order" and nothing else, I will count the list as ranked. If you don't like my interpretation, we can change it.
Your email may include the rest of the invitation, or not. Whatever you send in will be processed by a fairly sentient human, one with a great deal of experience deciphering ballots, and a more than passing knowledge of the subject matter. The artist name and title should be clear, but I'm pretty tolerant of typos. I keep a table of all the albums that have votes, so once an album has a vote, I have an authoritative record of how to format it. On the first vote for an album, I refer to my tracking file to check the formatting, and add it to the albums table. If there are any discrepancies, I double check, and correct my references. (If I find I made a mistake later, or just come up with a better way of formatting, I correct the albums table, and the change is propagated to every ballot voting for that album.) Thus your accuracy checks mine, and vice versa. It helps if you provide the label, but often it is not necessary. Again, the main value is error checking. Albums released in the previous year will be marked (e.g., '25 for 2025). For Rara Avis albums, I like to include the recording dates (year or year range). I keep track of that, and if I don't have it, I'll look it up (or ask you if I can't figure it out). If what you send in isn't clear, or is incomplete, or is just beyond my comprehension, I will make an educated guess, and check back with you. I may raise questions, or (rarely) make suggestions, but in the end your ballot is your own. We need to agree on that.
Once I have processed the ballot, I will return to you a formatted copy as entered it into the system. I mean to check this before sending it out, but I tend to work fast, and I'm a notoriously poor proofreader. You should check what I return, and let me know if there is anything you think should be changed. I am happy to revisit formatting details on albums (more on that below). Some are ambiguous and open to interpretation. Typos are certainly possible. One thing to be careful of is that sometimes I come up with the wrong title, maybe even the wrong artist. If I am not conscious of an artist having multiple titles, I may too quickly pick the first one I find. Some lines may be totally wrong. I am looking up names, and typing in numbers. Sometimes I read a number wrong (bad eyes), and sometimes I type it wrong (bad hands).
The formatted ballots have an optional line for a note. I may use this for a variety of reasons. Initially, it's usually to flag a discrepancy or raise a question. Most of those I clear up before publication. But I can use the line for further information. One example we've done is providing a link to the voter's separately published lists, which may be longer/different and/or provide more explanation, possibly because they aren't constrained by the poll rules and conventions. If you wish to add something to the notes line, or change something I have written, let me know. We have no hard and fast rules for its use.
The following sections (some with linked files) go into more detail, including some examples of things we need to deal with. While the sections cover general issues with albums, see Notes on Albums for specific cases of albums that we've needed to review this year.
As noted above, this year is 2026. The purpose of the poll is to identify the best albums newly released this year. However, we always have allowed voters to pick albums released in the previous year (in this case, 2025). The main reason for this is that some albums were only released after the previous year's voting period, so (unless voters received advances) couldn't have been considered in their release year. There usually aren't a lot of December releases, but it does happen, and it would be self-defeating not to allow them to be considered. Many more albums were released earlier in the year, but didn't get a hearing until after the deadline. Sometimes publicists picked them up late. (I've also seen them give new release dates for previously released albums.) More often it was just word of mouth, which only gradually kicks in. Every January, I discover dozens of previous year albums I was unaware of — many I only found out about from End-of-Year lists and polls.
Robert Christgau, in his Pazz & Jop poll, stressed subjective experience: he looked past release dates to allow picking albums that were new/important to him in the year. He usually includes more than a dozen albums with older release dates in his Dean's Lists (one in his 2025 list was released in 2018). Few voters go that far, but there are cases where albums get significant numbers of votes in two years: Michael Jackson's Thriller won P&J for 1983 after it was released in 1982 (only 1 of its 7 hit singles were released in 1982). Pazz & Jop had a rule where two years' votes would be added together if the album got more votes the second year than it did the first. We've followed that rule, but it is rarely invoked. The only case when it had an impact was in 2019, when Eric Dolphy's Musical Prophet won Rara Avis by adding its 2018 votes (when it was initially released on digital) and its 2019 votes (when it finally came out on CD).
As the Dolphy set shows, albums can have multiple release dates, varied by label and/or format, sometimes crossing year boundaries. We generally prefer the first release, but may go with the first one to get a vote, on the theory that it got recognized while the earlier one(s) didn't. But this can get messier when later releases add, remove, or reorder songs. These have to be resolved case-by-case. In the simplest of cases, a series of releases may be treated as logically one release. But at some point a release can turn into a reissue (which would make it belong to the Rara Avis category, even if the music is recent). It's not obvious what to do with remix albums, partly because they aren't common in jazz.
For the End-of-Year poll, we check previous year votes and add them in if the album gets more votes (not more points) than in the previous year. I don't recall this ever affecting more than 5 albums in a year, with 1-2 typical. But we do not apply this rule in the Mid-Year polls. One problem is that we get significantly fewer voters in Mid-Year Polls (at least so far). That makes cross-year comparisons more troublesome. Another is that it distracts from the focus on 2026 releases. Allowing votes for 2025 releases is one thing, but providing a mechanism to give them extra weight seems unwarranted.
Some polls try to divide calendar years, so that their eligibility overlaps the calendar year boundary (DownBeat, for instance; also the Grammys). Because we start polling in November, with a deadline before the end of December, we used to do this, picking various dates from Nov. 15 to Dec. 1 as cutoffs. What I've found is that most voters have no idea of (or interest in) precise release dates. (Some even have trouble with years.) If they get an advance for a December (or even a January) release, that's "new to them." So the simplest (least confusing) way is to use the calendar: January 1 marks the new year, so I'm happy to accept votes for December advances, but I refuse to count a January advance. Also, if we allowed votes for next year albums, the job of combining multi-year votes becomes much worse. (Also, I let one slip through that was scheduled but never actually released, permanently mucking up our records.) The only reason for voting for a next year advance is confusion, so we reject that.
We also reject albums released before the previous year, so in our case, that means anything released in 2024 or earlier. It's a fact of life that every year you're going to discover older music you haven't heard before. But it's not really germane to running a poll of new music. Any time I get a ballot with an album released outside of our time band, I discard that vote, note it, and refer the ballot back to the voter, who may then use the slot for another album.
The only difference for the Mid-Year Poll is that 2027 advances are extremely unlikely. The nominal calendar period for the Mid-Year poll is January 1-June 30. It is quite likely that voters will have advances of albums scheduled for later releases this year, and since they have them, some will vote for them. I advise them not to, if only because they can't get a fair hearing. But it doesn't strike me as a rule worth enforcing. So, in effect, it is allowed.
Albums are identified by artist credit, title, and label. While in most cases this is straightforward, this can get ambiguous, or even messy. We try to manage with this consistently and sensibly with a number of heuristics. These have never been properly codified, and probably never will be. (When AI automates this, it will not be by deriving the rules, but by pattern matching derived from thousands of examples humans have already figured out, which come to think of it is pretty much how humans figured this out in the first place.)
Once we figure out the artist/title/label for a given album, we store that information in a single "albums" table, with each album identified by a unique number. That number is then used in tables which encode ballots. (Voters also have their own unique number.) That way, album information is formatted the same on every ballot that votes for the album, and in the totals.
We can at least explain some of the conventions used to define albums. Some of these were originally set down as format preferences by Francis Davis. Others were derived from Tom Hull's long experience in managing his own database of over 46,000 rated albums. When we print out an album, we format it as follows:
Artist-Name, Title (Label)
Which is to say, the Artist-Name is in bold, the Title is in bold italics, and the Label is in regular type. Artist-Name and Title are separated by a comma and space. Label is inside parentheses. The parentheses may also include numbers before and/or after the label name. Before specifies he year, or range of years, when the music was recorded. We do this for rara avis releases, to show not just that they are old music but because that information is good to know. This is separated from the following label name by a comma and space. We use four-digit years, unless the second date in a range of years is in the same century, in which case the second date is reduced to two digits (e.g., 1968-72). The recording dates are usually omitted when they have already appeared in the title. Titles are not edited to conform to this formatting. Rarely, we include the recording date for a more recent album. This is due to some confusion, such as a voter voting for a recent album as rara avis. We may allow that, but feel it deserves a note.
The second optional number in the label field is a two digit release date, preceded by an apostrophe (like '25). This notes that the album was released in that year. The only case you should see in 2026 is '25, because albums released in earlier years are ineligible, and votes for them are rejected. (Same for releases scheduled for future years.)
Capitalization is generally normalized. All caps are usually turned into initial cap and lower case, except where the word is clearly an acronym. English language articles and short (1-3 letter) conjunctions and prepositions are lower case, except first and last words. Implicit breaks, such as between title and subtitle, are usually rendered with a colon. Ignorance of other languages is assumed, even when one knows better (so "Les" for French "les," "Y" for Spanish "y," but French "a" (unaccented) gets a pass. In former years, we never ventured beyond the Latin-1 character set, but recently we've started including UTF-8 encoded as HTML entities. (The website still uses Latin-1 as its default character set. Sorry about that.)
Brackets are used for extraneous notes, some derived from the package (like "[Original Soundtrack]"), some from the editor to help clarify some item.
Some field-specific notes follow:
Artist Name:
If two artists are given, they are listed separated by ampersand (&). If three or more artists are given, they are separated by hyphen and spaces ( - ). Francis Davis believed that slash (/) should only be used if the artists appeared separately, so we are following his rule. The extra spaces distinguish separate names from hyphenated names. One could use larger dashes for this, but this is easier and at least as clear.
Only the first tier of artist names is used. A second tier in smaller type or separated by a title is not used. "Featuring" artists are generally not used. Multiple artists are listed in whatever seems to be the most normal order (usually left-to-right, top-to-bottom). Rules like this tend to have exceptions, and we make them when they seem to make sense.
If a title is meant also to be a group name, but the artist names also appear on the cover, the artist names are used.
If there is no artist name, or it is listed as "various artists," the artist name field is left blank.
If the artist name is incorporated into the title, we may leave the artist name field blank. We don't do this if the artist name and title are simple, especially if the title alone could be misinterpreted, but we may suppress the redundancy of longer cases. This is a judgment call.
The "albums" table has a sort field, which is used for sort order within tied results. Individual artist names are sorted by surname. Leading articles are ignored for sorting group names. The sort key for an album with no artist name is the title, stripping leading articles.
Titles:
Subtitles aren't necessarily included, but when they are, they are separated from the main title by colon and space.
Labels:
Multiple labels, regardless of the number, are separated by slash (/). This may happen with joint releases, or with separate releases on multiple labels (e.g., in different regions or on different media) which we wish to combine together.
Albums with no known label are assigned to "self-released."
Or new music vs. old music. For previously unreleased music, we draw a line in the sand 10 years before present (for this year, that dividing line is between 2016 and 2017), with anything recorded since that line eligible for New Releases, and anything recorded before that line relegated to Rara Avis. Reissues, even of more recent music, are considered Rara Avis. In the past, we've tried forcing this rule, mostly to make sure that each record gets a fair count in its own slot. However, some cases are ambiguous, and some voters insist on their own definitions, so my rule here is to allow voters to decide which category any given record belongs in. (I may point out discrepancies, but will count the vote either way.) To help with fair totals, we may add points in the correct category for albums voted for in the wrong category.
The earliest polls had a Reissues category, but previously unreleased music of any vintage had to be evaluated under New Albums (sometimes called New Releases). This caused confusion, especially as most other polls combined reissues with previously unreleased older music in a "historical" category. When Davis changed the rule by introducing the current 10-year fence, he started referring to the historical category as Rara Avis. We've stuck with the term, although its applicability is somewhat less than obvious.
For New Albums and Rara Avis, voters are asked to submit ranked lists, in descending order. Each rank is converted to points, using a sliding scale.
For up to 10 New Albums, the scale is: 1: 3.0; 2: 2.4; 3: 2.0; 4: 1.8; 5: 1.6; 6: 1.4; 7: 1.3; 8: 1.2; 9: 1.1; 10: 1.0.
For up to 5 Rara Avis, the scale is: 1: 2.0; 2: 1.6; 3: 1.4; 4: 1.2; 5: 1.0.
This point scale was introduced for the 2024 Mid-Year Poll, and was continued for the 2024 and later Polls. Previously, the point scale was linear: from 10 to 1 for up to 10 New Albums, and from 3 to 1 for up to 3 Rara Avis. While the linear scale was easier to calculate, Hull felt that the range between top and bottom points was more likely to distort the results. Experience with assigned points polls, like the Village Voice's Pazz & Jop, suggest that, even when given complete license to assign weights to ranks, voters will either assign equal rank or stay within a fairly narrow band. (In P&J, voters can divide 100 points among 10 albums, with a maximum of 30 and a minimum of 5. That's 6-to-1, but maximum votes there are rare, so a 3-to-1 range is more typical of voters who seriously rank.) Davis considered and rejected the Pazz & Jop formula, figuring it was excessively burdensome both for voters and for himself. Part of Hull's reasoning was that in the Mid-Year Poll, there would be fewer voters, so the point-scale distortions of the linear system would be exaggerated. I doubt there is any objective way to optimize a points system, but intuitively the more compressed scale seems to be better at reflecting both the popular breadth and the depth of enthusiasm voters hold for their picks.
Davis also allowed the option of unranked lists, probably because some voters insisted on them. (In later years, he disparaged them, sometimes not even mentioning the option in his invite letters -- but we did always count them.) Davis simply divided the total points that would be given for a ranked list (55 for 10 New Albums picks, 6 for 3 Rara Avis picks) and divided them by the number of picks, so they were scored at 5.5 and 2.0 points. Hull found that when he mentioned the option of unranked lists, the number of critics opting for them roughly doubled. So he proposed scoring unranked lists at 1 point each. This gives voters an incentive to rank, but it's not so large as to greatly distort the results. In the 2024 Mid-Year Poll, the share of unranked ballots was about the same as it had been in recent years, so it didn't seem to be much disincentive at all. But voters who rank their lists are doing extra work, and providing us with extra information, which would seem to justify the extra points.
Several voters objected that giving extra points to ranked lists was penalizing voters who preferred unranked lists. For 2025 (starting with the mid-year poll), we decided to change the keep the point system, but change the sort, so the primary key is the number of votes (where both ranking and unranked voters have the same weight), and making points the secondary key, acting as the tiebreaker.
Early on, Francis Davis decided to add three special categories: two to highlight styles or genres that are indelibly significant parts of the jazz legacy but which tend to be overlooked or marginalized in top-ten lists; and one to help identify new musicians who are likely to become future stars. He asked voters to pick one record in each of these categories, and stressed that if they had picked any special category in their top-ten, they should name it (their highest pick, if multiple) as their category choice. In later years, Hull tried to generate more data by allowing voters multiple picks.
While the results were sometimes interesting, this added a lot of extra work to the poll, for not much gain. Last year's "Notes for Voters" took 14 paragraphs to try to explain these categories, not counting individual album case studies, but the bottom line was that they're pretty much whatever voters wanted them to be. Nearly half of the voters skipped one or more categories, with many skipping all three.
I dropped the categories from the Mid-Year Poll in 2024, but brought them back at Davis's request for the EOY Poll. Same for 2025, except that Davis had passed, so inclusion was meant as honoring his wishes. This year I think they should finally be eliminated.
If we later decide that it would be a good idea to spotlight these (or any other) categories, we could collect the data from the overall results, as part of our post-poll analysis.
Some questions from a previous year's draft.
Lots of things I need to write about here:
Also see the FAQ, which repeats some of this but in a form more oriented to the general public.