Every number on this site is computed from one data file. This is what is in that file, what the rules for reading it are, and where those rules stop working.
One file, 174 records
The source is Wikipedia's list of Saturday Night Live cast members, used under CC BY-SA 4.0 and last retrieved on 21 September 2026. Nothing is typed in by hand. A script parses that article's tables into a single JSON file, and the roster, the charts, the biographies, the statistics and these essays are all built from it. It holds 174 records. Each carries a name, a link to the person's own article, one or more runs of years, a sourced count of seasons appeared in, seven role flags, and, where a freely licensed photograph exists on Wikimedia Commons, an image with its photographer and licence. That last part covers 129 of the 174; the other 45 have no picture rather than something found loose on the internet. The photographs are listed with their credits on the image credits page, and the provenance is on the about page.
The rule that matters more than any of that is smaller and duller: every figure quoted anywhere on this site is computed from that file at the time of writing, never recalled and never carried across from another page. It sounds like bookkeeping. It is most of the difference between a reference site and a summary of one.
Here is what the rule is worth, because it has caught four errors in the pages already published here, and all four are corrected now. It caught more than four in drafts, which is a different and less interesting number, because a draft is allowed to be wrong. The statistics page opened its section on churn by saying that 1981 was the great emptying, with fourteen people leaving in a single year, and that the second largest was 1996 with twelve. Recount departures by the year each run ends: 1980 is the largest at fifteen, 1995 is second at twelve, and 1981 is ten, which makes it joint third alongside 1985 and 1986. No year in the file has fourteen departures, and 1996 has four. Every clause of that paragraph was wrong, the year included, and the one figure that survived the recount, twelve, belonged to a different year.
The quiz on the same page then asked which year saw the most cast members leave at once, and its answer key pointed at 1981. Of the four years it offers, the answer is 1995 with twelve, against ten for 1981, four for 2013 and nine for 2022. The key points there now, and the answer text says out loud that the true maximum, 1980 with fifteen, is not among the options.
The history page carried a different error from the same family. It called 1986 the largest single-year turnover in the show's history. Counting arrivals and departures together, 1986 is seventeen moves, which is fifth; 1980 is thirty. When that sentence was fixed, a sweep of the rest of the site found the biographies page making the same claim in different words, which is the part worth dwelling on. A wrong number does not stay where you wrote it.
What all four had in common is that they were plausible. 1981 is the season everyone remembers as the collapse, and 1986 is the year the cast visibly turned over. Recall produces the number the story wants. Counting produces the number the file has. The two agree often enough that it is easy to stop checking, which is why the rest of this page is rules rather than conclusions.
The rules, and where they stop working
A season here is a broadcast season, labelled by the year it started, so the 1975–76 season is "1975". A record reading 1975 to 1979 therefore describes four seasons and not five. Dan Aykroyd and John Belushi both read exactly that way, and both are sourced at four. The end year is the calendar year in which that person's last season finished, which is why subtracting the years and adding one reads a year too high on four records in five.
The cast-size chart on the roster page inherits that boundary and does not resolve it. It counts a person into every calendar year their run touches, first and last included, so Aykroyd's 1975 to 1979 occupies five columns on a chart of a career the file records as four seasons. The alternative is worse: dropping the final year would undercount everybody who left mid-season, and the chart has no way of knowing which of those two a given record is. The shape it draws is right and the column heights are a year wide in a way the sourced counts are not.
A run that is still going carries no end year at all, rather than a guess or the current year written in as though it were known. Seventeen people are stored that way today. Code that needs a number substitutes the current year at the point of use, and prose that quotes a figure involving them says so. This is the most dangerous field in the file: an absent end year read as a number comes out as zero, so a filter written casually drops all seventeen current cast members out of every date range without erroring, and only in the years that matter most.
A season's cast count includes anyone credited at any point in it, which is not the same as the number of people on screen on a given night. That is why the current column counts eighteen: the seventeen still there, plus Chloe Fineman, whose run ends in 2026. She is not in the cast now and she was in the cast during 2026, in the season that closed in the spring; both are true, and the count answers the second question, which a cast-size chart is for. It also means any total for a season still in progress is provisional in a way a finished season's is not.
The season count on each record is sourced rather than calculated, and it is the field most likely to look like a mistake. Take each run, subtract the start year from the end and add one, running anyone still on the show to 2027, the year season 52 ends, and the result disagrees with the sourced figure for 142 of the 174 records. Drop the added one and put a floor of one season under it, and the disagreement falls to 25. Those 25 are mid-season hirings and departures, which no arithmetic on a pair of years can see. The rule is that the sourced figure is never overwritten to make the column tidy, so a season count here will sometimes not match the year range printed beside it. That gap has a whole piece of its own, on joined, or arrived.
The limitation most likely to trip a reader is the one this site has tripped over itself. The seven role flags on each record are undated. Each is a bare true or false with no year, count or duration anywhere near it. A true writer flag means the person was credited as a writer at some point in their association with the show. It does not say in which season, for how many seasons, or whether the writing and the performing overlapped at all. Fifty-eight of the 174 records carry it, which supports the sentence "58 people have been credited both as writers and in the cast" and nothing whatever about when. The hosting flag, which 39 records carry, and the Weekend Update flag, which 24 carry, behave the same way. More than one essay here was narrowed in revision because a first draft read timing into a flag that has none, and that is a limitation of the data rather than a shortage of care.
What is absent is worth naming too. There are no episode counts, no sketch counts, no character lists, no salaries, no dates finer than a year, and no record of why anybody left. Questions of that shape get no answer here rather than an estimate dressed as one.
The distribution of tenures is a good place to stop, because it shows why rules this fussy are worth the trouble. Of the 174 records, 59 are of exactly one season and the median is three. Ninety-one are three seasons or fewer and 121 are five or fewer. Eleven people reach ten or more, and one reaches 24. Almost all of the weight sits in the first few seasons, which is where a rule that reads a year range one season too long does the most damage: on a record of one season, being out by one is being out by a hundred per cent. And the long end of that distribution is not finished. Four of the eleven who have reached ten seasons are still on the show, Kenan Thompson at 24 among them, so those figures are floors rather than final numbers, and they will be wrong in a known direction the next time the file is refreshed. Saying which numbers will move, and why, seems a better use of a page than pretending none of them will.
Cast dates and tenures on this site come from Wikipedia; the counting and the reading of it above are our own. More of these are on the articles page, and corrections are welcome at info@snlroster.com.