Thirty-nine of the 174 people who have ever been in the cast have come back through the front door as the host. That is a little over one in five, and the fifth of the roster it picks is not a random one.
Thirty-nine names, and a median that gives the game away
The roster stores seven role flags against each person, and one of them records that they hosted. Thirty-nine of the 174 records carry it, which is 22.4%. The flag is thin in a way worth stating plainly: a single true or false, with no date, no count and no gap attached to it. Somebody who hosted once in 1979 and somebody who has hosted six times since 1994 are the same value in this file. What follows is therefore a piece about a group and not about a sequence. It asks who ends up in the hosting fifth. It never asks when they got there.
The data does hint at timing, but only sideways. None of the 39 is in the cast now, and every one of the seventeen people currently employed on the show sits in the group without the flag. That is consistent with hosting being something that happens to former cast members rather than serving ones, which is also what common sense expects, but consistent is as far as it goes. A flag with no date cannot establish an order of events.
Now the tenure figure, which is the reason this piece exists. The median cast member who has hosted served 6 seasons. The median cast member who has not served 2. The roster median is 3, so the hosts sit well above the middle and everyone else sits below it, a factor of three between the two groups. That is not obviously what you would predict. Hosting looks from the outside like a fame test, and fame and tenure are not the same thing.
Two objections have to be cleared first. One is that unfinished careers distort the comparison: all 39 host careers are over, while the group without the flag contains seventeen people still adding seasons. Recompute on finished careers only, 39 against 118, and the medians do not move.
The second objection is about direction. Sort the roster by length of service and the share carrying the host flag climbs with it. Of the 59 people who lasted a single season, 7 have hosted, or 11.9%. Of the 46 who lasted two to four seasons, 8 have, or 17.4%. Of the 58 who lasted five to nine, 21 have, or 36.2%. Collapsed into one comparison: 24 of the 69 people with five or more seasons carry the flag against 15 of the 105 with fewer, 34.8% against 14.3%, which on a two-proportion test is z=3.17, p=0.0015. Length of service predicts the invitation.
The top band breaks the pattern. Of the 11 people with ten or more seasons, only 3 have hosted, a rate of 27.3% that sits below the band beneath it. But eleven people is small enough that one name moves it several points, and four of the eight without the flag are Michael Che, Mikey Day, Colin Jost and Kenan Thompson, still in the cast and so not yet eligible for a flag that appears to describe alumni. That is a band too censored to argue from, not a reversal.
The same censoring runs through the decades and caps the headline. By decade of joining, the host rate runs 33.3% for the 1970s intake, 22.0% for the 1980s, 29.7% for the 1990s and 45.5% for the 2000s, then falls to 6.9% for the 2010s and to nothing at all for the 2020s: none of the 21 people whose first year falls in that decade has hosted. That is not a verdict on them. If the invitation takes years to arrive, the newest names have not had them. Joiners before 2000 carry the flag at 27 of 102, joiners from 2000 onwards at 12 of 72, so 22.4% is then a floor rather than a settled rate.
The seven who did not need the seasons
Seven of the 39 lasted exactly one season: Billy Crystal, Robert Downey Jr., Paul Shaffer, Martin Short, Sarah Silverman, Ben Stiller and Damon Wayans. They are the whole of the overlap between the host flag and the 59 people who lasted a single season, and if the tenure story were the entire story they should not exist.
The file offers no shared mechanism for them. Four of the seven have a writing credit and three do not. Three were hired straight into the repertory company and never carried the featured billing at all, while the other four were featured players who never made repertory. Their arrival years are spread from 1979 to 1993. Nothing in the roster separates these seven from the other 52 people who lasted a single season and do not carry the host flag.
What distinguishes them is not in the roster, and this is the one place on the page where that has to be said out loud. Each of the seven was known to an audience for work that was not this show. Billy Crystal, a television actor with a sitcom following. Robert Downey Jr., a film actor. Paul Shaffer, a musician and bandleader familiar from late-night television. Martin Short, a sketch comedian with a set of characters built on another programme. Sarah Silverman, a stand-up with her own audience. Ben Stiller, a film comedian and director. Damon Wayans, a sketch performer known from a different sketch series and from the movies. Those seven descriptions are general knowledge, not data: the file holds names, dates, season counts and seven booleans, and nothing resembling a career summary. They are the obvious reading of the group, offered as a reading rather than a result.
Take that reading and the tenure correlation stops being mysterious. The invitation is a fame test, as it always looked. Tenure predicts it because for most performers this show is where the fame comes from, and six seasons is the usual price of it, though talent could just as easily be what buys both the seasons and the fame. The seven are the people whose fame did not have to be made here, so the price did not apply. They are not a hole in the rule. They are what the rule is made of.
There is a second institutional blessing in the data, and it does not pick the same people. Twenty-one records carry a flag marking the person as appearing in a "best of" compilation. The tenure profile is identical to the hosts': median 6 seasons against 2 for everyone else. The membership is not. Sixteen of the 21 also hosted; five carry the compilation flag alone, being John Belushi, Chris Kattan, Tim Meadows, Cheri Oteri and Gilda Radner; and twenty-three people hosted with no compilation flag against their name. Between them the two cover 44 of the 174, leaving 130 marked with neither.
The sharpest difference is at the bottom. Not one of the 21 lasted a single season, and the shortest tenure in the group is Chevy Chase's 2. The host flag reaches down into the one-season group seven times; the compilation flag never does, not once in fifty-one seasons. That is a difference in kind. A booking is a decision about a person, and a person can arrive with the qualification already in hand. A compilation is assembled out of material, and material only exists if somebody was there long enough to make it. No amount of outside fame produces sketches that were never taped. The compilation flag is the tenure half of the host flag with the fame half removed.
One caveat on those 21, in the spirit of the caveat on the 39. The earliest joined in 1975 and the latest in 2001, and by decade of joining the group runs 4, 5, 11 and 1. That does not look like an honour the show keeps awarding; it looks like a record of a particular set of compilations that were made and released, with an end date on it. Treat the 21 as a snapshot, and do not read a modern performer's absence from it as a judgement.
What is left when both caveats are paid is one clean finding. Coming back as host is the closest thing the roster has to a measure of what a cast member became, and it tracks how long they stayed at three to one. The show's history is full of people who arrived unknown and left famous, and the counting is what that path would look like from the outside: stay, get known, get asked. The seven took the other one. The roster table sorts by seasons served and the cast biographies carry the flags person by person, so both routes are visible even though the reason for either is not.
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.