There were a lot of ways to get 2022 wrong. The most common, and by far the most forgivable, came from simply not following the race closely. This is pretty common: people simply do not follow politics regularly, and of those who do, even fewer care about midterm elections. And for that subset who do care about midterm elections, they likely only know one thing about them: that the President’s party loses seats. And throughout the year, no President looked more primed to face a shellacking than Joe Biden. The economy was doing poorly, he was deeply unpopular, and Democrats did terribly in the 2021 elections. At a glance, a Republican sweep made total sense. But, as we all know, the story was far more complicated. Backlash to the Supreme Court’s decision to overturn Roe v. Wade in the summer provided Democrats with a massive boost that resulted in a historic overperformance for an in-power party in a midterm. There were many strong indications that this would happen for months leading up to the election. These were indications that should have been easy to track and understand for anyone who regularly follows elections, whether it be professionally or just as a hobby. For months, they presented a clear picture: a Republican blowout was not in the cards, Democrats were clearly favored in the Senate, and the House was meaningfully competitive. Of those who were unwilling and/or unable to heed these signs, the first were the obvious Republican partisans. Despite being perfectly capable of knowing better, they were so desperate to see their party win in a landslide that they refused to believe anything otherwise (or just had a monetary interest in lying). This group ranged from teenagers on Twitter to entire outlets like RealClearPolitics, which would cook their own data to paint as rosy a picture for Republicans as possible. It even extended to people outright involved in Republican politics. In the weeks leading up to the election, practically everyone who had ever worked for a right-leaning campaign was talking to some reporter about how Georgia and Pennsylvania were locks for their side and to watch for an upset in Washington. But it would be a disservice to pretend that the only people to have a bad 2022 were accounts named BlakeMastersEnjoyer or polling firms named BlackpilledData. Marching to the same tune as those morons was practically every political news outlet in the country. It is this class that stands as by far the most egregious offenders. Every single one of those desks, websites, and ratings organizations were full of professionals who were perfectly capable of seeing the very real signs of a close year for what they were. But they would refuse to do so to a degree that would reach outright absurdity. And no organization would be as consistently, arrogantly and disastrously wrong in 2022 as Nate Silver’s FiveThirtyEight. While Silver and his allies heavily contest the assertion that they even got 2022 wrong (they did), one hard fact remains: they had Republicans favored to win the Senate and they lost it. That made it the first election since 2016 that they outright called incorrectly, which is striking in and of itself. But even that still understates the extent that they overestimated Republicans. For the House, which they would only barely call correctly, their actual model (the “classic” forecast) would be off by 10 seats compared to the final results—a record worse than competitors ranging from other statistical models to subjective ratings. That is not at all what you would expect from the gold standard of political analysis. Instead of revealing the true signals of the election, FiveThirtyEight emitted a cacophony of noise. The end result was coverage that left its readers less informed the more they read it. Now, it will no longer exist, at least not in the way we’ve known it for the past decade. Rumors about the site’s fate had been swirling around since the start of the year, and they were just recently confirmed: the site’s contract with ABC News will not be renewed when it expires this year. Half of the staff has already been laid off, and Silver himself has confirmed his eventual departure. Thanks to the yeoman’s work put in by his lawyers, he will still have access to his models, which will invariably be put into action somewhere for next year’s election. But it will be under a different brand, with a different staff, and perhaps most important of all, with a far different reputation. No matter what form his work will take in 2024, Silver will never be seen as he once was, and is no longer capable of ever being seen in that way again. That, in itself, is a major development. So, how did we get here? What happened over the past decade that caused Silver and his website to go from being revered as the cutting edge to utterly incapable of handling an election like 2022? To fully understand, you need to start at the very beginning. Origins If there’s one thing that even FiveThirtyEight’s most virulent critics have to give it credit for, it’s that it was a legitimate pioneer in its field. Despite how much attention U.S. elections have always received in the media, the quality of campaign coverage had never come close to meeting the importance of the topic. Not only were elections covered like horse races, but even the horse race coverage was totally inadequate. At best, it would consist of unsophisticated debates about the results of individual polls. At worst, it would devolve to pseudo-scientific old wives tales with names like “The 13 Keys to the White House.” This incompetence would continue into the internet age. Even new outlets completely dedicated to campaign coverage completely miss the mark. There are a lot of examples from this era, but my personal favorite will always be the final prediction made by RealClearPolitics for the 2000 election. On November 6th, 2000, the day before the election, the self-declared “website for the political junkie” declared that “the real debate is not who is going to win the election, but whether Bush will win 308 electoral votes or 474 electoral votes,” and that “George W. Bush has a better chance of carrying New Jersey and Vermont than Al Gore does of becoming the next President of the United States.” Their final forecast was for Bush to defeat Gore by a margin of 10 points in the popular vote and by 446 to 92 in the electoral college. They seriously thought this would happen. Of course, this one prediction was far from the entire reason why the public was becoming disillusioned with traditional campaign coverage. The back-to-back missed calls of the 2000 and 2004 elections by cable networks played a far larger role. I bring RealClearPolitics up, however, not just because it is funny (although it is), but because it shows the standards that existed even at what was supposed to be the highest level of analysis. By 2008, an outlet that cast itself as the place for the deepest political discourse was just one cycle removed from sincerely forecasting George Bush as favored in California and Illinois. It was a deeply inefficient and highly unoptimized field. A correction was more or less inevitable. The only question was who would be the one to stake their claim. And, as fate would have it, that someone would end up being Nate Silver. Before he started modeling elections, Silver had essentially no connection to politics. He graduated with a degree in economics at the University of Chicago in 2000 and spent the first few years of his career alternately working as a consultant and playing online poker. He would begin his first work in “popular statistics” by making a model for baseball statistics in his spare time, which he would sell in exchange for a position as a writer at a Baseball Prospectus, a sabermetrics site. Silver would work diligently in his new job, personally producing the site’s advanced statistical forecasts over several years while also writing hundreds of articles. After a couple of years, during the doldrums of the 2008 Democratic Presidential primary, he would seemingly decide he was good enough at his work to branch out into an entirely new field, making his first foray into electoral analysis writing anonymously on The Daily Kos. The article, titled “HRC Electability in Purple States,” is an interesting piece to look back on now, over 15 years later. Compared to his staid, above-board persona now, “poblano,” as Silver called himself, was a virulent partisan: he referred to George Bush as “Shrub,” referred to the Democratic and Republican parties using “us” and “them” respectively, and declared in the replies that he would “make no apologies about [his] support for Obama.” The article itself consisted of the averages of crosstabs across a single poll which were used to make the point that Hillary Clinton was less electable than other Democratic candidates. Silver’s piece would catch the attention of the site’s readers, who would engage in heated debates over his points in the replies. He would go on to post near-daily over the course of the primary, steadily building up an audience. His pieces would, more often than not, follow a theme: setting up faulty mainstream media analysis as a foil, and then explaining why it was wrong—almost always because it was underestimating Obama. His clear pro-Obama bias frustrated Clinton supporters reading The Daily Kos, but as the primary continued and Obama kept rising, nobody could say he was actually wrong. By the time the primary began, Silver had already built a model for it. He made his first big call just ahead of Super Tuesday, predicting that Obama would win 833 delegates. It would end up being within a dozen delegates of the actual results. This gave him his first burst of attention, leading him to start his own site in March, which he would name FiveThirtyEight after the number of votes in the electoral college. There, he would cover the election by following a playbook that he has continued to use ever since: using polls, adjusted according to ratings given to account for accuracy and bias, to forecast the electoral college. It would make sense to say that Silver had his big break by getting the election right and that everything grew from there. But Silver started receiving intense media attention before he had been tested in a single general election. In June 2008, months before the actual election, Silver made his first mainstream media appearance with an interview on CNN. Over the course of the election, he would appear on MSNBC, Fox News, and even The Colbert Report. That this much attention was being given to a formerly anonymous baseball writer with no experience covering politics or even a track record for his model speaks to the sheer size of the gap that existed in American political coverage—a gap Silver was eager to fill. While he was, and still is, relentlessly self-effacing about his “wonkishness”, it was that exact attention to detail and fealty to numbers that made him such a compelling figure in 2008. The state of the nation seemed to get worse by the day, and the stakes of the election seemed larger than ever. But when anyone, from average citizens to journalists themselves, tried to assess the actual state of the race, all they could find was the same interminable morass of hackery, spin, and amateurism that had been woefully inadequate for decades. Amid this din, Silver offered something that stood out: a trustworthy signal against the noise. His backstory fit with the times, too: the arc of creating a profile out of essentially nothing to provide a totally new method of political analysis was a perfect match with the newly pervasive cultural archetype of Silicon Valley “disruptors” changing their industries forever. What Steve Jobs did for computers and Mark Zuckerberg had done for social media, this man would do for politics. And it certainly didn’t hurt that Silver would go on to have a near-perfect record in the 2008 election. His website contrasted sharply with the general consensus for the entire election: while the media would treat the context between Obama and McCain as a competitive race, Silver would essentially call it for the Illinois Senator in March. State by state, his model would have the eventual winner favored 49 out of 50 states, with Indiana (a very hard call) being the only miss. This genuinely impressive feat would rapidly accelerate Silver’s budding profile: just five days after the election, The New York Times would write a lengthy profile of him titled “Finding Fame With a Prescient Call for Obama.” The Times detailed his life story from his early childhood, describing him as someone who “believed in numbers the way authors believe in words, as capable of expression and provocation, since he was young.” There were stories, like from his brother, who told of how he once announced “Today, I’m a numbers machine” as he was dropped off in preschool, and then proceeded to spend the entire day counting. There was his ability to multiply double-digit numbers at kindergarten, the multivariable regressions he ran at 11, his statistics-driven fantasy baseball team at 13, and the video game he created himself after his parents wouldn’t let him buy them. All of this combined to present an image of a preternatural, savant-like talent, capable of disrupting entire industries in what was essentially just his spare time. Who knew what he would be capable of next? It was here that Nate Silver, the myth, was born. Dizzy with Success At the end of their post-election profile of him, both the Times and Silver played coy about his future. He said that he expected for the spotlight to wear off with the election over and floated the idea of making a model to predict Congressional votes—although he was bearish on its potential. Voters, he snidely remarked, “don’t care about bills being passed.” Putting aside the elitism—a tendency worth keeping an eye on—Silver had keyed onto something: people really were interested in his election models, far more so than anything else he had worked on. Within the month, Silver would sign a publishing deal with Penguin Books to write two books, pulling in a $700,000 advance. By the start of the new year, he would be invited to speak at both a TED conference and at South by Southwest. Esquire hired him to write a monthly feature column, and he began occasionally writing for the New York Times and Wall Street Journal. And in recognition for his success at predicting 2008, he was named one of the 100 most influential people in the world by Time. Baseball was over. Although he would still cover sports from time to time, Silver would resign from his leadership positions at Baseball Prospectus and almost completely stopped writing articles on MLB. Politics was his future now. It was only a question of when, not if, a major outlet would pick him up. In mid-2010, he would find his suitor in the New York Times, signing a three-year contract with the site—a tenure that quite notably overlapped with the 2012 election. In this, Silver was betting on himself almost as much as the Times was betting on him. Making calls about the Democratic primary behind a pseudonym or setting up models on your own site was one thing, but now, he was plastering his predictions and his name right in the middle of America’s paper of record. There would be nowhere to hide if he slipped up. He had chosen to put himself at the center of the upcoming presidential race. He would start off with something of a test run with the 2010 midterms, an easy call if there ever was one. By the fall campaign, it was pretty much established Democrats would lose the House but keep the Senate because of the cushion provided by their massive majorities. This made it easy for Silver to continue his perfect streak, and he would even key in pretty closely to the final results far ahead of time, predicting in August that Democrats were on track to lose six to seven Senate seats (they would lose six). It was a feather in his cap, to be sure, but it wasn’t what he was there for. 2012 was always the real prize. As that race began, Silver would once again find himself benefitting from the low bar set by mainstream media coverage. Just as always, the 24-hour news cycle treated the race as hyper-competitive to keep interest high. But unlike in 2008, where journalists were forced to speculate on the likelihood of McCain making up a ten-point gap in Pennsylvania, 2012 would actually see close results in national polls between Obama and Romney. It seemed obvious to everyone that the contest was a tossup where anything could happen, with any given event in a news cycle holding the potential to decide the Presidency. Silver, once again, broke from this consensus, arguing that, once again, that Obama was being drastically underestimated. While national polling was relatively close, that’s not where the race was actually going to be decided. That was going to be in the country’s swing states: places like Colorado, Virginia, Ohio and Iowa, which were set to hold the balance of power in the electoral college. And in these states, Obama was holding solid, consistent leads across the board. As a result, Silver’s model was highly bullish on the President’s chances, as was he. As early as February of 2012, he would begin writing articles providing an optimistic picture of Obama’s odds. In order to explain why, he began to attempt to push forward concepts such as “tipping point states” in a bid to shift the focus of election coverage towards the points in the electoral system that determined the balance of power. All of this would utterly infuriate Republicans. After years of seeing Obama with mediocre approval and their massive success in the 2010 midterm elections, conservatives were coming into 2012 expecting to be in a strong position to defeat the President. It was something they wanted to do very, very badly. But now, well before election day, they had to see this self-admitted Obama supporter plastered across the New York Times politics section telling everyone that Romney barely had a chance—while being regarded as having the utmost credibility. It was not what they wanted to see at all. As Silver remained steadfast in his rating of the race, conservatives began fanatically attempting to disprove him by any means necessary. They attacked his models, called him a biased liberal hack, and would even attempt to “unskew” polling results that they thought didn’t give Romney an adequate level of support. Silver would not only completely ignore these critics, but would become emboldened by them. Over the course of the campaign, his model would give Obama increasingly favorable odds. By the morning of November 6th, it gave the President a 91% chance to win re-election, a rating that once again stood in stark contrast with the competitive race being presented on television. Once again, the die would be cast, this time with even higher stakes for Silver’s reputation. Who would be right: the forecaster or his conservative critics? It wouldn’t take long for there to be an answer. The race would be called for Obama before midnight on the East Coast—meaning that Silver had once again gotten the winner right. And a few days later, when all the votes were counted, and Florida was called for the President, he had accomplished something even more remarkable. He had called every single state right. Silver had gone 50 for 50. Mic Drop After 2008, it was still possible to dismiss him. It was easy to say that maybe it was a fluke, and he just got lucky, and that it wasn’t really a close election either way. But after 2012, nobody serious in politics could deny Nate Silver any longer. This was an election that practically everyone but him considered to be a tossup, and it turned out exactly as he said. Between the brightness of the spotlight, how much criticism he had received, and how loudly he had called his shot, it’s hard to imagine a more resounding vindication. Everyone was listening to him now. But if you looked closely, you could see that this success had planted seeds—seeds that would grow to make Silver’s task of keeping his new reputation extremely difficult, if not outright impossible. To start with, it’s important to understand what Silver’s models were actually doing. After 2008 and especially 2012, he had been credited as the guy who called 49/50 or 50/50 states correctly. But Silver never actually made calls in that sense. His election forecasts were, and have always been, probabilistic: merely representing the chance a given candidate had at winning a given race at a given time based on the information available in that moment. In practically every race people were following, he always gave both sides a possibility of winning. Even his 2012 model left open the possibility of a Romney victory. In that sense, his predictions were unfalsifiable. If the candidate he had favored won, then he simply got it right. If the underdog won, he was right, too: he gave them a chance, and unlikely things happen all the time, so if you’re saying it means he got it wrong you just don’t understand statistics. I don’t think presenting his forecasts this way should be controversial: it’s exactly how he’s done so himself, to the extent that he’s compared his recent forecasts to normal distributions. The second overlooked factor in Silver’s models is the degree to which he, for lack of better words, simply got lucky as to the quality of the data that was available to him in the 2008 and 2012 elections. The specific media environment that existed during the period facilitated the creation of a massive amount of high-quality, nonpartisan polls that proved to be remarkably accurate in both years. As a result, Silver’s only real task was transforming this data into probabilities, a mostly custodial job that consisted of choosing which polls to rate as good, rate as bad, or leave out entirely. Of course, he had proven himself remarkably adept at doing this job, but it also meant that his own role in the creation of his ratings was far smaller than assumed in popular media. The result of this was a massive gap between the public image of Silver— a maverick who drew on his innate genius to confidently nail election after election—and Silver in reality, who essentially just served as the chief custodian of an advanced polling aggregator that never actually made a single call. It’s very possible to become a celebrity off of the former image. The latter, not so much. And given the extent to which the conception of Silver in the media had been built on a combination of luck and misconceptions, it would have been very reasonable to worry how long it could sustain itself. This was hardly an overlooked possibility. Even on election night itself, there were some who attempted to pour cold water on those attempting to coronate Silver as the new king of politics, declaring that those doing so were making the “same mistake that Silver’s critics made last week,” having “confused his projected odds with hard-and-fast predictions, and underestimated the accuracy of polling.” They also pointed out that, beyond the high-profile presidential race, Silver had made some big misses in major Senate races. He gave Republicans a 92% chance to win a seat in North Dakota that they ended up losing, and even went against the polls to give them a 66% chance of winning the seat in Montana, which they also lost. Even at the heights of his triumph, these signs seemed to indicate that Silver was not quite the undefeated savant the media was making him out to be. The possibility seemed to have been high on Silver’s mind after the election, where he said in an online chat a week after the results that he felt the temptation to “pull a Jim Brown/Sandy Koufax and just mic-drop/retire from elections forecasting,” quitting while he was ahead. He dismissed that notion, however, and said that he planned to once again make forecasts in 2014 and 2016. He seemed to lack much enthusiasm for the former election, which he outright dismissed as “boring”: guaranteed to be a sweep for the out-of-power party like midterms always are. The next Presidential race, however, did catch his interest. “The 2016 G.O.P. Primary,” he declared, “seems almost certain to be epic.” This moment, and the decisions made after it, would prove to be absolutely crucial for how the upcoming decade would unfold for Silver. He was on the top of the world after the 2012 election, with everyone desperate to hear from the race’s second biggest winner on how he got it so right. He could have tempered their excitement, explaining the limits of his own role in his own forecasts, how he never technically made any calls, how much he relied on the collective polling industry getting it right. Instead, he played right into their mythical conception of him, taking full credit for “calls” as noncommittal as the 50.2% chance he gave for Obama to win Florida. There would never be a pained explanation as to why he didn’t technically get the election right, like how he explained after 2016 and 2022 that he didn’t get the election wrong. He was going all in, betting that he could fully sustain his new image as a clairvoyant mastermind. Earth Wizard Silver’s first major move in this regard would be in 2013, when he let his contract with the Times expire without renewal. The possibility that Silver could voluntarily leave a top position at one of the most prestigious outlets in the country speaks to just how well-known and well-regarded he had become in the aftermath of the 2012 election. He had become his own brand, no longer needing the support of legacy media to have a large audience. He would find his next home with ESPN, who would sign him a contract built around giving him almost total independence and flexibility with FiveThirtyEight. He would be his own editor-in-chief for his site, which would have its own brand and URL independent of anything else. He would get to choose his own staff, which would add up to 20 full-time journalists by the time the site launched in 2014. There, they would do whatever he desired in accordance with his vision, from “data-literate reporting” to original documentary films. As for elections, Silver would bring his model back for the 2014 midterms, where it successfully predicted that Republicans would take control of the Senate. Although their prediction record was hardly perfect—they drastically overestimated the chances for the independent challenger in Kansas, had favored a sitting Democratic Senator in North Carolina who ultimately ended up losing, and underestimated Republican strength in Iowa, Georgia, Louisiana, Arkansas and Virginia—the extent of the year’s overall Republican lean kept this from causing an incorrect topline forecast. Either way, it’s not like anyone was paying attention to the midterm, which would see historically low turnout. Perhaps the most notable thing about the race was a remarkably consistent and sometimes quite large polling error in favor of Democrats, but those sorts of issues were thought to be inevitable. Just one election could hardly be indicative of any sort of concerning, industry-wide trend. Of course not. In any case, a mere midterm election was hardly the sort of event the new FiveThirtyEight had been created for. That was the 2016 election, and the site was committed to covering it from beginning to end. They would start as early as 2014, analyzing all of the people the mainstream media had decided were real contenders for the Republican nomination. This, in and of itself, represented a major shift for the site. They had become famous entirely because of their models, but it’s obviously impossible to credibly model something as chaotic as a presidential primary over a year away. So, in the absence of that, how would FiveThirtyEight cover the primary? The answer would be simple: punditry. But it’s not as if they would provide blind analysis like all of those useless columnists. FiveThirtyEight had a gimmick: that it was endorsements, not the polls, that were really worth keeping track of. This gimmick would define how they viewed everything over the course of 2015. By tracking the attitudes of establishment figures, rather than the attitudes of the people, it necessarily moved them towards being very bullish towards establishment candidates, such as Hillary Clinton, and instinctually dismissive towards outsiders, like Bernie Sanders. This would lead the site to confidently publish gems like a declaration that if Sanders was to run, it would amount to “some plaudits on the left, but that’s about it.” And that was just the site’s coverage of the Democratic primary, which was seen as a lower-profile, more predetermined and far more boring of the two nomination contests. On the higher-profile and more chaotic Republican side, they would take a similar stance, predicting early on that the race would just be a contest between the likes of Scott Walker and Jeb Bush. And as for the joke candidate for them to confidently dismiss, they found their man: Donald Trump. Negative 10% Silver had built his career by making the mainstream media his foil. Assuming that the insular, data-deficient and sensational press was always wrong about something, he would take their stories, examine them, and provide a counter-narrative. In 2008 and 2012, this had worked perfectly for him, and for 2016, he was trying to do the exact same thing, starting as early as the primary. To him, Trump was an obvious opportunity for him to complete his hat trick against mainstream pundits. While pundits were discussing his chances seriously, it was obvious to FiveThirtyEight that Trump was doomed. He didn’t have the resources, elite support, or even polling numbers necessary to compete in what they imagined a modern major party primary to be. So, Silver and his staff started loudly, confidently, and repeatedly dismissing Trump’s chances of not only winning the nomination, but even being a contender at all. They dismissed him when he entered the race. They dismissed him as he surged in the polls. They said that this was nothing new, that “we’ve seen this show before,” that the only reason he got any attention was because “people have really short memories in politics”—unlike them, of course. To make their stance clearly, undeniably, empirically clear, they decided to write out what odds they thought each candidate, in each party, had in 2016 primaries. Silver would be the most generous to Trump, giving him an entire 2% chance of winning. His guest, Katherine Miller of Buzzfeed, would not be so kind, giving Trump a flat 0% chance of winning. And Harry Enten, one of Silver’s writers, would be even more emphatic, declaring that Trump somehow had a negative 10% chance of winning. That’s how doomed Trump was! His standing was so bad that it broke the laws of probability itself! To explain why they thought this way about the polling leader, their logic was simple: Trump’s lead was just because of the inordinate media coverage he had received since his announcement. This should be expected to only last for a month and a half, and that a month has already gone by. History would remember him as nothing more than another Herman Cain. But two weeks passed, and he was still ahead. Then came another month, and he was still ahead. Then another month. And then another. As the dates of the actual primaries kept coming closer and closer, Trump’s lead over the rest of the field would only continue to expand. But FiveThirtyEight would remain virulently opposed to not only the idea that he was a frontrunner, but that he even had a real chance at all. Speaking to a live audience in September, Silver would tell them to “calm down”, because Trump wouldn’t win the nomination. Their estimates of his chances would stay under 10% throughout 2015. But as the primaries drew closer, Silver would find himself increasingly alone in his insistence in Trump’s imminent doom. He would only double down. Silver started directing his takes on Trump directly at the media, taking an immensely arrogant tone that, as described by Dave Weigel, treated any journalism that was not data journalism as “inherently stupid and wrong.” To him, it wasn’t merely just that Trump was not going to win. It was that anybody who took Trump’s chances seriously was just “freaking out”—both stupid and crazy. Polls that had him ahead “weren’t news.” Anyone saying anything otherwise was committing journalistic malpractice: utterly ignorant about what they were talking about and doing their readers a disservice. Friendly figures tried to give him outs. He wouldn’t take them. When shoe-leather journalists who had taken Trump and Sanders seriously before practically anyone playfully asked him if he was ready to admit he was wrong, he would snidely retort, “It seems like you’re being pretty straw-manny and cherry-picky.” The battle lines could not have been drawn more clearly. It was all or nothing. Silver would either prove everyone wrong once again, or he would be off by more than he had ever been in his life. Then Donald Trump steamrolled through the entire Republican Party. And to make matters even worse, Bernie Sanders, the other outsider candidate FiveThirtyEight had been dismissing out of hand for a year, took Hillary Clinton down to the wire, winning 23 contests and 44% of the overall vote. This was not good for Nate Silver. In the aftermath, Silver and his colleagues could bring up the success of his actual models, explain the actual definition of probabilities, or point figures at the rest of the Trump-skeptical media all he wanted. At the end of the day, his site was now in a position that it was designed to never be in. No longer the savvy outsider, he was now the face of a class of out-of-touch pundits incapable of understanding something really important. They had done so badly that even their lauded model would end up with a worse prediction record than Carl Diggler, a fictional character meant to parody pundits by getting races wrong. When Trump officially clinched the nomination, he would be forced to write a long-form mea culpa admitting he had “screwed up.” While it was advertised as an admission that he had made a mistake by trying to act like a pundit, it mostly consisted of him complaining that it wasn’t fair to be mean to him for not being able to predict Trump, because predicting things is hard. He’d articulate this indignant attitude far more succinctly in a tweet posted at around the same time, wherein he whined that people should stop judging him just because he “didn’t predict the Republican Party would lose it’s fucking mind.” But no matter how much ink Silver spilled about how mean and stupid everyone was, one singular fact remained. He had been caught completely out of his element when tasked with doing the one thing he was supposed to be the best in the world at. If he and the other writers of FiveThirtyEight expected to use the reputation they had gained in 2008 and 2012 to be able to take consequence-free risks as amateur pundits, they were wrong. No number of articles and tweets saying “Ackchyually, that’s not how probabilistic forecasting works” could change that. The reverence the site had once enjoyed was replaced with a harsh, critical eye, eager to render an indictment not just on Silver, but on the entire school of journalism he had pioneered. Maybe this wasn’t right. Perhaps it was too presumptuous for everyone to suddenly take his success in 2008 and 2012 for granted. Maybe that the entitled, what-have-you-done-for-me-lately attitude from his critics was unfair. But, at the end of the day, nobody had put Silver in this position but himself. Nobody forced Silver to play into an image of himself as an infallible savant. Nobody forced him and his staff to recklessly play pundit throughout the primary. Nobody had forced them to treat everyone else like ignorant children for disagreeing with them. They all did that themselves. And it’s not even that the calls they had missed were even all that hard. Both Sanders and Trump had their polling surges as early as the summer of 2015. It was entirely reasonable to treat them like serious candidates. Especially so as they maintained their support for months on end. Everyone else in the media did, at least to some degree. But FiveThirtyEight simply refused to do so. They emphatically refused to do so. And now, the site was back at square one. Silver and co. would once again be forced to nail a Presidential election to silence a vocal cast of doubters. Bug It was a very hard task. And here, I’ll say something that not many who are critical of FiveThirtyEight are willing to say about their performance in the 2016 general election: I think they handled it admirably. Straight up. If you think election modeling can be useful, and it can be, there is very little to criticize about how FiveThirtyEight handled the 2016 general election. They did terribly in the primary, absolutely. But they handled the actual election itself fine. Yes, they got the results wrong. Very famously so. They projected that Clinton would win, and she lost. They had her favored in five states (and a congressional district) that ended up going for Trump. To this day, it remains their least accurate forecast. But all of this must be understood in the context of what their model actually is. It is only as good as the data it has to work with, as Silver loves to point out. And in 2016, the data was very poor. There was a systematic error in Clinton’s favor at practically every level. No reasonable person working with the information available at the time of the election could have said that she was anything but ahead of Trump. And as far as FiveThirtyEight was actually involved with their model through their role of selecting which polls to include in it, there isn’t any evidence that they neglected obvious signs that the polls they were including were somehow overly biased towards Democrats. In fact, if there’s any argument to be made in that regard, it's that their decision making was slanted in favor of Republicans. The site still happily included numbers even from openly partisan firms like Rasmussen, with only minor adjustments. But even that didn’t really make a difference. That’s just how off the polls were that year. The end result was about the best you could get for Trump without just outright tilting numbers in his favor. Their final model gave Clinton a 71% chance to win, meaningfully leaving open the possibility of a Trump victory. And it was hardly the case that anyone at the time considered these numbers to be too bullish for Democrats: on the contrary, Silver and his site would receive substantial criticism from liberals and others who believed that his odds were too good for Trump. They were angry that Silver was giving Trump a chance at all, seeing it as cowardly hedging he was only doing to make up for his infamous coverage of the primary. Copycat sites with their own models began competing with each other to contrast the most with FiveThirtyEight by giving Clinton the best odds. The Daily Kos gave her a 92% chance of winning. The Huffington Post gave her a 98% chance. The most bullish of all, however, would be Dr. Sam Wang of Princeton, who outright called the race for Clinton in mid-October and promised to eat a bug if Trump even managed to receive over 240 electoral votes. (If you’re wondering, Wang would make good on his promise, eating a cricket live on CNN.) Against these headwinds, Silver ended up spending much more time before the election arguing against the candidate he had favored in his own model. And in contrast to his shambolic attempts at punditry in the primaries, many of Silver’s arguments in this closing stage of the election would end up being spot-on. He pointed out that Clinton’s lead had dwindled to within range of a normal polling error. He made the distinction that her strength in the popular vote didn’t necessarily mean strength in the electoral college. And he was right! It was exactly what happened. It really was Silver and his team at their best, using their strengths with numbers and data to disprove weak arguments. But nobody cared. Those subtleties didn’t matter. People cared about what FightThirtyEight’s model said, not what other site’s models said or the articles they had written ahead of time. They were supposed to be the gold standard and their model said that Trump would lose. Because of this, their reputation would never recover. Maybe it could have if the site had not so loudly dismissed Trump early on, but they did. Now, it was impossible for people to forget. The very purpose of the site, and data journalism in general, was now being questioned. Silver had fashioned himself as someone who could see what the masses couldn’t: capable of picking up, as per the title of his book, the signal from the noise. But when he faced the sort of test he should have been made for, he failed to see it coming over and over again. Was this an unfair reading of FiveThirtyEight given the impossibility of their task at hand and the emphasis they gave to Trump’s strength before the election? Maybe. But Silver was very far from blameless here. After his success in 2012, he took a bet that he was smarter than everyone else and intensified the risk at every opportunity. He fully played into the story the messianic image had created around him. He took a condescending and adversarial tone towards his peers in the press. He set the expectations around himself impossibly high, left himself no margin for error, and made an entire industry salivate at the chance of knocking him down a peg. He thought that he could live by the sword, but not die by it. And then he got the 2016 election wrong. Twice. His bet just didn’t pay off. It’s hard to feel much sympathy for that. Continued in Part II