Back in mid December I warbled like Admiral Ackbar that I thought the polls showing Donald Trump so shockingly ahead of Joe Biden in key swing states were a trap. I explained why using the issue of “registered” vs. “likely” voter screens, a concept that haunts me in my continuing analysis of (and perplexity at) the 2024 presidential race. So, as a predicate:

On its surface, you might be excused for thinking [today’s NYT/Siena poll] another great one for Trump: He nominally leads in every swing state except Wisconsin and is up 46–44 among registered voters over Biden. Except here’s the thing: For the first time, NYT/Siena has applied a “likely voter” screen to its numbers as well (basically, if you voted in 2020, you’re a likely voter), and when that happens, the numbers perfectly invert: Biden goes from losing by 2 percent to leading by 2 percent (47–45) nationwide. As the Times’ Nate Cohn points out, this is down to the staggering polling delta between those who voted in 2020 (Biden +6) and those who didn’t (Trump +22). Those nonvoters include many of the young voters whom Biden has been so infamously hurting in recent months, who seem like quite the difficult “get” for Donald Trump given that their disaffection with Biden stems from those aspects of his policies that aren’t progressive enough. The LV number shows clearly enough that no matter how much people might support you, it doesn’t mean anything if they don’t actually vote. I am skeptical they will.

I am significantly less confident in my final contention now, with the arrival of the latest New York Times/Siena polls, which raise the matter afresh. First of all, let us summarize this poll of both the national electorate and various swing states: Boy does Joe Biden not want to be the poor bastard on the receiving end of these numbers. As Jim Geraghty aptly points out this morning, at a whopping 50–38 Trump, Nevada honestly no longer looks too terribly swingy anymore as a state, although Lucy and the Culinary Union have yanked the football away from the GOP too many times to quite believe it. The registered-voter numbers show him leading in Pennsylvania 47–44, Arizona 49–42, Michigan 49–42, and Georgia 49–39. Only in Wisconsin does he trail, by two. It’s mid May. A little more than five months remain before the election.

At this point it’s worth pausing to note that many of us in the media have been in a bit of a daze about all of this. Sean Trende of RealClearPolitics once wrote eloquently about the lessons of 2016: about how election analysts were blindsided by Trump’s victory, despite its clear plausibility given polling, because they suffered from “unthinkability bias.” After all, Trump wasn’t actually going to be elected president, was he? Because c’mon now — that would be totally insane. So, far too many otherwise intelligent people conjured extracurricular explanations for why the data “will be wrong this time” or “things are different now” rather than believe the clear import of what the numbers in front of their faces were telling them, which is that Trump had a roughly one-in-four chance of winning the 2016 race. “So Hillary has a three-out-of-four chance? Then the race seems over to me. (Let’s not worry about that 25 percent, because that’s obviously not going to happen, seriously now people you’re scaring me.)”

And then of course Trump pulled an inside straight on Election Night.

This time around, Trump does not need an inside straight. He leads handily in all or most swing states, has done so for over a year, and still nobody is truly hitting the panic button quite the way you would expect them to. Even as poll after poll has come in showing Trump securely leading Biden, the response has been muted warning or caution verging on unstated disbelief. Some continue to insist the race will not truly begin until after the conventions.

I’m not so sure. The old adage that “the voters don’t start paying attention until Labor Day” may just be that: an old adage. Does it apply in situations where the voters already made up their minds long ago? Does it furthermore still apply in a technological and media ecosystem where politics are shoveled down American throats regardless of personal desires? And most specifically, does it apply to a 2024 campaign where there will be no debates, no face-to-face confrontations, and no predictable inflection points beyond the party conventions?

(That brings me to an important but parenthetical aside: Right now each team thinks it holds a potential trump card — the term, alas, cannot be avoided — in events yet to come. Democrats are counting on a conviction in Trump’s hush-money case sometime in the summer, harming his numbers enough marginally in the polls to squeak it out in key states. Meanwhile Republicans anticipate — oftentimes with frankly appalling glee — the August Democratic convention turning into Chicago’s Greatest Hits: To the Kneecap and Skull, Vol. II. With any luck, both parties will be profoundly disappointed — but if both were to happen, or even only the Democrats’ preferred scenario alone, there is now real cause to doubt whether Dems would truly gain from events.)

The real reason for doubting Trump’s otherwise sure-seeming victory always hid inside the question, Who will actually show up to vote on Election Day? (Especially in a nation that increasingly employs mail-in balloting and a candidate explicitly associated with vehement personal opposition to it.) In the earlier Times poll, the delta between registered voters (who favored Trump) and likely voters (who narrowly favored Biden) raised a significant red flag and gave Biden’s supporters a hope they could cling to: It suggested that when the time came to vote, their people would actually get to the polls.

Today’s Times/Siena poll offers no such comfort. This time, when the screen is shifted from registered to likely voters, the picture barely changes. Head to head, Trump still leads nationally 49–43, and in Nevada his numbers remain equally strong, at 51–38 — will Charlie Brown finally have his day in Clark County? In Wisconsin (where he trailed among registered voters, notably) he now leads 47–46. (Meanwhile Biden pulls the reverse trick in Michigan, vaulting from a seven-point deficit to a 47–46 lead.)

My interpretation is this: This poll portends doom for Joe Biden. As things currently stand, he is facing worse prospects than Donald Trump did in 2016 against Hillary Clinton. Again, five months of campaigning (or rather, “campaigning”) remain, so ask me again in October. But these numbers cannot be waved away, and now point toward a Trump victory. The likely-voter screen does not save Biden anymore, and in any event the critical issue of what the likely-voter model in 2024 looks like is left unaddressed. (The standard metric is voter participation in the previous two cycles — but 2022 and 2020 were two of the most abnormal electoral cycles in living American memory.)

In 2016, from the surprisingly strong (albeit structurally weaker) position he was in, Trump had to pull an inside straight to win the presidency. Biden currently needs an inside straight-flush. Things may change — remember those pocket aces both parties believe themselves to be holding — but for now these numbers suggest to me that, were the election held today, Trump would win every single one of these states, Michigan and Wisconsin included.

Does any of this sound like special pleading ? Or the bleat of a man sick in the throes of his own “unthinkability bias”? I hope it is the opposite: My earlier assumptions have been undermined, and new data have superseded old data, forcing a reconsideration. Maybe I will have cause to reconsider yet again, and in any event the race will look much clearer come September, but for now, Donald Trump holds all the cards, and Joe Biden is praying for a miracle, floating the river with the weakest electoral hand of any incumbent since George H. W. Bush.