It's gospel among Trump supporters that public polls systematically understate the millions of Americans who will vote for the President on Nov. 3. If that's the case, then the probability President Trump will prevail should advance to 50-50, or even show him in the lead, if he's still trailing but pulls within a couple of points of Joe Biden.

One way to test the "if Trump gets close he wins" hypothesis is to examine whether the odds on political betting sites do indeed rate the contest a toss-up when Trump goes from lagging by a furlong to within a few meters, turning what looks like a romp into a horse race. That data could provide a guide to how much of the current gap the President must close to get a good shot at victory on Nov. 3.

The odds offered on the betting sites are arguably the best guide to handicapping the outcome. That's because those odds are set by people putting "skin in the game" by risking their own money—including professional political gamblers who carefully study voting patterns and shifts in momentum. There's no better estimate of Notre Dame's or the Dallas Cowboys' chances of beating a point spread than the odds posted at the casinos or online digital sports betting venues. The gamblers can be wrong in presidential elections, as in 2016, but in general, the dollars shifting from blue to red and back should be an excellent bellwether for where voters are headed.

Gambling on U.S. elections is mostly banned stateside. The only major site approved to run a political futures market in the U.S. is PredictIt. But sundry online gambling sites in Europe are offering lines on the contest for the White House. They include Betfair, Bovada, Smarkets, and Spreadex, all of the U.K.; bwin of Austria, Betsson of Sweden, and Unibet of Austria. In recent weeks, the action's been terrific. Paul Krishnamurty, a professional gambler and freelance writer who pens insightful analysis on presidential gaming, notes in a recent post that Betfair saw over $25 million wagered on U.S. elections the week ended Oct. 16, equivalent to 15% of the total placed over the past four years.

RealClear Politics publishes a daily average of nine of the latest postings from the major sites. (PredictIt isn't included in that measure; more on that shortly.) For this analysis, I'm using that data alongside the RealClear Politics average of polls for the five swing states that most experts reckon will decide the election: Pennsylvania, Florida, Wisconsin, North Carolina, and Arizona. So let's look at how our leading indicator—those betting odds—have moved as Trump's poll numbers have waxed and waned, and what that tells us about the gap he must shrink to stand a chance on Election Day.

Ups and downs

From July 21 to Aug. 23, Biden led Trump by substantial margins of between 4 and 6 points in the swing states. Trump's odds of victory, as shown on the betting sites, toggled between 36.1% and 43.4%. In other words, the President looked like a loser, as he trailed Biden's odds by at least 16 percentage points during that monthlong interlude. But starting the last week in August, Trump went on a tear. The virtual Republican convention hosted from Aug. 24 to 27 appeared to give Trump a big boost. By Aug. 29, he'd narrowed Biden's lead from 4 points two weeks before to 2.7, and he clung that close through Sept. 1. On that date, Trump posted his best odds of the entire election cycle, as measured by betting sites, hitting 49.5% and pulling within 0.5% of the former vice president.

The bump was short-lived. By Sept. 10, Biden had regained a 3.9-point edge in the polls, and he maintained a lead in the high-3s and 4s before peaking at a 5-point advantage on Oct. 11. The rise in Trump's polling deficit from 2.7 to 5 points mirrored a steep drop in his odds, which cratered from the 49.5% summit at the start of September to 34.7%. That made Biden a two-to-one favorite to take the White House.

The road to the White House took still another surprise turn on Oct. 12. Trump emerged from his bout with COVID sporting an air of triumph and jumped back on the campaign trail, hosting a raucous rally that evening in Stanford, Fla. The sight of the President healthy and recharged for combat gave his poll numbers their second big lift in seven weeks. By Oct. 19, Trump had cut Biden's swing-state lead from 5 points to 3.9. Trump's odds followed his polls, jumping from the late Oct. 11 near-low of 34.7%, to 40.3%. (They shrank to 38.7% the next day, but the polling gap held at 3.9%.)

It's perhaps prophetic that Trump's ups and down in the polls appear to trigger more or less predictable gains and losses in his odds. The rise in his polling gap of 2.3 points, from 2.7% to 5%, between Sept. 1 and Oct. 11 caused his odds to retreat from 49.5% to 34.7%, or 14.8 points. For every 1 point his deficit grew, his odds dropped by 6.4 points. Similarly, his just-over-1-point move from a 5% to 3.9% deficit in the following eight days brought a rise in his betting line from 34.7% to 40.3%, or 5.6 points. So it's likely that a 1-point shrinkage in Trump's swing-state polling gap increases his probability of winning around 6 points.

Shrinking the edge

Trump's odds of clinching reelection now stand at around 40%. By our formula, to gain the extra 10-plus points needed to take the lead on the betting sites, he would need to shrink Biden's edge at the polls by around 1.6 points, taking him from 3.9 to 2.3 points behind. Keep in mind that on Sept. 1, Trump's odds trailed Biden's by just a hair when he lagged by his best reading ever of 2.7 points. It would appear that Trump needs to pull within the mid-2s by Election Day to pass Biden and win by a nose. That is, if the bettors are right.

Of course, squeezing the gap to 2.5 points still assumes that the polls are way off, and that millions of voters are still too embarrassed to tell neighbors and pollsters they're for Trump. Two highly respected prognosticators, Nate Silver's FiveThirtyEight and The Economist, put Biden's likelihood of victory at 80%-plus, and they reckon that the gamblers on the betting sites are now far overestimating Trump's fast-fading chances.

The odds on PredictIt show that Trump has gained recently in several states. He's regained the lead in Florida that he lost on Oct. 1, and North Carolina has moved to even money despite a 2% Biden lead in the polls. A recent jump has given the President a commanding lead in PredictIt's futures market in Ohio, and his deficits are narrowing in Arizona and Wisconsin. Nevertheless, those wagering on PredictIt still give Trump just a 35% probability of winning, little changed in recent weeks. That's several points below his standing in the RealClear Politics readings.

Still, you need to give a lot of credence to folks who put their money on the line. In fact, you've got to be an oddsmaker to decide who's closer, the bettors or the experts like Silver and The Economist. My vote's with the gamblers—who still give Trump a far less than 50-50 chance of winning. They just think getting close, though even that's a long shot, may be enough to keep Trump in the White House.