Here we go again: Donald Trump prepares to return to the Oval Office and once again showcases how “unpredictable” he is, convinced that playing “crazy” makes him strong. That force, he and his followers insist, will intimidate adversaries from Russia to China, Iran or North Korea. He will even discipline America’s friends if they misbehave. Strength through madness and peace through strength.

Asked if he would use military force against China if the latter threatens Taiwan, Trump replied: “He won’t have to, because he respects me and he knows I’m crazy.” That, in this case, is Chinese President Xi Jinping.

Trump makes the same case for Vladimir Putin. The Russian president would never have dared to invade Ukraine if Trump was in the White House, the incoming president is convinced. That’s because he remembers warning Putin that “‘I’m going to hit you so hard, you’re going to stop living in your world.’ Peace is near, Trump claims.

This projects to all politicians around the world, even to America’s best friends. In his first term, while discussing trade deals with South Korea, Trump instructed his negotiator to “tell them if they don’t make concessions now, this lunatic is going to walk out of the deal.” This time, Trump is unleashing more serious threats, in the form of devastating tariffs on imports from Canada, Mexico, China and other nations, regardless of economists’ warnings of massive damage to American consumers and businesses. Trump is the “bad guy” and the others will back down before he even has to do anything.

This so-called “madman theory” of conflict and competition has a long pedigree. Machiavelli was probably not the first to observe that “at times it is very wise to play the madman.” The nuclear age was chilling. In 1959, Daniel Ellsberg, a general later famous for releasing the Pentagon Papers, argued that a “blackmailer” could coerce opponents by acting “crazy.” Thomas Schelling, who later won the Nobel Prize for game theory, also concluded that in certain cases pretending to be insane is beneficial.

The first (until Trump) president to apply the theory was Richard Nixon. “I call it the Madness Theory,” he told his chief of staff: “I want the North Vietnamese to believe that I went so far as to do anything to stop the war. We’ll just tell them that “for God’s sake, you know Nixon is obsessed with communism. We can’t contain him when he’s angry – and he’s got his finger on the nuke button.”

Yet with Nixon as with Trump, there was always an obvious irony: Both rationalized his absurdity, which hardly made them credible. This is perhaps one reason why their theory has mostly failed, at least so far.

When Nixon launched a global nuclear alert in 1969, neither the North Vietnamese nor the Soviets were alarmed, and the war continued. After Trump first threatened North Korea with “fire and fury the likes of which the world has never seen” and then exchanged “letters of friendship” with its dictator, Kim Jong Un accelerated rather than halted his nuclear weapons program.

Spurred on by Trump’s first term, academics have taken another look at the madman theory. The bottom line was that the theory “doesn’t work” or is a “myth”. Some analysts, however, now believe that both the first generation of analysts (Ellsberg and Schelling) and the second missed a “window” that could help us explain a second Trump term.

Let’s explain madman theory with a simple game theory experiment, called “chicken”. (It’s a variation of another game called Prisoner’s Dilemma). In this crash model, two drivers drive their cars toward each other in a head-on collision course. The one who turns first loses, the other wins. If they both veer off course, it’s a tie, and if neither turns, they crash.

For Trump facing Putin, Xi or Kim, say, the math says:

One way to win would actually be to pretend to be insane. Trump could, for example, rip out the steering wheel and throw it out the window. Even then, he still won’t know if his opponent is sane or insane. And if he survives this game, he will have more rounds ahead of him against the other opponents watching him.

That means his success — his power — depends on how people interpret Trump’s madness. We add to the analysis the theory of Rosanne McManus at Penn State, who has interpreted the theory of the insane by distinguishing four types of madness. Then we simplified it as follows:

Some leaders take enormous but still predictable risks only in specific contexts and for specific purposes. McManus calls them “fanatics.” As an example he cites Adolf Hitler during the crisis of 1938, when the British retreated, concluding that the Sudetenland was worth much more to him than to them. Predictable bigotry can be effective.

A leader who is predictably extreme in any confrontation, due to mood, is a “megalomaniac” according to McManus. He mentions Iraqi dictator Saddam Hussein and points out that this rarely ends well for the bigot (because enemies, in this case the US, will conclude, rightly or wrongly, that they have no choice but to go to war late or quickly anyway ).

A leader who explodes only on certain occasions but then becomes completely unpredictable is a “hothead”. McManus points to Soviet leader Nikita Khrushchev during the Berlin Crisis of the early Cold War. This guy could win at some points based on theory but he could also lose (Khrushchev eventually backed down).

And then there are those who are really unpredictable all the time because that’s their nature. McManus cites Muammar Gaddafi in this category, the former Libyan leader who met a particularly gruesome end. This kind of madness is a recipe for disaster.

What about today’s leaders? Putin, even when issuing nuclear threats, seems to fall into the fanatic category: extreme but still predictable, because he is focused on not losing in Ukraine (as opposed to Syria, say). The same probably applies to Xi, who would not be drawn into a destructive dominance game, but only if the battle was over Taiwan. Both seem to have context in madness, which according to McManus makes them strong in these particular situations.

North Korea’s Kim Jong Un, like his father and grandfather, is more difficult to interpret, but comes across as either bigoted or hotheaded—that is, either predictably extreme in his hostility toward South Korea and the U.S. or unpredictable in any other crisis. This makes the Korean peninsula more dangerous than today’s headlines suggest.

But what about Trump? He goes on to emphasize his inherent unpredictability. Or he could fall into the hot-headed category, since he also fancies himself a gifted dealer with a transactional approach that discriminates between situations.

Either way, nothing in modern madman theory suggests that he will actually be “powerful”. One problem with “inherent” madness is that opponents conclude that it could never be satisfied, would always have complaints even if they deviated from the first round of game theory.

Another problem is that America’s allies are also watching these rounds. If they can no longer rely on the US because it has been unpredictable for years, they will take measures for their security. South Korea, Japan, Poland, or Germany might build their own nukes and form axes with other powers, including those hostile to America.

This has hurt Trump’s credibility. John Bolton, one of Trump’s last national security advisers, argues that Trump wouldn’t necessarily win many or even any rounds because his power and madness are mere threats, when in fact he has “no principles, so it just threatens unpredictably.”

Trump could of course fix this by rationalizing his goals, beliefs and red lines so that, over time, friends will learn to trust the US again and enemies will fear crossing it. Then he would no longer be unpredictable, nor crazy, nor even a strong man.

Andreas Kluth is a Bloomberg Opinion columnist covering US diplomacy, national security and geopolitical developments.