Oscar odds: Bet safe, win small. Bet crazy? Put your money away.

So you just know Joaquin Phoenix is going to win the Oscar for his work in “Joker”? Well, if you’re right, that knowledge and $100 in Vegas will get you … about $3-$5, depending on the betting house.

“Our primary goal isn’t to be right; it’s to balance our financial risk,” said James Murphy, oddsmaker at SportsInsider.com. Murphy, who said he has minimal knowledge of some Oscar categories, like costume design, for instance, said during a recent interview that he was still setting “Knives Out” as the favorite to win the Costume Designers’ Guild Award for contemporary film; later that night, it did just that.

“We’re trying to reflect the marketplace. There’s buy low, sell high. Squares, or recreational bettors, bet teams, pros bet prices.”

In other words, it’s not so much the sure things where bettors make money; it’s longshots and upsets — within reason. After all, there’s a reason the odds for Antonio Banderas or Jonathan Pryce to upset Phoenix are set at +2500 (win $2,500 on a $100 bet): It’s not likely to happen.

Murphy has set Phoenix as a huge favorite at -2000. That means a $100 bet would net you only $5 in winnings. If that sounds unreasonable, veteran oddsmaker Johnny Avello has Phoenix at 1-25 in fractional odds, which would net $4 on a $100 bet. Oddschecker.com is even stingier: -3300, or $3 on that same wager.

“From a bookmaking standpoint, you almost have to make him a huge favorite since given the predominately male demographics of the sports betting industry. ‘Joker’ should get a ton of action in every category,” Murphy said.

Murphy has been involved in professional gambling for more than 25 years, his love of the Portland Trail Blazers being his gateway drug. Along the way, he has set odds on everything from how cloud cover would affect the visibility of an eclipse to “who would be killed next on ‘Game of Thrones.’ That did really well.” He says, though without verification, he’s “around 70%” with his Oscar predictions over the last “10-20 years.”

When calculating his odds, Murphy looks for every factor he can find, not just what the guilds choose, fans want, or critics say.

“You figure in the size of the voting body,” he says. “The CDG is 800; the DGA is like 18,000. Smaller group equals larger volatility.”

He looks at front-runner fatigue (anyone remember when “The Irishman” was atop all the prognosticators’ lists? Or “La La Land”?) and has his own theories about why, say, Julia Louis-Dreyfus didn’t win an Emmy for the final season of “Veep” and Tom Hanks hasn’t won an Oscar in 25 years.

“The Emmys — they’re very protective of the history of television. Julia Louis-Dreyfus, maybe they don’t want to put her on the level of Lucille Ball,” he says. “Tom Hanks has two; maybe they don’t want him to tie the record,” which is three (Daniel Day-Lewis, Jack Nicholson and Walter Brennan each have three, but Day-Lewis is the only man with three lead-acting Oscars; Hanks has two for leads).

If you insist on betting money on the Oscars, you might consider what he says separates recreational bettors from the pros — betting the “prices.” If you won a $100 bet on “1917” to win best picture at Murphy’s number of -175, you’d collect $57.10. But investing in an upset — say, putting down $100 on “Once Upon a Time … in Hollywood” at +350 — could win you $350.

“1917″ director Sam Mendes has emerged as the director favorite (-750) following his DGA win, but Murphy isn’t willing to completely wave the white flag on two of the other candidates.

“Still a lot of support for [‘Parasite’s’] Bong [Joon Ho at +450] and [‘Once Upon a Time’s’ Quentin] Tarantino [at +750] in the market, and their appeal among the ‘new blood’ [among voters] makes them attractive to bettors, so I’m keeping the prices down on them just as a precautionary measure.”

Murphy’s longest odds in the main Oscar categories fall to “Little Women” or “Ford v Ferrari” winning best picture (+5000). That means you could win $5,000, or 50 times your money, on a correct $100 wager.

Or you could just set a C-note on fire.