Grok 4 Predicts Dodgers for World Collection Win—However Different AIs Aren't So Certain – Decrypt

Among the many demos Elon Musk confirmed off throughout Grok 4’s launch on July 9 was a banger asking the AI to foretell which crew will win Main League Baseball’s World Collection later this 12 months.

After 4.5 minutes of number-crunching that analyzed information from Polymarket, the Ethereum-based prediction markets platform, and utilizing what xAI calls its “Heavy” reasoning capabilities, Grok 4 delivered its verdict: The Los Angeles Dodgers are the almost definitely crew to win the 2025 World Collection. Grok gave L.A. a 21.6% probability to win all of it—larger than some other crew, however nonetheless famous they may be overpriced.

Grok’s predictions are definitely consistent with different main platforms, together with ESPN BET, which reveals the Dodgers sitting at +225 because the MLB season approaches the All-Star break. The Detroit Tigers (+750), who’re working away with the AL Central, have emerged as a darkish horse contender with baseball’s greatest document at 59-35.

Merchants on X are giddy in regards to the potential of getting a private Grokstradamus and calling the outcomes an “infinite cash glitch.”

However we wished to know: Did the opposite main AI fashions agree with Grok?

Seems, not totally.

What different AIs assume

ChatGPT’s o3 mannequin gave the Dodgers a 26% probability whereas flagging them as overpriced. The mannequin recognized Detroit as providing the very best worth with a 16% win chance in opposition to market odds implying simply 12.5%. Its reasoning centered on Tigers ace Tarik Skubal’s dominance and the crew’s league-best pitching workers.

DeepSeek doubled down on Los Angeles with a 23% chance, however famous the Dodgers may be using an excessive amount of constructive sentiment. Regardless of favoring LA to win, the mannequin stated it might slightly wager on the Phillies as a result of the risk-to-reward ratio was extra compelling.

Since we’re poor and our paymasters have been unlikely to approve Grok 4 Heavy’s $300 subscription for only one query, we requested the lighter Grok 4 model obtainable by way of the $30 tier. Apparently, it gave the Tigers a razor-thin edge over the Dodgers—lower than one share level separated their odds.

All three fashions flagged related components: Detroit’s elite pitching rotation, the Dodgers’ harm issues, and historic patterns suggesting the market overvalues defending champions.

It is all within the immediate

Whereas Grok 4’s “Heavy” reasoning is spectacular, you don’t want a $300/month plan to get strong predictions. With good prompting, even fundamental fashions can ship sharp insights. We discovered that profitable prompts want at the very least these three most important parts:

First, role-play. Inform the mannequin who it must be and how it ought to act. Attempt one thing like: “You’re an professional Prediction Market Analyst with deep data of Bayesian forecasting and threat administration.”

Second, the methodology: Inform the mannequin what you need and what steps it ought to comply with with a purpose to succeed. Ask the mannequin to collect present betting odds from a number of sources, examine them in opposition to analytical projections, and determine worth bets. Fashions carry out higher after they can examine market consensus in opposition to their very own calculations.

That is what immediate engineers name Chain-of-Thought—if the mannequin is aware of precisely what to do, it offers higher outcomes. Do not know easy methods to information it? Ask the mannequin individually for the steps wanted to finish your job.

Third, level towards analytical sources. Mentioning Baseball-Reference simulations or FanGraphs projections helps floor predictions in established frameworks, slightly than pure hypothesis.

For these taken with making an attempt this themselves, we constructed a customized GPT that replicates what xAI demonstrated with Grok 4. It was only a enjoyable experiment, but it surely gathers odds, analyzes crew efficiency, and identifies potential betting worth via pure dialog.

We additionally tossed our prediction market immediate on GitHub if you wish to experiment with your individual chatbot.

Use at your individual threat, naturally. We’re not monetary advisors, and neither are these AIs. Should you lose, do not blame us—but when it helps you win massive, then we cannot say no to a beer.