Last November, Kalshi ran a trader spotlight on someone who'd made half a million dollars on music and movie markets. No Ivy League degree, no network, he just figured it out. Eight months later, a Netflix documentary called the same trader "The Whistleblower" after he flagged what he believed was fake Spotify streaming used to swing a Kalshi market.
He's Caleb Davies, a dad of three in Minnesota with a full-time IT job, who trades as Gaeten Dugas. He built his edge in markets most traders ignore: Spotify charts, Rotten Tomatoes scores and, lately, the midterms, where he has about a quarter million dollars on the line.
Justin Herzig sat down with Caleb to talk about how he got started, the Spotify episode and what it changed about how he manages risk, how he models Rotten Tomatoes, how he's trading the midterms, and the mistakes he sees beginners make over and over.
Watch and Listen to The Interview Here - Read Below
From PredictIt to Kalshi
Justin: How did you get here? What has your journey been like?
Caleb: "I originally had a pretty deep interest in politics. I'd sit and debate people all day, which, looking back, was a tremendous waste of time. I took that interest and moved it into prediction markets.
"Back in October 2015, I saw an ad asking whether Joe Biden was going to run for president. Hillary Clinton was the candidate. It was her turn after getting beat by Obama in 2008, and she and Biden occupied the same lane. So I figured he wasn't going to run. I deposited about $300 and won.
"At the time, PredictIt was only politics, and politics on a day-to-day basis isn't that interesting. Elections are huge, sometimes there are legislative fights, but there isn't much to do every single day. It wasn't a full-time hobby. I'd actually withdraw my money when it wasn't interesting, which was dumb, because PredictIt had a 5% withdrawal fee, so I kept eating into my winnings. Over time, though, I was making five figures a year just doing politics.
"When Kalshi and Polymarket came on board, two big things changed. First, they offered a ton of topics. Instead of just politics, you could expand into movies, music or whatever interested you, and that runs year-round without the political cycle. Second, they don't have limits. On PredictIt, the most you could bet on a single market was $850 per bracket, which set a ceiling on what you could earn. That ceiling was gone. So I expanded into more topics, bet more money, and that's how I got here."
Justin: What would you say your edge is? Your IT background, modeling, data, gut?
Caleb: "A little bit of all of that. On politics, it's mostly looking at incentives, figuring out why people are doing what they're doing and what that means for their next step. A lot of that is just following politics for a long time. You'll often see a House vote look like it's going to fail, but if you've followed politics long enough, you know that's showmanship. People who normally don't get attention are taking advantage of their one opportunity, and it's ultimately going to pass.
"On elections, Rotten Tomatoes and Spotify, it's a lot more math. I don't have an extensive math background. I've got finance and business degrees, so I studied some, but it wasn't a focus. The most important thing is a good understanding of statistics. And you have to be a pretty smart person. You'll see a lot of traders do the 'aw shucks, I'm so dumb' thing, but if you're consistently making money, you're looking at the same markets as everyone else and doing better than average over a long period. That takes some smarts."
Justin: Are you using AI to build tools or analyze data?
Caleb: "Almost all of my analysis is done in Claude now. That's a change I made this year. I upload the data and ask it questions, so I don't need to figure out formulas or queries. It's been a huge time saver. My morning routine used to take quite a while, and now it takes a couple of minutes."
Justin: Do you think you're the exception, or can people with full-time jobs be very profitable in prediction markets?
Caleb: "I'm probably the exception, and several things work in my favor. My job isn't super demanding. I'm senior level in IT, and the little unspoken secret of office work is that you're not busy most of the time. I've got my work computer to my left and my personal computer to my right, so I'm almost always available to monitor positions.
"The subject matter helps too. Politics has deadline days where you need to watch constantly, but in general it's controlled. News tends to come during business hours, not at 1 a.m. Rotten Tomatoes runs on a weekly cycle: the embargo lifts, you monitor until Monday morning, then the next one. Music is even easier, mostly one data update a day. So it's a job that doesn't need 100% of my attention plus markets that don't need a huge time commitment."
The Spotify Market That Made Him "The Whistleblower"
Justin: Let's go back to Sunday night, June 28. You're looking at the Spotify US chart for the month-end market and you see a rising line on Olivia Dean's "Man I Need." Walk us through why that mattered over the next 48 hours.
Caleb: "Some background first. These markets started this year. They asked who would have the number one song on Spotify Global and Spotify US for a month. They turned into a perfect opportunity for what we call bonding: loading a lot of money onto a low-risk position for something like a 1% return in three days. That doesn't seem like much, but done over and over through the year, it compounds.
"New music typically drops on Friday, so after the last Friday of the month, you know the pool of songs. And there isn't much variance. The number one song on a given day is almost always number one the next day and the day after. At the end of June, Friday came and went with no new threats. Ella Langley's 'Choosing Texas' had been number one forever. I'd have said she was over 99% to win out the month.
"But somebody was paying 11 to 15 cents for Olivia Dean to take first. I tweeted a joke about some idiot thinking there was an 11% chance. Olivia Dean had only hit number one once, the previous November, and the song was about 11 months old. There was basically no chance. I went to bed.
"The next morning I thought, wait. Somebody's paying a lot and accumulating a lot of shares. So I dug into the data and saw Olivia Dean's streams had spiked from Friday to Saturday, which is normally a drop. An old song spiking for no reason was concerning. Then I found someone on Polymarket who'd accumulated more than 100,000 shares of Olivia Dean, and who'd made about $80,000 earlier in the month off another odd day of streams. I found he used the same screen name across accounts, including one offering stream-boosting services for money. So he had the means, he'd done it before, and he was accumulating shares. I stood to lose about $13,000 if it was Olivia Dean.
"I contacted Spotify, Polymarket and Kalshi and said I thought fraud was happening. Spotify opened a ticket. Kalshi's initial response was essentially that it was Polymarket users doing it and there was nothing they could do. My point was, do you want to pay out based on a fraudulent result?
"The next data update, Saturday to Sunday, when songs almost never go up, Olivia Dean and Drake both jumped again. They didn't quite reach first, but they were clearly on a path to it. I was nervous. At the time I had a very cooperative relationship with Kalshi, and they connected me with someone to work with on it. I assumed that if a fraudulent result happened, they'd make it right.
"Then the result came out, and it was Malcolm Todd's 'Earrings.' There was no chance that song would get first normally. Its streams would have needed to spike more than 70%, and there was no reason for that. I unfortunately had money on it not being Malcolm Todd. The funny part, after all the blaming of Polymarket, is that Polymarket didn't even have a bracket for Malcolm Todd. That's how unlikely it was. Whoever made money off it wasn't trading on Polymarket.
"I told Kalshi there was 0% chance this happened naturally, and they paid it out anyway. That's what bothered me. I went to the media, the story got out, and then Netflix contacted me."
Editor's note: Spotify later confirmed streaming fraud on "Earrings" and removed more than 500,000 streams from the song, according to Stereogum and Prediction News. Caleb's statements about who was behind it and how the market was resolved are his account.
Justin: How did that change your process going forward? Are you avoiding Spotify markets?
Caleb: "They removed the monthly Spotify markets altogether. If it's that easily manipulated, you can't really keep running them. There are still year-end ones, and Spotify does a full cleanup before releasing year-end results, so I'm not worried about fraud there. I still play those. But the monthly markets are gone as a category for me.
"More broadly, before this I always assumed that if a resolution source was fraudulently modified, Kalshi wouldn't pay out on it. Now I've seen that's not true. If somebody finds a way to manipulate a resolution source, the market can still pay out on it. That affects how I bond things. If a market pays out on a number on a website, anyone who gets into that website can change that number. So I'm more conscious of it. It's one more risk I weigh in every market."
Justin: How has it changed your relationship with Kalshi? Are you still trading there?
Caleb: "I withdrew a lot of money for a while, hoping Polymarket would become a stronger competitor on the US side by election season, but they're not quite there yet. When I loaded my portfolio back up for the midterms, I went with Kalshi."
Justin: What was the documentary process like?
Caleb: "It was a wild experience. Somebody from Netflix DMed me on Twitter and asked to set up a call. We talked on a Wednesday afternoon. I said I was going to Cancun for the weekend and we could meet when I got back, and they asked what I was doing tomorrow. By 9 a.m. the next day they were on a plane. They checked out my work area, which is pretty boring, and they loved the theater room, so that's where they filmed most of it, plus my backyard. They were at my house for five hours, an hour and a half of setup and teardown and three and a half hours of questions, for about a four-minute segment."
How He Built the Spotify Model
Justin: You've described your edge as 95% math, 5% gut. What went into the Spotify model?
Caleb: "Spotify has the history of every daily chart since 2017. I pulled all of it in and built profiles of songs, which gives you a range of performances. When a new song comes out that you think could threaten for number one, you compare it to previous songs and come up with ranges and probabilities. If prices stray too far from your ranges, you buy, and if they later stray the other way, you sell or hold. It really wasn't super complicated. Profiles, mins and maxes, averages."
Justin: In my world, every year or two I have to figure out my new edge, because the markets change or people catch up. Do you feel the same way?
Caleb: "Even before the Spotify markets went away, most of the edge in the daily markets had vanished. What I was doing was maybe a two out of ten in complexity, so it didn't take the market long to figure it out. Most days you'd look and the prices were fine. Then the edge came from being available. I'm on Twitter most of the day, so I might see a song announced before other people and move first.
"I do think I have a big edge on the year-end markets. Either people aren't paying attention or they aren't doing the same analysis. This year the front-runner is 'End of Beginning' by Djo, the guy who plays Steve Harrington in Stranger Things. The finale dropped around New Year's Eve, his song spiked, and it stayed up for months, building enough of a lead that every other song was chasing it. I'd track the songs outstreaming it daily and their decay rates to see if any would pass it. About a month ago we were down to two possibilities, and then the last real challenger dropped off enough that it clearly wasn't going to catch up. The market didn't seem to catch on. So my edge was paying attention when other people weren't, or doing analysis they weren't."
Justin: Who's on the other side of your trades? In sports, people assume it's firms like Susquehanna or Jane Street. Is that the case here?
Caleb: "It's definitely other retail traders. There's not enough money or volume in a lot of the culture markets to draw that level of interest. It might be a big Taylor Swift fan who knows she can't possibly lose. When her album came out last year, almost all my profit came from people who didn't realize Bad Bunny had an insurmountable lead by the time she dropped it."
How Rotten Tomatoes Markets Work
Justin: Rotten Tomatoes markets fascinate me. There's evaluating the movie, and then the meta of how Rotten Tomatoes reviews actually work. What's the process, and what makes these markets hard?
Caleb: "Rotten Tomatoes seems dumb, but a lot goes into it, because there are phases. Before anyone's seen the movie, you know the director, producer, actors and trailer. I've found that's not very useful and I don't have an edge there, so I don't usually trade that early.
"Then the social media embargo lifts, and people post reactions. At that point almost everything is positive, because everyone's hyping the movie, not critiquing it. So it's about reading between the lines. You notice who saw it and didn't say anything when they normally would, which probably means they didn't love it. Or a reaction that looks positive at first glance, but when you read closely, that's all they liked. That's usually where I start taking a large position.
"Then the review embargo drops, and from then until Monday morning you're looking at every review and the likelihood it gets counted. You'd think a given critic's review either always gets included or never does, but that's not how it works. A handful go in 100% of the time, but most are on a spectrum. Some go in 5% of the time, some 95%. So you figure out the pool, how likely each one is to count, and calculate the range of likely scores. It gets pretty complex for something as dumb as movie reviews."
Trading the Midterms
Justin: When this goes live, we'll be about a week from the midterms. How have you been trading them, and what are your plans heading in?
Caleb: "I've probably got a quarter million dollars on midterm markets right now. Start with the narrative. If you'd said, with no evidence at all, that an unpopular president's party loses seats in the midterm, you'd have been right in every midterm back to 2002. And 2002 was the exception because George W. Bush was still riding the post-9/11 high.
"So that's the baseline. Then you layer on this year. Trump went to war with Iran and prices spiked, the economy isn't doing great, and affordability is a huge issue. Those all push toward an anti-Republican year before you look at a single poll. Then the polls back up the same story, and now early votes do too. Everything points to a blue wave of some size. Whether it's huge or smaller is still up in the air, but it's looking on the larger side.
"Right now, Democrats taking the House is at 90 cents, which I think is absurd. With about 14 days to go, I don't see anything that saves the House for Republicans.
"From there you fine-tune. A lot of states have early voting, but how much you can learn from their reporting varies widely. Nevada is one of the best. It has partisan registration, about 80% of the vote comes in before Election Day, and registration lines up pretty closely with how people vote. Compare that to West Virginia, where almost half the state is still registered Democrat but a Democrat isn't winning statewide anytime soon. So you look at states like Nevada to gauge how big the wave is and bet accordingly. I've got some positions on Democrats winning in Texas, and on Troy Jackson beating Susan Collins in Maine, because the environment looks too blue for her to survive."
Justin: I'm extremely interested in these markets, but I don't think I've done the legwork to have an edge. Hearing from someone with a quarter million on the line is fascinating.
Caleb: "There's so much expertise needed to analyze early votes state by state, so don't do it all yourself. There are plenty of accounts on Twitter run by people who've genuinely been analyzing their own state for years and post the data. Rather than becoming an expert on every state and building models for each, it's more like being a CEO. You've got information coming from different places, and you decide what's reliable and what to act on."
Common Beginner Mistakes
Justin: For people new to this space, what are the common mistakes you see from beginners?
Caleb: "There are a lot of them. Last year there was a Taylor Swift market where people set a ton of money on fire, and I wrote an article about the basic trading mistakes they made.
"First, make sure you have an edge of some sort. Put in enough work that you're not just guessing. It's surprising how many people are comfortable throwing money at what's essentially a guess.
"Second, longshot bias. People see Taylor Swift shares at 1 cent and a 100x payout, and they start fantasizing about what they'll spend it on. In reality those shares are worth zero. The only time you should buy a 1-cent share is if it's more than 1% likely to happen. It's like a Powerball ticket: you can win hundreds of millions, but it's always negative expected value, and the same thing happens here on a smaller scale.
"Third, doubling down blindly. You buy at 30 cents, something happens, and it drops to 15. A lot of people double down, figuring if it was good at 30 it's better at 15. But if it fell from 30 to 15, it's usually for a reason. The market doesn't do that randomly. Consider what happened and what the updated odds are.
"Fourth, AI. It can look like an easy way to win, and it isn't. If you just ask Claude or any of these tools what the price should be, it's almost always going to be wrong. Use it like an advanced calculator. Have it do analysis, but don't have it come up with conclusions. It'll often give you hallucinated information. I'd see it in Kalshi comments all the time, people citing something as the reason a market moved, and it clearly came from an AI tool and wasn't true.
"Last, leave emotion out of it. Have a process. If your shares drop, take an analytical view of why. People have an almost reflexive instinct to sell and stop the pain. If you're down $200, you don't want to be down $300, so you sell. That's not always right. Why did it drop? Should it have dropped that much? Don't be an emotional trader. Be stone-cold analytical."
New to prediction markets? Our Prediction Markets 101 covers how pricing and order books work in plain language.
Justin: What's the one piece of advice for someone who wants to take prediction markets seriously and turn it into a real side hustle?
Caleb: "Put in the work. You're potentially competing with hundreds or thousands of other traders on every question. Why do you think you'll answer it better than anyone else? If you're not putting in a lot of effort to get a better answer than the average trader, you shouldn't expect to win. Whether it's building advanced models or tracking votes better than other people, it's got to be something, or you'll lose.
"I can't tell you what that is for you. That's up to each person. But it helps to pick topics you're interested in, because you'll spend a lot of time on analysis and models. If I didn't care about movies, spending thousands of dollars setting up Rotten Tomatoes models would have been pretty tedious. Having my trading line up with my hobbies has really helped."
Where to Find Caleb
Justin: People can find you on X at @GaetenD. Anything else coming up you want people to check out?
Caleb: "Just the midterms. That's my focus right now. I'm following Twitter all the time for the latest results, and that'll keep me busy until the election. Then it'll determine how many vacations and home remodels I can do afterward."
Follow Caleb on X at @GaetenD.
Want to talk through culture markets, Rotten Tomatoes or the midterms? Jump into the PredictQ Discord. If you're opening accounts on the exchanges Caleb trades, check the current sign-up offers first.
This article is an adapted interview based on a real conversation about how one trader approaches prediction markets. It is not an exact transcript, a directional prediction, a trade recommendation, or investment advice. The guest's statements about suspected manipulation and market resolution reflect his account and should not be treated as independently established findings. Prices, positions and odds discussed were quoted at the time of recording (October 20, 2026), are point-in-time, and may change. PredictQ is a marketing partner of Kalshi and Polymarket and may receive compensation for referrals.