Guest column: Caleb Davies, known on Twitter and Kalshi as GaetenD, is a long-time prediction markets trader currently ranked fifth in lifetime earnings on culture markets on Kalshi.
When trading in Spotify prediction markets, there are a ton of variables that you need to consider: day of the week, genre, days since release, new releases, and prior weeks’ performance. Now there’s a new one: Somebody appears to have manipulated streams in a way that benefited prediction-market positions.
Botting on Spotify
Stream botting is artificially inflating a song’s play count using automated tools rather than real listeners. The methods range from simple scripts running on a loop to sophisticated stream farms to malware that infects devices and streams the songs. The goal is to make fake plays look real. The more the streaming resembles actual listener behavior, the harder it is to detect.
Spotify isn’t oblivious to this. They conduct audits, remove suspicious streams, and have pulled royalty payments from artists caught botting. The audits are done daily before each chart is published, a little more thoroughly at the beginning of each month, and one final, very thorough review is done before Spotify Wrapped is released at the end of each year. But identifying botted streams is an imperfect art and not all are caught quickly, if at all.
People do it for a few reasons. The most obvious is royalties: Spotify pays out per stream, and if your bots are convincing enough to survive an audit, you can make a haul from your shitty songs. Labels and smaller artists have used it for years to juice chart positions, manufacture the appearance of momentum, or hit thresholds that trigger algorithmic playlist placement. Get on Discover Weekly or New Music Friday and the fake streams pay for themselves many times over.
But what if instead of needing to play the long game of playlisting, royalty collection, and building popularity, you could just 100x your money in a day? And that you could pull it off without any risk of Spotify blacklisting you? That’s where prediction markets come in. And there is strong evidence that someone has already found this one neat trick (traders hate it!).
Examining the evidence
Spotify makes daily stream histories available globally and by country here. Every morning, I put on my slippers, grab the WSJ from the porch, pour a glass of bourbon, then download the day’s US and global Spotify data and update my projections.
When you do this every day (the Spotify part, not the rest of it), you notice when things are off. The charts are a dance of day-of-week trends, new songs and their rapid decays, the long-term decay of established songs, and the occasional viral hit. When something spikes outside of that framework, it warrants a closer look.
There are several things I look for if I suspect fraudulent streams:
Geographic targeting – Is the surge in streams in a single country or worldwide? If it’s widespread, it is less likely to be artificial. A botted stream that doesn’t impact the targeted country’s chart is a wasted one.
Impact on the rest of the artist’s catalog – Some events are song-specific (a music video release, a song catching fire on TikTok), but others like concerts, television appearances, or a biopic release lift the entire catalog. If only a single track is impacted, it is more suspicious than if the entire catalog is.
Suddenness of the trend – Viral hits tend to build up over time rather than jumping up the charts in a single day. Likewise, their declines tend to be gradual rather than sudden. If a spike lasts a single day or a handful of days then drops again suddenly, that’s suspicious, too.
Violation of day-of-week patterns – Fridays are the strongest streaming days. Sundays are by far the weakest. The rest of the days are roughly even. Saturday to Sunday increases are rare. Anything outside of these norms raises flags.
Applying your knowledge
Now that you know how to detect fraud, I’m going to put you to the test.
Click on this chart. It shows the US/global ratio for every song that charted on both charts during the time period. The grayed-out lines are within a normal range. The highlighted lines show some very unusual activity. We can see that the US-to-global ratio skyrockets for several Bad Bunny songs on Monday, February 2 and again on Sunday, February 8. It also has a big jump for “Opalite” by Taylor Swift on Friday.
What do you think? Any reason to suspect fraud here?
Of course not! That data is exactly what we’d expect. Bad Bunny won Album of the Year at the Grammys and his catalog got a boost. Taylor Swift dropped a video for “Opalite” that was initially available only on Spotify Premium and Apple Music, before later being released on YouTube, so it makes sense that would spike her US/global ratio. And you can see that the curves are smooth. No reason to suspect manipulation here.
If you got that right, pat yourself on the head. If you didn’t, you’re really bad at this. I labeled the damn thing for you! Now click on this one. It again shows the US/global ratio for every song on both US and global for the date range.
Anything suspicious here?
I hope you said “Yes.” This is about as blatant as it gets. This is a streams boost targeted exclusively at the US that violates the normal day-of-week flow in a very sudden way. And this chart actually understates the severity because it uses filtered streams. On unfiltered streams, the cheating is ALS-ice-bucket-over-the-head-while-getting-kicked-in-the-nuts obvious:
I had to expand the y-axis from 150 to 400 to fit these lines into it! On Tuesday, there were more botted streams removed from the US than there were legitimate streams in the rest of the world for “Janice STFU.”
The Malcolm Todd case
On June 22, there was about $2k in open interest in the Malcolm Todd bracket and the share prices were in the single digits. Because his song “Earrings” had been stable for a while 500k+ streams below the chart leaders, if anything, being in the single digits rather than $0.01 was generous to its chances. Over the course of the week, open interest increased to about $76k, which is a lot for a dead bracket. Most of the dead brackets were in the low thousands in open interest. Somebody accumulated a lot of shares that were extremely likely to become worthless on the morning of July 1.
On Tuesday, June 30, Spotify market enjoyers of all shapes and sizes rolled out of bed, grabbed their WSJs, and poured their own bourbon while anxiously awaiting the results. All eyes were on Drake and Olivia Dean. Would Spotify be able to stop this dubious plan from working?
They did! The results were delayed for hours, and when they were posted, traders breathed a sigh of relief as the top US song was not “Janice STFU” or “Man I Need.” No, it was Malcolm Todd’s “Earrings.” Wait…what? How the f……? Those shares were worthless, but now they became worth $1 each? A previously stable song at 500k+ streams below the leader won? To illustrate how unlikely that was, here is every Sunday to Monday gain for “Earrings” since it emerged on the charts:
This isn’t just an unlikely result, it’s as close to a mathematical impossibility as you can get. If you and I both chose a star at random from all the stars in the observable universe, the odds that we chose the same one would be 77,000 times as likely as this increase occurring by chance. (Nerds: The same day-of-week comparisons of streaming changes for a stable song are close to a normal distribution. This result was an 11.24 sigma outcome.) Spotify was so busy filtering out artificial “Janice” and “Man I Need” streams that they missed Malcolm Todd’s artificial surge, which you can see in the earlier graphs as well.
On July 1, Spotify removed the artificial streams, but by then it was too late for Kalshi traders* because the market had already been paid out. Spotify’s auditing process was sufficient for Spotify’s purposes, but too slow to prevent a Kalshi market from being paid out based on artificial streams.
(*Polymarket did not have a Malcolm Todd “Earrings” bracket, so no Polymarket trader profited from this specific fraud. However, their daily markets remain fundamentally exposed to the exact same structural exploit.)
The fate of Spotify markets
As much as it pains me (and my wallet) to say: Spotify daily and monthly markets are fundamentally broken. I’ll walk through why.
The cost of manipulating the resolution source is small compared to the potential profits – It simply doesn’t cost very much to bot streams. There are a ton of websites that can easily be found on Google that offer those exact services.
The odds of getting caught are very low – Short of a full audit of a trader’s financial statements, linking a specific trader to a botting effort is extremely difficult. Even if a single trader owns all the shares of a winning fraudulent bracket, how do you prove they are the person who manipulated the results? A sophisticated fraudster could have somebody else do the botting, making it even more difficult to pin on them.
There are few obvious consequences for failing – Unlike an artist or a label that boosts their own songs, a prediction market trader receives no compensation from Spotify itself, so Spotify has no way of holding them accountable for the likely fraudulent streams. Spotify just removes the streams and the fraudster is only out the cost of them. They have no risk of losing royalties or being delisted.
Spotify’s internal controls aren’t calibrated for prediction markets – Spotify’s cadence of daily, monthly, and annual clean-up efforts work just fine for Spotify. By their end-of-the-year Spotify Wrapped, their data is as clean as it will get. But that’s not good enough for prediction markets, which pay out daily. Fraudsters can get rewarded immediately even if Spotify later catches and removes their streams.
Broader implications
Let me lay out an example I totally made up on my own and didn’t steal from anywhere:
A nuclear power plant dumps its waste into a river. Because of the toxicity of the river, the fish grow a third eye and taste terrible, ruining the lives of all fishermen who depend on the river. Those fishermen have a legal recourse through the Clean Water Act. They can hold the nuclear power plant liable for the pollution. Our legal system can handle this case.
But what if, instead of the nuclear power plant dumping its waste into a river, a prediction market platform creates a market on whether the river will be closed for contamination at any point this year? This creates a financial incentive for traders to contaminate the river and, if they do, it is not obvious under the current legal framework how the prediction market platform that created that incentive would be held liable.
This concept applies to the Spotify markets listed on Kalshi and Polymarket. Those markets provide an incentive for traders to manipulate stream counts, which means that Spotify has to dedicate more resources to their audits through no fault of their own. Yet because no Clean Water Act-type framework is in place for prediction markets, Spotify does not have an obvious legal recourse against the prediction market platforms even though their negative externalities cause Spotify harm.
Recommendations
Kalshi paid out a market based on incorrect results over calls for a delay in payment while an investigation was performed. (To be clear, Kalshi is investigating the situation, and The Event Horizon is aware that Kalshi is in contact with Spotify.) Regulators have fallen behind the industry and need to catch up. I have recommendations for each on how to move forward.
Before the June 29 results posted, I personally warned Kalshi about the potential for fraud via messages to their CEO Tarek Mansour, COO Luana Lopes Lara, and Head of Enforcement and Legal Counsel Robert DeNault. After they posted, I updated my messages with the new data which was extremely likely to be fraudulent and requested a delay on payout while an investigation was performed. Less than six hours later, DeNault replied that only Spotify could verify whether those streams were real, and Kalshi then paid out the market. The following day, Spotify removed those streams, meaning Kalshi had paid out based on incorrect results.
Trust is vital on a prediction market platform, and Kalshi’s response, in my view, looked too focused on plausible deniability and not focused enough on pausing settlement while the anomaly was investigated. In my opinion, that is what happened here. I think the proper way forward for Kalshi to rebuild that trust is to:
Halt trading on the July Spotify markets and refund positions.
No longer create daily or monthly Spotify markets based on streaming counts or charts.
Make the traders who lost in the incorrect Malcolm Todd “Earrings” bracket whole. The ideal outcome would have been to void this bracket entirely, but the early payout removed that as a viable option.
For policymakers and regulators (CFTC, Congress, state legislatures), our current legal framework is not equipped to handle prediction markets that create negative externalities. The act of creating a prediction market creates incentives that previously didn’t exist, and the costs of those incentives often fall on third parties such as Spotify. Regulators need to take this into consideration as they create laws and issue guidance regarding which markets are permitted. They should also create a framework for harmed third parties to obtain legal recourse either through removal of markets, prevention of the creation of new ones, or seeking damages from the offending platforms.
Anyway, that’s what I think. Now I need to get a refill.
This is Dustin…thanks to Caleb for taking the time to make us all smarter about prediction markets about Spotify, and all the issues at play. A few addendums:
Other reporting:
Spotify Challenges Prediction Markets After Song Chart Rigging (Bloomberg): “Spotify has asked Kalshi and Polymarket to remove its logo and clarify that neither company has a partnership with the streaming service after it identified users manipulating rankings that are tied to prediction market bets.”
Spotify Confirms Streaming Fraud After Kalshi Trader Cries Foul (Wired): “We’re in touch with Spotify and are actively investigating this matter,” Kalshi spokesperson Elisabeth Diana tells WIRED. …
“Nobody from Polymarket profited from the fraud. That’s what undermines Kalshi’s argument, because they didn’t have a Malcolm Todd bracket,” Davies tells WIRED.
Polymarket refutes this theory as well. “It’s actually not plausible since we didn’t even have Malcolm Todd as an option on this Spotify market,” said spokesperson Annabel Walsh. The company confirmed it’s reviewing the broader streaming manipulation situation, but hasn’t identified any immediate manipulation thus far.
📍 My take: Most of this could have been avoided — and perhaps even paraded as a good news story — If Kalshi had taken Caleb’s advice and just waited to pay out the market. I don’t understand the rush to resolve it when there was pretty clearly an issue. More to come…
You can’t pause current events when your payments can’t keep up. With every failed deposit or lengthy withdrawal process, customers lose trust in your brand. And when money doesn’t move fast enough, your liquidity suffers. PayNearMe’s PayXMTMplatform is purpose-built for Payment Experience Management, giving you the fast, familiar payment options your customers want while streamlining back-end operations for your team. Get real-time payments for real-time events—from one simple integration.
Prediction markets news roundup
🚨 Important stuff
North Carolina OSB Tax Hike, Predictions Tax Headed To Governor (InGame): “Both chambers of North Carolina’s General Assembly on Wednesday passed a conference report of its budget bill, which includes an online sports betting tax increase, a tax on prediction markets, and a change in how gambling losses may be deducted on personal tax returns. SB 257 will now be sent for approval to Gov. Josh Stein, who has 10 days to sign or veto. If he does neither, the bill will become law.”
⚖️ Legal and regulatory news
Nevada Supreme Court Upholds Kalshi Geofencing Mandate (Sports Betting Dime): “The Nevada Supreme Court on Wednesday denied Kalshi’s emergency motion for a stay pending appeal. The prediction market operator will be required to implement geofencing technology to prohibit certain event contracts in Nevada throughout the course of the state’s lawsuit against the company.”
I’ll be at NEXTPredict in October; a lot of big names and smart people from the prediction market industry will be there. Get your tickets now for 15% off using this link and promo code CLOSINGLINE15.
📖 Everything else you should know/read
ADI Predictstreet Expands into New Prediction Markets Following Successful Launch for FIFA World Cup 2026 (press release): ADI Predictstreet, the global prediction market platform and Official Prediction Market Partner of the FIFA World Cup 2026™, today announced the next phase of its growth strategy after successfully completing the initial phase of its regulatory framework in Gibraltar.
Following strong performance across its regulatory obligations – including operational resilience, consumer protection, compliance and market integrity – ADI Predictstreet has received permission to expand beyond football into a significantly broader range of regulated prediction markets.
At the conclusion of the FIFA World Cup 2026™ tournament, the platform will progressively introduce new markets across additional sports, entertainment, culture, weather, and selected political events, further advancing its long-term ambition to become the leading destination for understanding what the world believes will happen next.
The expansion represents an important milestone in ADI Predictstreet’s broader mission to bring prediction markets into the mainstream by enabling individuals to forecast real-world outcomes through transparent, regulated and accessible markets.
“Successfully completing this first phase with Gibraltar validates both our platform and our approach to responsible innovation,” said Dimitrios Psarrakis, CEO of ADI Predictstreet. “While sport was the ideal place to introduce our prediction market platform to a global audience, our ambition has always been to build one where people can participate in forecasting the events that shape our world – from sport and entertainment to culture, weather and beyond.”
“When Gibraltar licensed ADI Predictstreet, we saw an opportunity to lead the development of a new regulated sector founded on innovation, integrity and consumer protection,” said Nigel Feetham, KC MP, Minister for Justice, Trade and Industry, Government of Gibraltar. “In just a few months, that vision has gained global recognition through the FIFA World Cup™ and a series of landmark international partnerships. It demonstrates how forward-looking regulation can create entirely new markets, attract world-class businesses and reinforce Gibraltar’s position as one of the world’s leading regulatory jurisdictions.”
Michigan Gaming Control Board Leaves NCPG Over Kalshi’s Membership (Legal Sports Report): “In a letter addressed to NCPG Executive Director Heather Maurer, the regulator noted Michigan’s temporary restraining order against Kalshi that prevents its operations in the state through July 13. The letter to Maurer was addressed Wednesday but the MGCB announced the news Thursday afternoon.”
Most prediction market contracts have low volume, leaving users exposed to volatility and bots (CNBC): “A CNBC analysis found that about 70% of all closed markets on Polymarket saw under $10,000 in reported volume from 2021 to the end of May this year, according to Polymarket’s Gamma API. The Gamma API records notional volume on both sides of the trade.”
⏱️ Day eight of asking for more than one CFTC commissioner
This is day eight of me asking the Trump administration to appoint more commissioners and at least one Democrat:
