If you go on Robinhood you can bet on stocks. You can make as many bets as you want: You can just buy an index fund and hold it forever, but if you’d like more excitement you can buy a bunch of single-stock options every morning and sell them in the afternoon. If you make a lot of bets, you will win some and lose some: Even professional trading firms with very good records might win, you know, 53% of their bets. You’d be quite happy with 51%. If you make 1,000 trades a year, make $10,000 each on 510 of them (51%) and lose $10,000 each on the other 490, (1) that’s fine work and (2) you’re up $200,000 for the year. And then you have to pay taxes. Intuitively, you have $200,000 of income from your trading; if your tax rate on this money is 37%, you pay $74,000 and keep $126,000. More technically, what you have is $5.1 million of short-term capital gains and $4.9 million of short-term capital losses, which you can use to offset the gains. [1] So $200,000 of net short-term capital gains, but when you fill out the form, you’ll list both the $5.1 million of gross gains and the $4.9 million of offsetting losses. If you go on DraftKings you can bet on sports. You can make as many bets as you want. If you make a lot of bets, you will win some and lose some. If you make 1,000 bets a year, make $10,000 each on 510 of them (51%) and lose $10,000 each on the other 490, (1) that’s fine work and (2) you’re up $200,000 for the year. And then you have to pay taxes. It’s the same analysis as above: $5.1 million of gross gains, $4.9 million of offsetting losses, $200,000 of net gains, $74,000 of taxes, $126,000 of after-tax income. [2] Or that was how it worked. But, as we discussed last week, the taxation of gambling winnings has changed: “One sentence of the 940-page [‘One Big Beautiful Bill Act’ passed last week] says gamblers can deduct only 90 percent of their annual losses, instead of 100 percent, which has been the norm.” [3] So now the math is: $5.1 million of gambling winnings, $4.9 million of gambling losses, but only $4.41 million of those are deductible, so you have $690,000 of net taxable gambling winnings — more than three times your actual net winnings. And you pay $255,300 of taxes. But you only made $200,000. The tax rate on your winnings is 128%: The IRS takes more than all of your income. [4] This seems bad, if you are in the gambling business: Your margins were always slim, and now they might flip to become negative. The Athletic reports: “Not to be hyperbolic, but it’s an extinction of professional sports betting,” a gambler who goes by the pseudonym Jack Andrews told The Athletic. Andrews is a professional gambler with a popular YouTube channel and a contributor to a MasterClass on sports and gambling. “It is the end of my career, potentially.” You can still trade stocks on Robinhood though. That’s fine. Nothing has changed. Your losses on stocks are still 100% deductible against your winnings. [5] The new tax bill discourages gambling, but it doesn’t discourage stock investing, and it doesn’t even discourage stock gambling. Just regular gambling. That is a subtle distinction. You can bet on sports on Robinhood. In February, Robinhood rolled out “event contracts” on the Super Bowl through Kalshi Inc.; it quickly canceled those contracts after regulatory pushback, but it restarted a few prediction markets in March, including contracts on the college basketball tournaments. And Kalshi, a prediction market registered as a futures exchange with the US Commodity Futures Trading Commission, offers many more event contracts to US retail traders, including lots involving sports. When I went to Kalshi.com this morning, the top market it offered me was on the winner of the women’s Wimbledon semifinal; the front page also offered contracts on the baseball all-star game and the home run derby, as well as the winner of the Chicago mayoral election, the next chief executive officer of X, the winner of the Nobel Peace Prize, the price of Bitcoin, the number of tornadoes this month and the Rotten Tomatoes score of Happy Gilmore 2. So, lots of bets, some of them on sports. We talked about this last month, when I wrote: Kalshi offers a prediction market where you can bet on sports. No! Sorry! Wrong! It offers a prediction market where you can predict which team will win a sports game, and if you predict correctly you make money, and if you predict incorrectly you lose money. Not “bet on sports.” “Predict sports outcomes for money.” Completely different. The difference matters for somewhat baffling legal and jurisdictional reasons. The legal distinction we were discussing last month is that commodity futures markets are regulated by the CFTC (which is quite Kalshi-friendly these days), while gambling is regulated by the states. Several states have sent Kalshi letters saying “hey you are doing sports gambling without complying with our regulations, knock it off,” and Kalshi has sent back letters (and lawsuits) saying “buzz off, states, we’re a commodities exchange.” So far Kalshi has been pretty successful: CFTC regulation really does preempt state regulation, so if a state thinks that a Kalshi sports contract is unregulated gambling, that’s the state’s problem, not Kalshi’s. But now there is another important legal distinction. Commodity futures markets are financial markets, meaning that your losses can 100% offset your winnings for tax purposes, but sportsbooks are gambling, meaning that your losses can only 90% offset your winnings. (Not tax or legal advice, and this is all pretty new, but: probably.) If you bet on sports on DraftKings and win 51% of the time, after taxes, you will lose money. If you bet on sports on Kalshi and win 51% of the time, after taxes, you will make money. That is considerably better! The Athletic goes on: Speculation online about the reasons for the provision has run wild, ranging from straightforward accounting explanations to conspiracy theories that sportsbooks were behind the change in an attempt to eliminate sharp bettors or that the provision benefits new prediction market firms such as Kalshi, which are seen as rising competitors to sportsbooks. “You can deduct your full losses under this bill if you’re one of these predictive sites because those aren’t technically online betting sites,” [Representative Ro] Khanna told The Athletic. (Kalshi Chief Executive Officer Tarek Mansour told me that “people keep thinking we were behind” the new provision, “but I didn’t even know about it. I’m more focused on building a 10x product.” Still, good deal for him.) And the tax advantage of Kalshi is not just about winning gamblers, or the new 90% limit. What if you are a losing gambler? What if you win 49% of your bets, rather than 51%? (What if you lose all of them?) Generally speaking — even before last week’s changes — net gambling losses are not deductible. [6] If you make $150,000 a year at your job, and lose $20,000 a year betting on sports, you still pay taxes on $150,000 of income: If you’re a losing gambler, then intuitively you are gambling for entertainment, and money that you spend on entertainment is not tax deductible. On the other hand, if you do your sports betting in your brokerage account, and you lose money, you should be able to offset those losses against your other brokerage winnings. [7] If you buy a broad index fund in January and sell it for a $200,000 gain in December, you have $200,000 of taxable income. If you also lose $200,000 betti—I mean, trading events contracts on sports, then presumably you could offset those losses against your gains and avoid paying any taxes on your stock sales. (Not tax advice!) Somewhat counterintuitively, losses can be useful for investors, [8] and a certain portion of the financial industry is dedicated to finding new kinds of tax losses. Kalshi has done that! One possible pitch is: - You are a sophisticated high-net-worth long-term stock investor who would like to optimize your taxes by strategically harvesting tax losses.
- You also like to bet on basketball, not because you are a hardcore grinding edge gambler but because you’re rich and basketball is fun. You lose more than you win, but you have a good time and you can afford it.
- If you do your recreational betting on basketball at Kalshi, that’s good for your stock portfolio.
I wrote last month that this situation — where Kalshi offers sports bets that compete with sportsbooks but legally don’t count as “gambling” — “strikes me as an extremely, extremely weird state of affairs,” and it still does. “Do your sports betting on a CFTC-regulated futures exchange because that offers tax advantages” is, you know, probably not an intended consequence of the US financial regulatory or tax systems. But here we are. There is something cynical about it. “I just don't really know what this has to do with gambling,” Mansour has said about Kalshi’s sports contracts, but I do! But it’s not entirely Kalshi’s fault. An important fact of modern retail financial markets really is that they have become fun places to gamble, and everyone involved is leaning into that. “The whole economy is a meme stock now, so enjoy the ride,” I grumbled in April. We talked last week about Robinhood’s push to sell private company stocks to retail investors without disclosure. Robinhood CEO Vlad Tenev argued that SpaceX is no riskier than many other things you can spend your money on, and I admitted that he’s not wrong: “The general public can buy all sorts of speculative things in the public markets (zero-day options!), or the crypto markets (memecoins), or sports betting,” and they are all increasingly substitutes for one another. When Robinhood rolled out Super Bowl bets in February, I wrote: Robinhood is right, isn’t it? “We recognize an opportunity to better serve our customers as their interests converge across the markets, news, sports, and entertainment.” My theory — that financial markets represent investments in real economic activity, while betting on sports represents zero-sum bets on sports — is outdated. The link between finance and real economic activity was always indirect and imperfect — lots of financial markets activity has always been speculative and irrational — and it is increasingly inessential. All of it is betting on sports. Sports are sports, and entertainment is sports, and politics is sports, and crypto is sports, and stocks are sports. Democratizing finance doesn’t mean giving people easier ways to invest in broad economic growth, or to finance new business ideas. It means letting them make the bets that they want to make. The modern way to gamble is through the financial markets, and people like betting on sports, so of course the modern way to gamble on sports is through the financial markets. Now that is also the tax-advantaged way to gamble on sports. I have written that there are two broad ways to think about using artificial intelligence in investing: - You build a machine learning system that predicts stock prices: It ingests market data, identifies useful signals, and finds trades based on those signals.
- You ask ChatGPT “hey what stocks should I buy,” but you write such a good prompt that it gives you the right answer.
The first approach is a traditional way to run a quant hedge fund, hiring skilled researchers to apply machine learning techniques to predicting stock prices. In this approach, the hedge fund researcher is essentially an AI engineer: Her job is to understand how to build machine learning systems, to have the right model-building skills and economic intuitions and inductive biases to turn data into signals. In the second approach, the hedge fund researcher is essentially a prompt engineer: Her job is to understand how to use AI systems, and the AI system’s job is to be generally very intelligent. If you have access to an AI system that is better than the best human at all sorts of knowledge jobs, it is a little silly to reinvent the wheel by building a specialized AI system that’s really good at picking stocks. Just ask the generally intelligent AI system “pretend you are a hedge fund researcher, and pick the right stocks.” In the second model, you might literally use ChatGPT or some other off-the-shelf model from a big AI lab. Or you might build your own model. But your model will target something like what the big AI labs target: Artificial general intelligence, human-like performance. If you succeed, the model will not be a box that an analyst manipulates to find stock alphas; the model will be more like an analyst itself. You will ask the model “hey what do you think about the market here” and the model will give you a coherent answer. You can combine the two approaches by asking a model to build you a machine learning system that predicts stock prices. Bloomberg’s Justina Lee reports: Man Group’s quant equity unit says it has begun using a new artificial-intelligence system that can generate, code and backtest trading ideas — marking the arrival of agentic AI at the world’s largest listed hedge fund. Boston-based Man Numeric says it built the tool — known internally as AlphaGPT — to mimic how its researchers develop new investment signals. It digs into data for ideas, writes the code for potential strategies and then tests them on historic data. Such a system that can autonomously perform multi-step tasks is known as agentic AI, one of the hottest trends right now in Silicon Valley. While humans still vet the outputs — and errors like hallucination remain a big issue — the firm says the goal is to automate more of the research pipeline and accelerate the discovery of smart, rules-based trades. Several dozen signals generated by the AI system have passed Man Group’s investment committee and are slated to be deployed in live trading, according to Ziang Fang, a senior portfolio manager. There is something a little indirect about asking a smart computer “please code me up a system to find good stock trades,” rather than “please tell me some good stock trades,” but I guess this is the state of the art right now. Hey good for them: Hunterbrook Global has been valued at $100mn after a recent fundraising, as the novel US newsroom-cum-hedge fund revealed to investors that it planned to move into litigation. … Hunterbrook’s fund generated a 31 per cent return in the second quarter of 2025 and a 16 per cent return year to date. We have talked about Hunterbrook — the hedge fund that’s also a newspaper — a lot, and one thing that I have said is that, if you are in the business of finding problems at public companies, you should really try to monetize that every way you can. The classic approach is “find bad companies and sell their stocks short,” but there are other approaches that are (1) maybe better and (2) certainly additive. “Find bad companies, sell their stocks short, sue them for securities fraud and file a whistleblower complaint with the US Securities and Exchange Commission” is strictly better than only doing one thing. Hunterbrook is using every part of the bad companies; the Financial Times reports: Hunterbrook also revealed to investors in a letter that it planned to further exploit its news gathering by launching a litigation business that would partner with law firms on cases enabled by the newsroom’s reporting. Also, if your business model is “find bad companies and sell their stock short,” you will investigate a lot of companies (to see if they are bad), and some of them will turn out to be good. You could … buy them? Hunterbrook initially envisaged that it would short stocks in instances where its newsroom exposed scandals, but this approach has been sidelined in an “irascible bull market”, according to the investor letter. The letter also details how Hunterbrook is generating a sizeable portion of its returns by taking long positions in businesses its journalists have investigated and found to be sound. They should publish those too. “BREAKING NEWS: We looked into this company to see if it’s a fraud, but actually they’re very nice and you should buy their stock.” There’s not enough good news these days; scandals are more newsworthy than “everything’s fine here.” Unless you are long the stock. Being the sell-side lawyer on public-company mergers is a lucrative business. When a company is selling itself, it will care a lot about getting a high price and certainty of deal closing, but it won’t care very much about its legal fees, because it doesn’t pay them: When the deal closes, the buyer will own the target and will have to pay its legal fees. Those fees are not the target’s problem. Charge whatever you want. [9] It is occasionally, though, a risky business, for the same reason: When the deal closes, the buyer will own the target, and will have to pay its legal fees. What if it doesn’t want to? This happens from time to time. In particular, sometimes (1) two companies will agree on a deal, (2) the buyer will try to back out, (3) the target’s lawyers will go to court to force the buyer to close, (4) they will work hard and do an excellent job while racking up a lot of billable hours, (5) they will win, (6) the buyer will be forced to close, and (7) now the buyer will be really mad. It wanted to get out of the deal, but now it has to do the deal, and it has to pay the lawyers that forced it to do the deal! Of course it will try to stiff them. The famous case is Elon Musk trying not to pay Twitter Inc.’s lawyers (who forced him to buy Twitter after he tried to back out), but that was not unprecedented. At the Financial Times, Sujeet Indap reports: Law firm Quinn Emanuel has sued a listed 3D printing company, claiming it was short-changed on a $30mn fee for winning a legal battle over a troubled merger. Quinn represented 3D printer Desktop Metal in a Delaware court fight to complete its $183mn acquisition by larger rival Nano Dimension. A judge ordered the deal completed in March but Quinn has yet to be paid, and now claims it is owed $90mn by Nano as the successor company. “Nano Dimension and its CEO, Ofir Baharav, have weaponised their newly acquired control over Desktop Metal, Inc. to exact revenge against the attorneys who defeated them — Quinn Emanuel,” the law firm said in the complaint filed in state court in Massachusetts, where Desktop Metal is headquartered. Yeah, on the one hand, of course they’re not going to be happy about paying the lawyers who defeated them. On the other hand, Quinn has already won once, so I guess I’d bet on it winning again. The classic advance fee scam is: I have a foolproof way for you to get $5 million, but to clear up the last few technicalities, you need to give me $100,000 now. Of course it is a scam, so if you give me the $100,000, (1) I will not actually get you $5 million but (2) no see that’s not my fault, there was an unanticipated holdup at the Ruritanian register of wills, but if you give me another $50,000 we can get that expedited. And so forth. I am trying out this scam on a lot of people, and most of them aren’t biting; if you bite once, I am going to come back to you over and over again. “You’ve spent so much already,” I will eventually tell you, “what’s another $170,000 to finally get that $5 million?” Ideally I would pull this scam on people who (1) are naive dreamers who believe that they are special and deserve to get $5 million out of the blue and (2) have a lot of money. At Bloomberg Businessweek, Brent Crane has a story about a company called PageTurner, which allegedly pulled a version of the advance fee scam on writers who dreamed of publishing books and getting them turned into movies. Maybe not the audience I would have chosen (are aspiring writers notably wealthy?), but after reading the story I guess I see their point: Alex Skywalker—a writer who, fearing for his safety, requested a pseudonym [10] —was in his mid-60s with a few self-published books to his name when he first started working with Sordilla’s company in 2018. He received an email from a PageTurner agent going by Danny Evans. In 1999, Skywalker says, he made a “decision to write down all of my personal experiences with intelligences beyond ours and past lives I had become aware of.” Two decades in, he’d completed books spanning the genres of new age, theology and science fiction and retired from his career in IT management, eager for his books to get the movie treatment. “I hoped for sales to entice movie production,” he says. He paid PageTurner about $2,500 for a publishing package that included copyediting, formatting and printing-on-demand for his latest book. After these services were performed, Evans reached out to say Netflix was interested in adapting a romantic sci-fi series Skywalker had self-published into a TV show and would pay him $8 million. Overjoyed, he agreed to move forward with Netflix, with his PageTurner handlers leading the way. “That’s when the bleeding began,” he says. He ended up paying a total of $130,000 for various movie-related services. “Often they brought happy news, saying that Universal or Sony wanted in, or that Ana de Armas and Matt Damon were interested.” They were not. Andreessen Horowitz: “We’re Leaving Delaware, And We Think You Should Consider Leaving Too.” Yaccarino Stepping Down as CEO of Musk’s X After Two Years. Musk Says Grok Chatbot Coming to Tesla Vehicles by Next Week. Italy’s Ferrero Agrees to Buy Kellogg in $3.1 Billion Deal. Meta Poached Apple’s Pang With Pay Package Over $200 Million. “I thought, I’m 47 years old, all my life I wanted to study lattices, if I don’t do it now then it’s never going to happen.” Can millennial dads have it all? “Who needs to be the founder of a tech company when you can write a newsletter on beehiiv?” If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks! |