When Investing Starts Looking Like Gambling — A Capital Wealth Read of Jason Zweig's Weekend Column

There is a column in the Weekend Journal this week that is worth pulling out of its folder and reading twice: Jason Zweig's B1 "Intelligent Investor," titled "It's Getting Harder to Tell Investing From Gambling." Zweig is one of the most careful writers in personal finance — he has been writing this column since 2008, took over Benjamin Graham's The Intelligent Investor annotations in 2003, and has spent his career studying the gap between what investors think they are doing and what they are actually doing. When he says the boundary has moved, advisors should listen, and clients should read the piece twice.

The argument is simple and uncomfortable. The infrastructure that used to separate gambling from investing — different apps, different account types, different language, different settlement systems — has been demolished. What used to live in a casino now lives one tab over from your 401(k). And the demolition was deliberate: prediction-market venues, options exchanges, ETF issuers and gambling-platform operators have spent the last five years studying brokerage UX and copying it, because brokerage UX is what builds long-dated trust and long-dated balances.

The piece names four specific forces that have merged the speculation track and the investment track into the same screen:

• Prediction markets at scale. Kalshi, Polymarket, and the rebuilt DraftKings "Event" product now run real-money contracts on rate decisions, earnings beats, election outcomes, and sports results. The UX is indistinguishable from an equities app — charts, watchlists, "positions" tabs, and dollar P&L. Kalshi is CFTC-regulated, which gives it the patina of legitimacy that Robinhood gave to options trading in 2019. Industry estimates put 2026 prediction-market open interest at over $3 billion across the major venues.
• Zero-day options (0DTE). Same-day expiry contracts now account for more than half of S&P index option volume on an average trading day, up from roughly 5% in 2016. The CBOE expanded 0DTE listings on the SPX to every weekday in 2022; the volume migration was immediate. These are not hedging instruments for retirees. They are lottery tickets priced by Black-Scholes, marketed as "tactical income," and held in retail accounts that have never seen a margin call.
• Single-stock-leveraged ETFs. 2x and −2x wrappers on names like NVDA (NVDL/NVDS), TSLA (TSLL/TSLS), and COIN now sit on the same screen as VTI. They reset daily, decay against the underlying in any choppy tape, and have produced 80%+ peak-to-trough drawdowns within calendar quarters. A 401(k) brokerage window will frequently let you buy them. We have seen them in client outside accounts.
• The re-platforming of gambling apps. Several sportsbook operators have redesigned their interfaces to read like brokerages — price charts on each contract, watchlists, research tabs, P&L displays in dollar terms instead of "winnings." Settlement now flows through the same bank ACH rails that hold index-fund contributions. Visually and operationally, the two activities have converged.

Five Drawdowns, One Recurring Casualty

The argument we want to make — and the reason this column lands harder for our book than the dozens of similar warnings we have read — is structural. Every major drawdown of the modern era took its worst toll on the same kind of account: a portfolio whose bet sleeve had grown unchecked, unlabeled, and indistinguishable from the planning portfolio. The form changed each cycle. The pattern did not.

1987 — Portfolio Insurance As "Hedging"

On October 19, 1987, the Dow fell 22.6% in a single session. The proximate trigger was a strategy called portfolio insurance — dynamic hedging programs marketed by Leland O'Brien Rubinstein and others that promised institutional accounts they could replicate a protective put without paying the option premium. The label was "hedge." The reality was a synthetic short put, sized to the full portfolio, that mechanically sold futures into a falling tape. When everyone's "insurance" sold at the same time, the bid disappeared. The Brady Commission's report (January 1988) is still the cleanest postmortem ever written about a position that was structurally a bet but had been sold into the planning portfolio as protection. The lesson: if you cannot describe a position's payoff in a single sentence, it does not belong in the planning portfolio.

2000 — Margin, Day Trading, And The "Diversified" Tech Account

The Nasdaq's peak was March 10, 2000 at 5,048. The bottom was October 2002 at 1,114 — a 78% drawdown that took roughly 15 years to recover on a price basis. The accounts that came apart were not "tech investor" accounts. They were retirement accounts whose owners believed they held a diversified portfolio that happened to be tech-tilted. The forensics are unambiguous: NASD margin debt peaked at $278 billion in March 2000 (5.4x the level in 1995), and the typical retail margin account had concentration ratios above 60% in three or fewer names — usually Cisco (CSCO), Sun Microsystems (SUNW), JDS Uniphase (JDSU), and the dot-com IPO of the month. The day-trading firms of the era — Datek, A.B. Watley, MB Trading — sold the activity as a profession. Frank Partnoy and Jane Bryant Quinn wrote at length about the casualties. The labeling failure was identical: "these are my investments." They were bets, sized to the full portfolio.

2008 — "Cash Equivalents" That Weren't

The 2008 crisis produced the cleanest example of mis-labeled positions in any postwar drawdown: auction-rate securities (ARS). Brokerages had sold roughly $330 billion of these instruments to retail and corporate clients as "cash equivalents." When the auction mechanism froze in February 2008, holders discovered their "cash" was actually a long-duration, illiquid claim on student-loan or municipal collateral. Several major brokerages eventually settled with regulators and bought back the paper, but holders were locked up for 6–18 months — some of them during the worst liquidity moment of the cycle. Separately, the same year saw the failure of Lehman structured notes that had been sold to retirees as "principal protected." The protection turned out to be a Lehman credit obligation, not an insurance product. The position your broker told you was safe is the position you need to label most carefully.

2020 — The Retail Options Surge

March 2020 saw the fastest 30% drawdown in S&P history. The recovery saw the fastest retail-options surge in history. By Q4 2020, retail accounted for roughly 25% of equity-option volume, up from under 10% in 2019. Robinhood added ~13 million new accounts in 2020 alone; the median age of the new account holder was 31. The behavior that emerged was not "investing on a dip." It was concentrated single-name call buying, frequently on margin, frequently in companies the buyer could not describe a cash-flow case for. The casualties were not the people who bought GameStop (GME) in January 2021 — they made headlines. The casualties were the much larger cohort who bought Peloton (PTON) calls at $160, Zoom (ZM) at $560, DocuSign (DOCU) at $310, and Carvana (CVNA) at $370 in late 2020 and early 2021, sized those positions like investments, and watched 80–95% of the value evaporate over the following 18 months. The pandemic produced more new retail accounts than any event in U.S. history, and almost none of them were ever asked to label their positions.

2022 — 0DTE, ARKK, And Crypto Leverage

2022 was the modern template for what Zweig is warning about now. ARK Innovation ETF (ARKK) fell 67% peak-to-trough. The 2x leveraged Nasdaq ETF (TQQQ) fell 80%. Single-stock leveraged products on names like AMC and TSLA produced drawdowns above 85% inside calendar quarters. Crypto's collapse — Terra/Luna in May, Celsius in June, FTX in November — vaporized retail deposits at venues that had been marketed as "high-yield savings." And 0DTE volume crossed 40% of SPX options for the first time. The accounts that broke were the ones that had let any of these positions grow into "real money" sleeves without ever being asked: is this an investment, or a bet? The 2022 drawdown was the first drawdown of the post-Robinhood era. It will not be the last.

The repeating pattern across all five cycles is the same: a wrapper made a bet feel like an investment, the wrapper held until volatility tested it, and the holder discovered — at the worst possible moment — that the position had been sized as if it had a cash-flow claim when it never did. The accounts that survived these episodes intact were almost always the ones with a clear separation between the planning portfolio and the discretionary sleeve. The accounts that did not survive intact almost always had no such separation.

Why The Line Moved: The Behavioral Architecture

The reason the four forces Zweig describes have been so effective at blurring the line is not accident — it is design. Three pieces of behavioral-finance literature describe exactly the mechanism, and they are worth naming because they are the playbook every modern fintech UX team is reading from:

• Mental accounting (Thaler, 1985). Investors treat money differently depending on the account it sits in, even though the dollars are fungible. The fintech innovation of the last decade has been to break down the mental walls between "play money" and "real money" by putting both inside the same app. Once those walls are gone, the discipline they enforced is gone too.
• Prospect theory (Kahneman & Tversky, 1979). Investors are loss-averse on gains and risk-seeking on losses. This is exactly the wrong wiring for any sleeve with negative expected value or high-variance payoffs. 0DTE call buying and prediction-market contracts are textbook prospect-theory traps — small frequent wins that condition the buyer, followed by infrequent catastrophic losses that the buyer doubles down on rather than absorbs.
• Narrative economics (Shiller, 2017–2019). Asset bubbles travel on stories, not spreadsheets. The current narrative — that ordinary investors can earn institutional-grade outcomes if they just use the same instruments institutions use — is the same narrative that drove portfolio insurance in 1987 and the day-trading firms in 1999. The form changes; the function repeats.

Zweig's own book, Your Money And Your Brain (2007), is essentially a 350-page warning about exactly this set of mechanisms. The new column is the same warning, re-aimed at the new wrappers. If you have not read the book, the column is the abridged version. If you have, this weekend's piece is the update.

The Four Categories We Use

Zweig's column is effectively asking advisors to be able to point at every line item and say which category it's in. We agree. This is going to be in every Q2 review we run. The category is the position's label — it is not a moral judgment, it is not a recommendation to sell, and it is not a value judgment about the holder. It is a planning question: can I run a Monte Carlo through this position? If yes, it is an investment. If no, it is a bet, and it needs to be sized accordingly.

Sizing The Bet Sleeve: The Math Behind The 1–3% Cap

The 1–3% cap is not arbitrary. It is rooted in the Kelly criterion (Kelly, 1956) and its half-Kelly application, which is how every professional discretionary trader sizes high-variance positions. The full Kelly formula is f* = (bp − q) / b, where b is the odds, p is the probability of a win, and q is the probability of a loss. For most retail bet-sleeve products — 0DTE options, prediction-market contracts, leveraged single-stock ETFs — the honest input ranges produce a full-Kelly result between 2% and 5% of capital. Half-Kelly, which is what disciplined practitioners actually use because full-Kelly is too volatile to survive psychologically, lands at roughly 1–3%.

What that means in plain language: if your bet sleeve is larger than 3% of investable assets, you are over-sized relative to any defensible probability estimate of the bet paying off. If your bet sleeve is below 1%, you are probably fine and the sizing question can wait until next review. If you cannot estimate a probability for the bet at all, the correct Kelly size is zero. The category exists; the position can still belong to zero of the portfolio.

An Anonymized Case From Our Book

A client of ours — a recently retired public-sector employee with a moderate-conservative plan — came to a Q1 2026 review with a question about whether to add to a position. The position was a leveraged single-stock ETF on a mega-cap tech name. He had bought it 14 months earlier with what he described as "a small amount." Over the holding period, the underlying ran, the leveraged wrapper compounded, and the position had grown from roughly 2% of the household account to 11%. At 11%, it had stopped being a bet. It had become a planning-portfolio risk: a single-name, daily-resetting, decay-prone exposure whose payoff distribution was indistinguishable from a long-dated option struck near the money.

The conversation was not "sell it all." It was three sentences. One: this is a bet, not a position. Two: it is now eleven times the size a bet should be in your plan. Three: we are going to trim it back to the size we agreed it should be at, harvest the gains into the planning portfolio, and keep the residual sized as Category 4. He did. Two months later the underlying corrected 22%, the wrapper corrected 43%, and the planning portfolio was unaffected. The conversation that mattered was the one that happened before the correction, when the position was at 11%. Without the category labels, that conversation does not happen, because there is nothing to point at on the screen and say this is the line item that does not belong at this size.

What This Means Operationally For Our Book

Three concrete things are changing in our process as a result of this column and the historical pattern it surfaces:

• Q2 reviews will open with a category audit. Before we discuss markets, performance, or themes, we will walk through positions screen-by-screen and assign each to a category. We have done this informally for years. We are formalizing it.
• The intake form for new household plans now asks about prediction markets and leveraged ETF exposure explicitly. This used to be discovered in conversation. It will now be discovered in writing, because discovery in writing makes for cleaner reviews and cleaner compliance documentation.
• Our quarterly client letter will include a one-line "category drift" note for any household whose bet sleeve has grown past 5% of investable assets without a deliberate decision to size it there. The drift is the problem; the deliberate decision can be defended.

A Note On Gambling & Allocation Discipline

Capital Wealth does not have a moral view on prediction markets or single-stock leverage. We do have a structural view: a portfolio with unlabeled bets in it is a portfolio you can't plan around. You can't run Monte Carlo on a position that might go to zero on an election. You can't model retirement cash flows through a sleeve whose expected return isn't knowable. You can't build a withdrawal schedule across a position that resets daily and decays in choppy tape. The 1987, 2000, 2008, 2020 and 2022 drawdowns each had different proximate triggers and different headline products, but the casualty profile was the same in every cycle: households whose bet sleeve had grown into the planning portfolio while no one was watching.

Zweig's warning this weekend is the same warning he has issued before every mania of the last twenty-five years, and it has aged well every time. The honest reading of his column is not that the four forces he names are evil. It is that they are popular, they are well-designed, they are profitable for their operators, and they have a documented track record of doing the most damage to the accounts least equipped to absorb it. The defense is not abstinence. The defense is labeling.

If nothing else, his column is a prompt to audit your own screen. Open your brokerage's "positions" tab tonight and, for each line, ask yourself a single question: is this an investment, or a bet? Investments have cash-flow claims you can describe in a sentence. Bets do not. If you can't answer for a given line — or if the honest answer surprises you — that is the line we are going to start your next review on.