KINETIC ALPHA
Research
Prediction Markets · Household Portfolios · Market Structure

The Speculation Sleeve

Half of Gen Z investors moved money meant for investing into sportsbooks last year. The industry's answer — "stop betting, buy index funds" — is losing to a product that pays out Sunday night. The better answer is structural: a capped, guardrailed sleeve of exchange-traded event contracts that keeps the lottery ticket and caps the ruin. Built here with live Kalshi and Polymarket order books, and an honest look at who, legally, could manage it.

Gen Z investors who diverted investing money to sports betting, past 12 mo
52%
Betterment 2026 Retail Investor Survey (n=1,000, Mar–Apr 2026)
Treat sports betting as part of long-term financial strategy
26%
Gen Z, vs 14% Millennials · 6% Gen X · 1% Boomers
Gen Z in (or considering) prediction markets & sports betting
32%
Northwestern Mutual 2026 P&P Study (Harris Poll, n=4,357)
Gen Z high-risk investors who say they feel financially behind
80%
And ~80% believe high-risk assets reach goals faster
01 · The Problem Is Real, and Abstinence Is Losing

Gen Z didn't stop saving. They re-routed it through a worse market structure.

The August 2026 Betterment survey behind the recent Yahoo Finance coverage is blunt: 52% of Gen Z investors redirected money originally intended for investing into sports betting over the past year, and 26% describe sports betting as a deliberate, ongoing part of their long-term financial strategy — four generations of data collapsing to near zero by the Boomers (1%). Northwestern Mutual's 2026 Planning & Progress Study explains the why: among Gen Z investors holding speculative assets, 80% feel financially behind, and roughly the same share believe high-risk vehicles will get them to their goals faster than index funds ever could. This is not ignorance of compounding. It is a judgment — houses feel unbuyable, wages lag, time horizons feel unaffordable — that the only way out is convexity.

"When a prediction market or sportsbook starts to feel like a retirement strategy, we have a problem."— Sarah Levy, CEO, Betterment (2026 Retail Investor Survey)

The damage is measured, not hypothetical. Baker, Balthrop, Johnson, Kotter and Pisciotta's Gambling Away Stability — published in the Journal of Financial Economics in June 2026 — tracks ~184,000 households and finds legal sports betting access cuts net brokerage investment by roughly 20%, with heavy bettors cutting deposits by more than half; about 20 cents of every dollar sent to a betting app is money that never reaches long-term savings. Hollenbeck, Larsen and Proserpio find average credit scores fall ~12 points after online legalization, with bankruptcies and delinquencies rising. And the flow is accelerating: legal US sportsbooks took $166.9B of handle in 2025 and kept $17.0B of it — a blended hold above 10% — while Robinhood's event-contracts revenue hit $156M in Q2 2026, exceeding its equities revenue.

So the realistic design question is not "how do we get Gen Z to stop speculating." Fifteen years of responsible-gambling research says voluntary restraint tooling barely moves net losses. The question is: if a generation insists on holding a book of binary risk, what is the least destructive market structure for it, what guardrails actually bind, and can anyone legally manage it for them? Prediction markets are the only venue where that question has a defensible answer — because they are the only venue where the house edge is a fee schedule, not a business model.

Figure 1 · Structural cost per $100 staked, by venue
Expected cost (house edge / fees) Expected gain (index, 1yr real)
Structural take of each venue — before any skill or behavior. Prediction-market bars are worst-case taker fees at 50¢; resting (maker) orders pay ~zero on both venues.
Sources: AGA State of the States 2026 (2025 blended hold 10.2%); standard −110 vig math (4.76%); parlay holds 20–30%; Kalshi fee schedule eff. Jul 7, 2026 (taker 0.07·P·(1−P), max 1.75¢ at 50¢; maker multiplier defaults to 0); Polymarket US fee schedule eff. Jul 1, 2026 (taker 0.06·p·(1−p) ≈ $1.50 max per 100 contracts; maker rebate); Powerball expected-value analyses (~33%+); long-run US equity real return ~6.5–7% (Damodaran series). Behavioral asterisk: structure is not outcome — see Figure 2.
The honest caveat: cheap structure, expensive behavior

Karl Whelan's study of 300k+ settled Kalshi contracts found takers lost ~32% of stake on average (makers ~10%) — vastly worse than the fee schedule implies — because retail crosses the spread to buy longshots. The venue is near-zero-sum; the behavior is not. That gap between structural cost (~1–3.5%) and realized retail loss (~20–32%) is precisely the gap guardrails have to close. Every strategy below is engineered against the specific behaviors in that data.

02 · Why Event Contracts Can Carry Guardrails Sportsbooks Can't

A sportsbook is a counterparty. An exchange is a venue.

Four structural properties make CFTC-designated prediction markets — Kalshi, and since late 2025 Polymarket US through its QCX acquisition — a fundamentally different chassis for retail speculation than any sportsbook:

  • Zero-sum vs. negative-sum by design. Traders face each other, not a book that re-prices you on skill. Fees are the only structural leak: Kalshi charges takers 0.07·P·(1−P) per contract (max 1.75¢ at 50¢), Polymarket US 0.06·p·(1−p) with maker rebates. A sportsbook's blended 2025 hold was ~10.2% — and parlays run 20–30%.
  • Prices are probabilities you can audit. A 6¢ recession contract tells you the market's number. A +650 parlay price tells you the book's margin. Calibration is publicly measurable (Figure 2) — no equivalent exists for sportsbook pricing.
  • Position limits are already exchange-native. Kalshi operates under CFTC-certified per-market position limits ($25,000 for most markets). The rails for hard caps exist at the venue level — sportsbooks limit winners, exchanges limit positions.
  • The underlying set is diversifiable. Rates, inflation, payrolls, elections, climate, AI milestones, crypto ranges, sports — a book of uncorrelated binaries is constructible. A sports-only slip book is one correlated behavioral channel.

None of this makes event contracts an investment. Expected value net of fees is ~zero at best, and Section 05 is emphatic about what that means for fiduciaries. The claim is narrower and more useful: for the speculative dollars Gen Z has already committed to spending, this venue is an order of magnitude cheaper, transparent enough to audit, and — uniquely — programmable enough to wrap guardrails around.

Figure 2 · The favorite-longshot tax: price vs. realized win rate
Realized frequency (Kalshi, 300k+ settled contracts) Perfect calibration
Mid-priced contracts are honest. The lottery corner is not: 5¢ contracts resolve YES only ~2% of the time. Longshot buying is where retail losses live — so the guardrails ration it rather than ban it.
Stylized from Whelan, "Makers and Takers: The Economics of the Kalshi Prediction Market" (50¢ contracts win ~half the time; 5¢ contracts win ~2%) and the 2026 cross-venue calibration study of 353M trades (sports near-perfectly calibrated within 48h; favorite-longshot bias grows with horizon, slope ~1.32 beyond one month). Curve is illustrative of the documented pattern, not a fitted dataset.
03 · The Investable Universe

What a diversified event book actually looks like, priced this morning

Everything below is a real, open contract snapshotted from the Kalshi and Polymarket public APIs on August 17, 2026. Three things to notice. First, category breadth: a Gen Z saver can hold rates, recession, inflation, index ranges, climate, AI milestones, elections and crypto in one book — this is what makes a "sleeve" diversifiable where a betting slip is not. Second, liquidity tiering is extreme: the Nasdaq year-end tail has ~1.1M contracts of open interest at a 1¢ spread, while an April-2027 Fed strike quotes 32/83 — a 51¢ spread that would vaporize a retail order. Third, time is an asset class here: same-day weather, monthly CPI, year-end indexes, a 2028 election — expiries ladder from hours to years, which the strategies below exploit deliberately.

Figure 3 · Live contract universe — snapshot 2026-08-17
Kalshi Polymarket Deep spread ≤2¢ & OI/liquidity high Tradable Thin retail should not cross
YES price = market-implied probability. Filter by category. Thin rows are shown deliberately — they are what the liquidity guardrail screens out of every built strategy.
ContractVenueCategory YESSpreadVol / OI·LiqExpiresBook
Kalshi: yes_bid/yes_ask in cents, volume and open interest in contracts, from api.elections.kalshi.com/trade-api/v2. Polymarket: midpoint prices and liquidity in USD from the public Gamma API (global books; the CFTC-regulated Polymarket US venue lists a narrower subset at slightly different prices). Election-market "expires" shows the resolution event, not the API's post-settlement close date. Prices move continuously; this is a teaching snapshot, not a quote feed.
04 · Five Sleeve Strategies, With the Guardrails Load-Bearing

Structure the conviction; ration the lottery; cap the ruin.

All five strategies share one outer wall: the sleeve itself is capped at 5–10% of monthly investing dollars — Taleb's barbell applied to a paycheck. The other 90–95% goes where it always should (index funds, T-bills, the 401(k) match first). Nothing that happens inside the sleeve can touch the core: gains above a ratchet threshold sweep out to the core automatically, and losses stop at the sleeve's floor because event contracts are fully collateralized — no margin, no leverage, no debt. Within that wall, the five archetypes take genuinely different risk shapes:

S1 · Calibrated Carry — sell the lottery, don't buy it

Hold 65–90¢ favorites in liquid, short-to-mid-dated markets: the Fed-no-change contract at 74.5¢, hottest-year-on-record at 67.7¢, Democrats-take-the-House at 87.5¢. The favorite-longshot bias means favorites are the systematically underpriced side of the book, and calibration data says this is the one corner where retail expectation is near zero or better before fees. Returns are small, frequent, and boring on purpose — this is the strategy for the 26% who claim betting is their long-term plan: it forces the discipline of being the house-side of everyone else's lottery.

S2 · Diversified Event Ladder — the index fund of event risk

Eight to twelve mid-probability (20–65¢) contracts spread across at least five categories with expiries laddered from weeks to a year: September Fed strikes, October payrolls, year-end index ranges, midterm Senate control, hurricane counts, an AI-lab race market. Position cap 10–15% of sleeve per contract, category cap ~30%, liquidity floor mandatory. Uncorrelated binaries mean the book's variance falls with breadth — the closest thing this asset class has to diversification, and the natural default archetype.

S3 · Convex Tail Book — the fat tail, rationed

The behavioral research (Barberis's lottery-preference work) is clear: kill the jackpot and users defect back to parlays. So S3 keeps it — rationed. A hard-capped 100% of this small sub-sleeve goes to sub-10¢ contracts with plausible mispricing and huge payoffs: the S&P-below-4,000 crash tail at 2¢ (a 50× payoff that is also a genuine hedge on a Gen Z tech paycheck), Bitcoin-above-$200k at 3¢, OpenAI-declares-AGI at 8¢, recession-in-2026 at 6¢. The calibration curve says these are overpriced on average — the simulator below shows that cost transparently. The point is not that tails are good EV; it is that 20% of the sleeve in capped tails is survivable convexity, while 100% of a betting budget in parlays is a wealth transfer.

S4 · Barbell Sleeve — carry funds the tails (flagship)

The recommended default: ~80% Calibrated Carry + ~20% Convex Tail Book inside the same sleeve. The carry book's small favorite-side drift claws back much of the tail book's expected bleed: in the calibrated model below, the whole sleeve carries an expected cost of roughly 12% of sleeve dollars a year — about 0.6–1.2% of total savings at a 5–10% sleeve — versus the ~37% the sportsbook pattern extracts from the same dollars, while keeping a live shot at a 10–50× event. This is the structure that most honestly replicates what a Gen Z bettor says they want ("a real chance at getting unstuck") at a structural cost of roughly nothing, versus the ~10–30% annual tax the sportsbook charges for the same feeling.

S5 · Household Hedge — negative beta to your own life

The only archetype where a fiduciary case is clean, because it's insurance, not speculation: contracts chosen to pay out when the holder's finances get hit. Recession-in-2026 (6¢) pays when entry-level hiring freezes. The Fed-hike contract (24.5¢) pays when variable-rate debt gets more expensive. Rent-and-rates exposure would naturally sit in the CPI-above-3% contract, but that book quotes 76/92 — a 16¢ spread the liquidity guardrail refuses to cross — so the zero-cuts-in-2026 contract (85¢) carries the inflation-stays-hot leg instead. The Nasdaq-below-19,000 tail (5.5¢) pays into a tech-salary layoff scenario. A young renter with student debt has short exposure to exactly these prints; this book flips a slice of it long.

The guardrail stack — defaults, not prompts

The gambling-harm literature is unambiguous about what works: voluntary limit-setting prompts do not reduce net losses (a 4,300-user RCT found statistically identical losses vs. control), while hard caps, forced breaks and defaults do (mandatory 60-minute breaks raised same-day deposit stopping from 27% to 68%; "Save More Tomorrow" auto-escalation quadrupled savings rates by making the good path the default). Every rule below is therefore a default or a hard stop, not a nudge: (1) sleeve ≤10% of monthly investing dollars, auto-funded on payday, never topped up mid-month; (2) ≤10–15% of sleeve per contract; (3) ≤30% per category; (4) sub-10¢ longshots rationed to ≤20% of sleeve; (5) liquidity floor — no crossing spreads >3¢, no books thinner than ~2,000 OI; (6) no multi-leg/parlay contracts, ever — the 20–30% hold products exist to un-diversify you; (7) ratchet: sleeve balance above 2× annual budget sweeps to the index core; (8) circuit breaker: sleeve down 50% of annual budget → 30-day trading pause; (9) CFTC-regulated venues only — segregated funds, position limits, surveilled books; (10) maker-first order routing — resting limit orders pay ~zero fees on both venues and never cross a thin spread.

04b · The Sleeve Builder

Build it against the live book

The builder below constructs each strategy from the Figure-3 snapshot, checks it against the guardrail stack in real time, and Monte-Carlos a year of outcomes — using the calibration-adjusted probabilities from Figure 2, not the quoted prices, so the favorite-longshot tax and taker fees show up honestly in the distribution. Compare the sleeve's floor against the sportsbook counterfactual under every scenario: that difference is the whole argument.

Figure 4 · Strategy constructor — live snapshot, simulated year
Monthly budget deploys across real contracts; 3,000 simulated years; contracts assumed held to resolution, taker fees paid on entry where applicable.
$100 / mo
$1,000 / mo
PositionVenueYES @Weight$ / moPayoff if hitCalib. P(hit)

Guardrail check

One simulated year — where the sleeve ends up

Distribution of sleeve ending value ÷ total contributed, across 3,000 simulated years. True hit probabilities = quoted price adjusted by the Figure-2 calibration curve (longshots decay toward ~45% of quoted price below 10¢; favorites get a small boost); Kalshi taker fee 0.07·P·(1−P) charged on entry; positions held to resolution; monthly budget redeployed at the same book. The dashed markers show the same money's expected resting place in the two counterfactuals: re-betting each month's budget weekly at 2025's blended 10.2% sportsbook hold erodes ~37% of it before month-end (0.898^4.3 ≈ 0.63), while the boring index core sits near +7% real. Simplifications: independence across contracts (real books correlate — recession × Fed × index tails), static prices, no early exit, no reinvestment of interim wins. This is a teaching model, not a backtest.
What the simulator is honest about

Run S3 (pure tails) and look at the median: you usually lose most of the sleeve. The fat right tail is real — a 50× crash-tail hit dwarfs a year of contributions — but the mode of the distribution is bleed. That is exactly the trade Gen Z is currently making blind inside parlay slips at 10× the structural cost, with no position caps, no ratchet, and no floor. The sleeve doesn't make speculation smart; it makes the price of the lottery ticket visible, small, and survivable — while S1/S4 show that a disciplined book can hold the line near zero and still keep tail exposure alive.

05 · Who Manages This? The Missing Middle

Not a robo. Not a fund. A CTA wearing robo UX — and nobody has built it yet.

The natural instinct is "robo-advisor with a discretionary component" — Betterment-for-event-contracts. The regulatory plumbing says otherwise, and the difference is the most interesting finding in this piece. Event contracts are CFTC-regulated futures. The moment anyone exercises discretion over a client's event-contract account for compensation, they are a Commodity Trading Advisor (CTA); pool the money and they are a Commodity Pool Operator. That is the managed-futures regime — NFA membership, Series 3 principals, disclosure documents — not the Advisers Act regime robos live in. And the accounts themselves can't even cohabitate: CFTC-regulated event contracts must sit in a separate futures account at an FCM, walled off from the FINRA-regulated brokerage account holding the index core. A "robo with a sleeve" is, mechanically, two regulators, two account types, and two rulebooks stitched together at the app layer.

Three buildable lanes emerge, in ascending order of discretion:

Figure 5 · Three lanes to a managed sleeve — and the gap between them
Every component on the board exists today except the middle lane's assembled product.
Lane 1 — The Wrapper: prediction-market ETFs
Roundhill, Bitwise and GraniteShares have 24+ event-contract ETFs filed (Senate/House 2026, the 2028 race); the SEC delayed them in May 2026 pending disclosure review. If approved, the '40-Act wrapper re-securitizes the futures — a robo can then allocate a 2–5% satellite exactly like any ETF, zero CFTC contact. Cleanest path; least Gen Z appeal (the wrapper kills the "my call on the world" ownership that drives the behavior).
SEC pending
Lane 2 — The Guardrailed Allocator: rules engine + advisory overlay (the buildable one)
Self-directed execution on a regulated rail (the Apex/Kalshi stack already lets a fintech offer event contracts via an NFA introducing broker with Apex Clearing as FCM), wrapped in client-preauthorized standing rules: the ten guardrails of Section 04 as product defaults — sleeve cap enforced at deposit, position/category caps at order entry, longshot rationing, auto-ratchet to the index core, circuit breakers. An affiliated RIA advises the 90% core and treats the sleeve advice as incidental (CFTC Reg 4.14(a)(8) exempts RIAs whose commodity advice is solely incidental to securities advice). No discretion over individual contracts — the user picks the contracts; the structure picks what they can't do. This is Save-More-Tomorrow architecture, not portfolio management, and it is buildable today with zero novel relief.
Buildable now
Lane 3 — Full discretion: the retail micro-CTA
True "manage it for me" — an adviser runs the sleeve, selects contracts, rolls expiries. Requires CTA registration (or a pooled vehicle under a CPO, where the 4.13(a)(3) de-minimis exemption is unusually easy to satisfy since fully-collateralized event contracts use no margin). The mechanics exist — Kalshi's own FCM affiliate (confusingly named Kinetic Markets, no relation) enables institutional margin; quant desks and AI agents already run books. What's missing is anyone willing to be first through the fiduciary door.
Legal chassis exists
The gap — why the middle is empty
Three blockers, none of them registration mechanics: (1) Tax character is unresolved — the IRS has issued no guidance on whether DCM event contracts get §1256 60/40 treatment; platforms don't even issue transaction-level 1099s. A robo built on tax-loss harvesting cannot run on an asset whose basic tax character is unknown. (2) Rebalancing doesn't exist here — contracts expire to $0 or $1 in days-to-months; "allocation" is really continuous contract selection, which is discretion, which is the CTA line. (3) Fiduciary optics — recommending a ~zero-EV-minus-fees instrument as a strategy is hard to paper (Whelan: only ~7–13% of human prediction-market traders are profitable); recommending it as structured harm reduction for speculation already happening is a different and defensible document, but nobody has written it yet.
Unbuilt
Grey-zone prototypes already surround the gap: Polymarket copy-trading tools (Polycopy, PolyTrack et al.) are functionally unregistered CTAs on the offshore books; PolyFund runs pooled smart-contract vaults with performance fees — an unregistered CPO in code; and 30%+ of Polymarket wallets show AI-agent activity, with agents showing ~37% positive-P&L rates vs ~7–13% for humans. The demand for a managed layer is being met — outside the perimeter, without guardrails.

So: robo, discretionary, or something new? Something new, with a precise shape. Near-term, the answer is Lane 2 — a non-discretionary rules engine that automates the guardrails rather than the picks, attached to an RIA advising the core. It is not a robo-advisor (no discretion, no rebalancing, wrong regulator for the sleeve) and not a CTA (the user keeps contract selection). It is closer to a 401(k) plan document for speculation: a structure the saver adopts once, that then makes the destructive versions of their own behavior unavailable. The discretionary component arrives later, in one of two costumes — the ETF wrapper (if the SEC relents) for the passive version, or a registered micro-CTA (plausibly AI-driven, given the agent P&L data) for the active one. The window for whoever builds Lane 2 first is real: Robinhood's event revenue just crossed its equities revenue, the CFTC is suing states to defend the perimeter, and 52% of a generation's investing dollars are already in motion.

06 · Sources & Method

Sources