Per-Venue Edge Reality on $5,000

Why this post exists

A common reader question: "If I have $5K and want to test prediction-market trading, where should I actually put it?" The internet answer is "Polymarket and trust me." The honest answer requires venue-by-venue accounting of fees, oracle architecture, incentive programs, and US legality. We did the accounting and we're publishing the sheet.

The thesis: $5K of experimental capital, deployed across a structural-edge portfolio, returns 6–14% in the base case and 15–25% in an optimistic case. That's roughly $300–700 base, $750–1,250 optimistic. Anyone quoting more than that is selling a story.

The 2026 venue landscape — what's actually US-legal

The CFTC + state-by-state regulatory map shifted materially in 2025–2026. The current state:

Venue	Status	Key 2025–2026 change
Polymarket-US	Live, CFTC-licensed (QCEX)	Relaunched 2025-12-02. 0.30% taker / 0.20% maker rebate.
Kalshi	Live, CFTC-licensed	Liquidity Incentive Program (LIP) expires 2026-09-01.
IBKR ForecastTrader	Live	Launched 2026-05-14. Unifies Kalshi + CME + ForecastEx in one IBKR account. ~3.14% APY coupon on held collateral.
Bitnomial	Live (CFTC no-action Jan 2026)	Crypto + econ event contracts.
Coinbase Derivatives (CFM)	Live	US perpetual futures launched July 2025. 0.02%/contract, 10× leverage.
Robinhood / MIAXdx	Acquisition closed Jan 2026	Integration not yet customer-facing.
dYdX (US spot)	Live Dec 2025	Limited symbol coverage.
Limitless	Off-shore but accessible	100% maker rebate. Points / airdrop S2 program.
Hyperliquid / Deribit	US-blocked	VPN access violates ToS — do not use.
DraftKings / PrizePicks / DFS	Live	No algorithmic API — manual entry only.

Per-venue edge profile

Each venue has a structurally different edge profile because its fees, oracle, and incentive program differ.

Polymarket-US — primary structural-arb venue

The 0.20% maker rebate is the big news. It's the lowest taker cost in the regulated US prediction-market space and the first time a US venue has paid makers since pre-Robinhood Bitfinex. Combined with neg-risk market structure (where multi-outcome contracts trade as a single complex instrument with internal arbitrage), PM-US is the natural home for:

• E07 maker harvest — the rebate alone is the EV.
• E04 cross-venue arbitrage — same contract, Kalshi vs PM-US, fee differential pays for the trip.
• E15 oracle-lag — UMA's resolution oracle has a measurable settlement window.

Kalshi — fee-tier & oracle-lag venue (under LIP)

Kalshi's ~1.75% taker fee is high, but the Liquidity Incentive Program rebates most of it for qualifying volume. The LIP expires 2026-09-01 and is unlikely to renew on the same terms — so Kalshi's structural edge is dated. Best fit while LIP is active:

• E11 fee-tier optimization — cross tier boundaries strategically.
• E15 oracle-lag — NWS gridpoint weather, FRED nowcast.
• E16 settlement-window convergence — daily binaries with known resolution ts.

IBKR ForecastTrader — the institutional bridge

Unifying Kalshi + CME + ForecastEx in one account is operationally enormous. It enables cross-venue arbitrage that was previously a multi-day funding round-trip. ForecastEx pays ~3.14% APY on held collateral, which functions as a free yield on arb working capital. The downside: per-trade fees on IBKR are higher than direct-venue, so high-turnover strategies are punished.

• E04 cross-venue arbitrage — natural use case.
• Equity options: 30–45 DTE 20–25 delta SPX put credit spreads, the canonical retail-options structural edge.

Coinbase Derivatives — funding-rate arbitrage only

Coinbase's US perpetual futures have a published 8-hour funding rate. When the funding rate exceeds ~0.03%/8h (~33% APY), funding-rate arb against spot becomes profitable net of fees. This is the only edge class we currently trust on Coinbase Derivatives at $5K scale.

• E22 funding-rate arb — gated to funding rate > ~0.03%/8h.

Limitless — incentive farming

100% maker rebate. Points/airdrop S2 program. Off-shore so not the primary fleet venue, but a small allocation captures airdrop optionality with downside bounded by deposit size.

• E07 maker harvest
• E12 incentive farming

The recommended $5K portfolio

Below is the actual allocation we recommend for a $5K experimental seed. It's not the only allocation, but it's the one we'd run ourselves if we were starting from scratch today, and it matches what the proven-positive research wing supports.

Venue	Allocation	Strategy class	Expected gross
Polymarket-US	$1,500	Maker rebate + neg-risk arb	+8–18% base
Kalshi	$1,000	LIP volume optimization + oracle-lag (until Sep 2026)	+10–25% base
IBKR ForecastTrader	$1,500	SPX 30–45 DTE 20–25 delta credit spreads + cross-venue arb	+5–10% base
Coinbase Derivatives (CFM)	$500	Funding-rate arb (gated)	0 to +5% base
Limitless	$500	Maker + airdrop farming (S2)	0 to +40% optionality

Aggregated to $5K total, weighted by allocation, with realistic execution drag and ~5% standby cash:

• Base case: +6 to +14% per year, or $300–$700.
• Optimistic case (LIP keeps paying, Limitless S2 airdrop hits, funding rate elevated): +15 to +25%, or $750–$1,250.
• Downside (LIP cuts, no airdrop, funding flat, one bad regime month): –3% to 0%, or –$150 to $0.

This is not a yield product. This is an experimental capital allocation across genuine structural-edge venues. The reason we're publishing it is that nobody else does, and the absence of an honest baseline lets bot-vendors-on-Twitter quote 5–10× higher numbers without anyone calling it.

What we explicitly reject for $5K

• US equities — direct. Next-day direction has 0.498 IRL accuracy on our minute-bar corpus. Coin flip. The structural-edge alternative on equities is options (above).
• Cross-exchange spot crypto arb. Latency-contested down to roughly $0 capacity. Was viable in 2017–2020; not in 2026.
• Overtime on-chain sports — LP. LP fees on a $5K stake don't beat impermanent loss + gas. Forward-paper at higher capital first.
• Hyperliquid / Deribit. US-blocked. VPN is a ToS violation, not a strategy.
• DraftKings / PrizePicks DFS. No algorithmic API; manual-entry only. Not a programmatic edge.
• Naked options sold. Tail risk on $5K wipes the seed; never sell what you can't deliver.
• Anything with referral-affiliate compensation as a primary edge. Not the same thing as alpha.

How the allocation maps to the validation battery

Each line above is sourced from a strategy class that passes purged-CV in the proven-positive set (see the winner dissection) and is currently in forward-paper validation against the 26-method battery. Five strategies are battery-clean on Families A–D and forward-paper-positive, pending the 14-day SPAN gate (re-validation 2026-05-27). The allocation above assumes those five graduate. If they don't, we revise the allocation down — the same week.

The deeper point

A $5K trading seed is, candidly, a marginal use of capital. The same $5K deployed as marketing spend against a SaaS product with a 30% gross margin returns more reliably than the most optimistic forecast above. We continue the trading exploration not because the trading itself is the best ROI but because the methodology is the most rigorous machine we have for thinking about uncertainty under realistic constraints — and the methodology is the asset.

The trading P&L finances the methodology infrastructure. The methodology infrastructure underpins Validation-as-a-Service, the reproducibility datasets, and the public research log you're reading. That is the actual business. The $5K seed is the lab notebook.

If you're running your own $5K seed

• Verify the venue list against the current CFTC/no-action register before opening accounts.
• Start with paper-mode for at least 14 days per strategy. No exceptions.
• Track per-venue fees on every fill. Most "edge" disappears once fees are real.
• Allocate to structural, not directional. Reread the winners post.
• If you want a second pair of eyes on your strategy before deploying, see Validation-as-a-Service.