After Polymarket's $3.5B Election Boom, Prediction Markets Need a B2B Infrastructure Layer

TL;DR

The 2024 US presidential election was a watershed moment for prediction markets. Polymarket processed over $3.5 billion in handle and outperformed polls so convincingly that mainstream media started citing prediction market odds on live television. Kalshi reached a reported $11B valuation. But here's what nobody talks about at ICE or SBC: all that growth happened on consumer platforms, not on operator infrastructure. For the 6,000+ licensed iGaming operators worldwide, prediction markets remain a vertical they can see but can't easily touch — because the content infrastructure doesn't exist for B2B.

---

What Actually Happened After the Election Boom

Let's be honest about the state of things. Polymarket's election night was spectacular — their markets called the outcome faster and more accurately than any poll. But what happened after November 2024 was more revealing.

Trading volume dropped roughly 70% post-election. The casual bettors who'd discovered prediction markets through election coverage mostly didn't stick around. This wasn't a failure — it was exactly what you'd expect. Election markets are event-driven. The real question was whether the infrastructure, liquidity, and user base could find its next catalyst.

By mid-2025, the answer started becoming clear. Crypto markets on Polymarket rebounded. Kalshi expanded into sports-adjacent and entertainment categories. Metaculus continued growing in the forecasting community. And a new pattern emerged: the platforms that diversified their content survived the post-election hangover, while those dependent on a single category struggled.

For operators watching from the sidelines, this confirmed something important. Prediction markets aren't a fad — they're a content format. And like any content format, they need a steady supply of relevant, well-structured questions to sustain engagement.

The Operator's Dilemma

I've spent 14 years building sports betting platforms. When I talk to operators at conferences, the conversation about prediction markets follows a predictable pattern:

Phase 1 — Excitement: "We saw the Polymarket numbers. Our board wants us in this space."

Phase 2 — Reality check: "Wait, who creates the questions? Who sets the odds? Who resolves them? What happens when a question is ambiguous?"

Phase 3 — Shelved: "Let's revisit this in Q3."

The problem isn't demand. Operators can see their players are interested — many of them are already betting on Polymarket and Kalshi directly. The problem is content infrastructure.

A prediction market question isn't like a sports match where the event, participants, rules, and data feeds are all standardized. Every prediction market question is a unique content asset that requires:

• A clearly worded title that bettors can understand
• Unambiguous resolution criteria (this is harder than it sounds — "Will AI replace programmers by 2027?" is a terrible market question)
• A reasonable initial probability estimate
• Source data for resolution
• A mechanism for handling edge cases

Creating one good question takes effort. Creating thousands — across politics, entertainment, crypto, sports, technology, and local events — requires an entire content operation.

What the Numbers Actually Show

Rather than repeating inflated industry projections, here's what we can verify:

Polymarket (on-chain, verifiable): $3.5B+ cumulative handle through 2024. Post-election, monthly handle stabilized around $400-600M — still massive, and growing again through early 2026 as new events (European elections, crypto milestones, AI developments) create fresh catalysts.

Kalshi (regulated US exchange): Raised at a reported $11B valuation in February 2025 after winning its CFTC battle for election contracts. Volume figures aren't public, but the valuation tells you what institutional investors think about the market size.

Metaculus (forecasting community): 100,000+ registered forecasters, consistently ranked among the most accurate prediction sources. Their tournament-style approach has attracted corporate clients interested in decision-making under uncertainty.

The total addressable market for prediction-style betting through licensed operators? Nobody has a reliable number, and anyone who gives you one is guessing. But the directional evidence is strong: consumer platforms have proven demand, and operators have the distribution but not the infrastructure.

Why Operators Can't Just Copy Polymarket

Every few months, someone pitches me on "Polymarket but for regulated markets." It sounds obvious, but it misses the key structural difference.

Polymarket is a marketplace. Users create markets, provide liquidity, and trade with each other. The platform's role is primarily infrastructure — matching engine, settlement, custody. The content comes from the community.

Operators need something fundamentally different. They need a content feed — a stream of well-curated, properly structured prediction questions that they can present to their existing player base within their existing UI, risk framework, and compliance structure.

The differences are significant:

This is why operators need a B2B content layer, not a white-labeled consumer platform.

The Content Infrastructure Stack

At a technical level, a prediction market content feed for operators needs to solve four problems simultaneously:

1. Question Generation

The best prediction markets combine multiple content sources:

• Platform aggregation: Pull high-quality questions from Polymarket, Kalshi, Metaculus, and Manifold. Deduplicate across platforms using embedding similarity.
• News-driven generation: Monitor RSS feeds, financial data, sports calendars, and entertainment schedules. Use AI to generate structured questions from unstructured events.
• Operator-specific content: Let operators define their own question themes. "What will happen with the Champions League quarter-finals?" becomes a set of 20+ specific, tradeable questions.
• Local relevance: Questions about Finnish elections for Finnish operators. Questions about Premier League for UK operators.

2. Pricing

Consumer prediction markets use AMMs (Automated Market Makers) — typically LMSR (Logarithmic Market Scoring Rule) — where the platform provides initial liquidity and adjusts prices based on trading activity.

For operators, pricing needs to work more like a sportsbook: the operator sets odds (based on aggregated market data and probability models), manages their own book, and keeps the margin. This requires:

indicative_odds = aggregated_probability from multiple platforms
operator_margin = configurable spread (typically 5-15%)
displayed_odds = indicative_odds adjusted by operator_margin
exposure_limit = configurable per-market cap

3. Resolution

This is where most in-house attempts fall apart. Resolving "Will Bitcoin hit $100K by June 2026?" is straightforward — check a price feed. But resolving "Will the EU AI Act be fully enforced by 2026?" requires judgment.

A robust resolution system needs multiple layers:

• Data API resolvers: Automatic settlement for quantifiable outcomes (stock prices, weather, sports scores, lottery results)
• AI-powered research: For events that require verifying news sources and interpreting complex outcomes
• Consensus mechanisms: Cross-referencing multiple authoritative sources before settling
• Dispute handling: A process for when resolution isn't clear-cut

4. Risk Management

Prediction markets create risk exposures that differ from traditional sports betting:

• Markets can run for months (vs. hours for a football match)
• Correlated positions across related markets (if Trump wins, multiple markets resolve simultaneously)
• Information asymmetry (someone might know about a company merger before the public announcement)

Operators need position limits, exposure monitoring, and circuit breakers — familiar concepts, but applied to a new content type.

What's Actually Realistic for Operators in 2026

I'm not going to give you fantasy ROI projections. Instead, here's what I've seen work:

Small operators (under 50K monthly actives): Start with 50-100 active prediction markets, focused on sports-adjacent content that your existing audience already cares about. A pilot with a content feed provider costs far less than building anything yourself. Expect prediction markets to represent 2-5% of incremental handle within 6 months if you promote them.

Mid-size operators (50K-500K actives): You can afford a broader approach — sports, entertainment, financial markets. The audience is large enough that even niche categories find traction. Integrate prediction markets into your existing lobby as a new vertical alongside casino and sportsbook.

Large operators (500K+ actives): You might consider building some of this in-house — but even Bet365 and Flutter would benefit from content aggregation rather than creating every question manually.

The operators who will win in this space are those who move now while the category is still emerging, not those who wait for a fully mature market. First-mover advantages in content depth, audience development, and operational expertise compound over time.

Regulatory Landscape (March 2026)

The regulatory picture has clarified significantly since Kalshi's CFTC victory:

• United States: Kalshi's event contracts are live and regulated. State-by-state sports betting licenses don't automatically cover prediction markets, but several states are drafting frameworks. The CFTC's stance has become more accommodating of event contracts that aren't "gaming."
• European Union: Most EU jurisdictions regulate prediction-style betting under existing gambling licenses. The key question is whether event betting on non-sports outcomes requires additional authorization — most regulators say no if the operator already holds a general betting license.
• United Kingdom: The UKGC's position on "novelty betting" (their term for non-sports prediction markets) is well-established. Operators with standard betting licenses can offer these markets.
• Nordics: Estonia, Malta, and other Nordic gambling jurisdictions have relatively permissive frameworks for event betting.

The takeaway: if you hold a betting license in a major regulated market, you can almost certainly offer prediction markets under your existing authorization. Check with your compliance team, but the barriers are regulatory interpretation, not new licensing.

FAQ

How does prediction market content differ from sports data feeds?

Sports data feeds deliver standardized events (matches, players, leagues) with established rules. Prediction market content requires creating the events themselves — defining questions, criteria, and resolution mechanisms. That's why operators need a specialized content layer, not just another data feed.

What categories generate the most engagement?

Based on actual platform data: crypto and financial markets generate consistent daily volume. Politics spikes around elections. Entertainment (awards, reality TV, cultural events) brings in casual audiences. Sports-adjacent predictions (transfers, awards, records) engage existing sportsbook users most naturally.

Can I start with a pilot before full integration?

Yes, and that's the smart approach. Most content feed providers offer sandbox environments and limited market counts for testing. Start with 50-100 markets, measure engagement, then expand based on data.

What happens if a prediction market question is ambiguous?

This is the resolution problem, and it's the most operationally complex part. Good content feeds handle this through multi-source verification and predefined resolution criteria. The key is that criteria must be established when the market is created, not when it's time to settle.

How do prediction markets affect responsible gaming obligations?

The same responsible gaming frameworks apply — deposit limits, loss limits, self-exclusion, and reality checks. Some jurisdictions may require specific disclosures about the speculative nature of event betting. Prediction markets actually lend themselves well to responsible gaming because the binary nature makes risk clearer to players.

---

Last updated: March 2026