DraftKings Predictions Guide for Agents

DraftKings Predictions is CFTC-regulated event contract trading built by the largest U.S. sportsbook operator. This guide covers the platform’s exchange infrastructure, regulatory structure, market categories, developer access options, and how AI agents can integrate DraftKings Predictions data into cross-platform trading strategies.

What DraftKings Predictions Is

DraftKings Predictions is a standalone event contract platform that launched on December 19, 2025. Unlike the DraftKings Sportsbook — which operates as a traditional bookmaker under state gaming licenses — DraftKings Predictions functions as a federally regulated exchange under CFTC oversight. This regulatory distinction is the entire strategic play: DraftKings Predictions is available in 38 states, including massive non-sportsbook markets like California, Texas, Georgia, and Florida.

The platform lets users trade binary yes/no contracts on event outcomes. If you buy a “Yes” contract on whether the Lakers will win tonight and they do, the contract settles at $1.00. If they lose, it settles at $0.00. Your purchase price reflects the market-implied probability.

DraftKings’ subsidiary operating the Predictions product is a CFTC-registered Introducing Broker and Member of the National Futures Association. This means user funds and contract settlements follow commodities law, not state gambling regulations — a structural advantage DraftKings shares with Kalshi but not with offshore sportsbooks.

The Railbird Acquisition and Exchange Architecture

The exchange infrastructure behind DraftKings Predictions is evolving in phases. Understanding the current state matters for developers building agents that interact with the platform.

Phase 1: CME Group (Launch — Current)

At launch in December 2025, DraftKings Predictions connected to the CME Group exchange as its primary order book. This means all trades route through CME-owned infrastructure. CME Group also powers FanDuel Predicts, which creates a notable dynamic: DraftKings and FanDuel users can fill each other’s orders on the same underlying exchange.

For agent builders, this means pricing on DraftKings Predictions and FanDuel Predicts may be nearly identical on shared markets. Cross-platform arbitrage between these two is unlikely on CME-settled contracts — but arbitrage against Polymarket or Kalshi remains viable because those platforms run independent order books.

Phase 2: Crypto.com Integration (February 2026)

In February 2026, DraftKings partnered with Crypto.com Derivatives North America to add player-specific sports event contracts — a first for the platform. These contracts cover individual player performance metrics (points, touchdowns, strikeouts) similar to player prop bets but structured as exchange-traded event contracts.

The Crypto.com partnership also established the pipeline for future category expansion into culture, entertainment, and politics.

Phase 3: Railbird Exchange (Expected Mid-2026)

DraftKings acquired New York-based Railbird Technologies in October 2025 specifically because Railbird held a CFTC designation as a registered designated contract market (DCM). Once Railbird Exchange is fully integrated — projected for mid-2026 — DraftKings will control its own exchange infrastructure: order matching, market listing, pricing, and settlement.

This is the phase that matters most for agent developers. When DraftKings owns the exchange stack, it can set its own rules around API access, data feeds, and potentially programmatic trading. Whether DraftKings will open this infrastructure to third-party developers remains unannounced.

Acquisition Economics

DraftKings reportedly paid $50 million upfront for Railbird with an additional $200 million in performance-based incentives. Railbird’s CEO Miles Saffran and his team joined DraftKings to lead the exchange buildout. The deal logic tracks with DraftKings’ broader strategy of owning its technology stack — the company has historically moved from third-party infrastructure to in-house systems across DFS, sportsbook, and iGaming. DraftKings CPO Corey Gottlieb framed it explicitly: every major product line eventually migrates to vertical integration to maximize product control and margin.

The performance incentives suggest DraftKings tied the payout to Railbird’s successful integration and volume milestones. For agent builders, this means DraftKings is financially incentivized to make Railbird work — it’s not a side project.

2026 Product Roadmap

DraftKings Predictions SVP and General Manager Jeanine Hightower-Sellitto laid out a detailed roadmap at the March 2, 2026 Investor Day. The milestones, in order:

The in-house futures commission merchant (FCM) and derivatives clearing organization (DCO) are the final pieces. Once DraftKings controls clearing, it eliminates dependency on third-party exchanges entirely. This is the same playbook DraftKings ran in sportsbook: start with third-party odds feeds, then build proprietary pricing, then own the entire stack.

DraftKings excluded all Predictions revenue from its $6.5–$6.9 billion 2026 guidance, meaning any Predictions revenue is pure upside to current estimates. CEO Jason Robins described Predictions as “the most exciting new growth opportunity we have seen since PASPA was struck down in 2018.”

Market Categories

DraftKings Predictions currently offers contracts across three primary categories, with more planned.

The sportsbook-style presentation is deliberate. DraftKings Predictions displays contracts using moneyline-style pricing rather than probability percentages or market prices, making the interface familiar to its existing sportsbook user base. Developers working with the data should note that converting between moneyline odds and implied probability is required for any quantitative analysis.

Sports event contracts are not available in every state where DraftKings Predictions operates — some states only permit financial and crypto contracts. Agent builders targeting specific markets need to account for this geographic variability.

State Availability Details

DraftKings Predictions launched in 38 states, but the product availability is tiered. Sports event contracts from Crypto.com are currently available in roughly 17 states. Financial and crypto contracts are available more broadly. States where DraftKings already has a sportsbook license create a complicated dynamic: some state regulators have signaled that offering sports event contracts through a federal pathway could jeopardize a company’s state gaming license.

The states where DraftKings Predictions offers the most strategic value — California, Texas, Georgia, Florida — are precisely the states where DraftKings has no traditional sportsbook presence. These markets represent roughly 40% of the U.S. population that was previously unreachable. For agent builders, this geographic data matters when modeling total addressable liquidity per market.

Currently absent from DraftKings Predictions: Maine, New Hampshire, and Ohio.

The Shared Exchange Problem

A critical detail for arbitrage-focused agents: DraftKings and FanDuel both route contracts through CME Group. When limit orders launch on both platforms, a DraftKings user and a FanDuel user can fill each other’s orders on the same underlying exchange. This means identical markets on DraftKings Predictions and FanDuel Predicts share the same order book and the same pricing.

This shared infrastructure eliminates the most obvious arbitrage path (DraftKings vs. FanDuel on CME markets) but creates a different opportunity. Crypto.com-sourced markets on DraftKings are not shared with FanDuel — they run on a separate exchange. An agent monitoring both platforms can detect pricing gaps between CME-sourced and Crypto.com-sourced contracts covering the same underlying event.

The practical implication: build separate price feeds for CME-settled contracts (shared) and Crypto.com-settled contracts (DraftKings-exclusive) and look for divergence between the two.

Moneyline-to-Probability Conversion

Since DraftKings Predictions displays contracts in moneyline format, any quantitative agent needs a conversion layer. Here’s the math:

def moneyline_to_implied_prob(ml: int) -> float:
    """Convert American moneyline odds to implied probability."""
    if ml > 0:
        return 100 / (ml + 100)
    else:
        return abs(ml) / (abs(ml) + 100)

def remove_vig(prob_yes: float, prob_no: float) -> tuple[float, float]:
    """Remove overround to get fair probabilities."""
    total = prob_yes + prob_no
    return prob_yes / total, prob_no / total

# Example: Lakers Yes -150, No +130
yes_prob = moneyline_to_implied_prob(-150)  # 0.600
no_prob = moneyline_to_implied_prob(130)    # 0.435
fair_yes, fair_no = remove_vig(yes_prob, no_prob)
# fair_yes ≈ 0.580, fair_no ≈ 0.420
# Overround: 3.5% (the vig DraftKings/exchange collects)

Compare this fair probability against Polymarket prices (which are expressed as 0–1 probabilities directly) to spot cross-platform edge. A systematic agent monitors this delta continuously and flags markets where the DraftKings-implied probability diverges from the exchange-cleared price by more than transaction costs on both sides.

For deeper vig analysis methodology, see the Vig Index and the DraftKings vs. FanDuel vs. BetMGM odds comparison.

Developer and API Access

DraftKings does not offer an official public API for its Predictions platform. This is a meaningful gap compared to Polymarket (which has open CLOB, Gamma, and Data APIs) and Kalshi (which has a documented REST API with programmatic execution).

Third-Party Data Access

Developers currently access DraftKings Predictions data through several third-party channels:

OpticOdds provides structured JSON feeds covering DraftKings sportsbook odds and is expanding coverage to Predictions markets. This is the most reliable option for real-time price monitoring across multiple sportsbooks simultaneously — useful for cross-platform vig analysis and arbitrage detection.

Apify hosts a DraftKings Predictions scraper actor that extracts market data across sports and financial categories. This is suitable for periodic snapshots but not real-time monitoring.

Unofficial endpoints exist through DraftKings’ internal API (documented by community projects like the draftkings_client Python library), but these are unsupported, undocumented, subject to change without notice, and currently focused on DFS/sportsbook data rather than Predictions.

What’s Missing for Agents

The current DraftKings Predictions architecture has several gaps that limit autonomous agent integration:

• No programmatic execution: There is no API for placing or managing trades. All execution is manual through the app or web interface.
• No order book access: Unlike Polymarket’s CLOB API or Kalshi’s REST API, there is no way to query the order book depth, pending orders, or historical fills.
• No WebSocket feeds: Real-time price streaming is not available to external developers.
• No account management API: Position queries, balance checks, and settlement status are not programmatically accessible.

For agent builders today, DraftKings Predictions is a data source, not an execution venue. The practical workflow is: monitor DraftKings pricing via third-party feeds, detect edge against Polymarket or Kalshi prices, and execute on the platform that supports programmatic trading.

The Super App and Vertical Integration

At its March 2, 2026 Investor Day, DraftKings announced the Super App — a unified platform combining sportsbook, DraftKings Predictions, online casino, and iLottery (via its Jackpocket acquisition) into a single interface. The app dynamically adapts based on state availability: users in California see Predictions and casino only, while users in New Jersey see the full stack. Phase one rolled out during March Madness 2026, with additional upgrades planned throughout the year.

DraftKings projects prediction markets represent a $10 billion annual gross revenue opportunity, within a broader industry TAM of $55–$80 billion in gross revenue by 2030. The company is building a proprietary market-making division that leverages its existing sportsbook infrastructure — hundreds of data scientists and ML engineers who already model sports probabilities daily for the traditional book. This data science team now also prices prediction market contracts, giving DraftKings a structural advantage over pure-play prediction exchanges that rely entirely on peer-to-peer liquidity.

The Dual Revenue Model

DraftKings Predictions operates on a dual revenue model that differs from both traditional sportsbooks and pure exchanges:

Transaction fees: As an Introducing Broker, DraftKings earns a cut of every trade it routes to CME Group, Crypto.com, and eventually Railbird. This is margin on flow.

Proprietary market-making: DraftKings’ in-house market-making division will provide liquidity by quoting both sides of contracts. The spread between buy and sell prices is the margin. This is the same model DraftKings uses in its sportsbook (where the “vig” is effectively a market-making spread), now applied to event contracts.

The combination means DraftKings is not purely a broker (like a retail brokerage routing to an exchange) nor purely a bookmaker (like a traditional sportsbook taking the other side). It’s both — earning from flow routing and from providing liquidity.

For agent builders, the market-making arm is the higher-signal component. DraftKings-provided liquidity means tighter spreads and higher fill rates compared to thin peer-to-peer markets. But it also means DraftKings’ internal models set the pricing baseline, creating potential edge for agents that can identify when those models lag behind breaking information.

AI and Engineering Infrastructure

DraftKings reported a 40% improvement in engineering productivity year-over-year driven by AI tooling, and a 100% increase in AI-assisted development. The company’s data science stack — built over a decade of sportsbook operations — is now a competitive moat in prediction markets. DraftKings models sports probabilities daily across thousands of markets, and this existing infrastructure is being repurposed to price event contracts.

This matters for agents because DraftKings’ pricing is model-driven, not crowd-driven. When DraftKings’ ML models update slower than peer-to-peer markets (which react to Twitter/X, breaking news, and sharp money), a latency edge exists. Conversely, DraftKings’ models may occasionally lead the market when they incorporate proprietary data (injury reports, weather models, team analytics) that hasn’t yet been priced into exchange markets.

Agent Integration Strategies

Given the current platform limitations, here are the viable agent architectures involving DraftKings Predictions.

Cross-Platform Arbitrage Monitor

Build an agent that monitors DraftKings Predictions prices alongside Polymarket and Kalshi prices for the same underlying events. When DraftKings’ moneyline-implied probability diverges from exchange prices by more than the combined vig, execute on the platform with open API access.

This works because DraftKings’ ML-driven pricing may lag peer-to-peer markets that react faster to breaking news and social sentiment. The three-way comparison page covers the structural differences in pricing mechanics.

Vig Analysis and Market Efficiency Research

Use DraftKings Predictions pricing data as an additional input to vig analysis workflows. Because DraftKings prices contracts using sportsbook-style models rather than pure market clearing, comparing DraftKings’ implied probabilities against exchange-cleared prices reveals where the bookmaker model and the crowd disagree.

Signal Extraction for Sportsbook Betting

DraftKings Predictions contracts on financial events (Fed decisions, S&P targets) settle on hard data — there’s no “house edge” interpretation involved. Agent builders focused on sports betting can use financial market contract prices as macro-economic sentiment signals. For example, a sharp move in rate-cut probability contracts may correlate with shifts in sports betting handle or sharp money flow.

Regulatory Architecture

Understanding DraftKings Predictions’ regulatory structure matters for agents operating across jurisdictions.

Federal vs. State: The Core Tension

The Trump administration has treated event contracts as commodities, placing them under CFTC jurisdiction rather than state gaming commissions. This is what allows DraftKings Predictions to operate in 38 states — including states that have not legalized sports betting. The legal theory: prediction market contracts are financial derivatives, not wagers.

This framework is actively contested. Over a dozen state regulators are challenging the legality of prediction markets within their borders, arguing that sports event contracts are functionally identical to sports bets. The outcome of these legal battles will directly affect market availability.

For agent builders, the regulatory risk manifests as geographic instability. A state could restrict sports event contracts tomorrow, removing an entire market category overnight. Building agents that depend on DraftKings Predictions availability in a specific state carries this tail risk. Financial and crypto contracts face less regulatory pressure because they don’t directly overlap with state-licensed sports betting.

The Introducing Broker Structure

DraftKings Predictions operates as an Introducing Broker (IB), not as the exchange itself. In commodities law, an IB solicits orders from customers and routes them to an exchange for execution. The IB earns commissions on flow but doesn’t hold customer funds directly or match orders.

This structure is important for agent builders because it means DraftKings does not control the order book on CME-settled or Crypto.com-settled markets. DraftKings cannot create custom market types, set margin requirements, or offer API endpoints that aren’t exposed by the underlying exchange. Once Railbird integrates, DraftKings will function as both IB and exchange operator — a vertically integrated structure that gives it far more control.

Tax Treatment

DraftKings Predictions profits are treated like capital gains under commodities law, not gambling income. For systematic agents tracking positions across prediction markets vs. sportsbooks, this distinction affects tax reporting. Prediction market gains/losses may be subject to Section 1256 contract treatment (60/40 long-term/short-term capital gains split), though specific treatment depends on the contract type and taxpayer situation.

DraftKings Predictions vs. Polymarket and Kalshi

The competitive landscape is detailed in the DraftKings vs. Polymarket comparison and Kalshi vs. DraftKings Predictions comparison. The condensed version:

For the Agent Betting Stack, DraftKings Predictions maps to Layer 3 (Trading) but with significant constraints at the execution layer. Agents can read DraftKings data (via third parties) but cannot write — no placing orders, no managing positions, no automated settlement collection.

What to Watch

Several developments in 2026 could materially change DraftKings Predictions’ relevance for agent builders:

Railbird Exchange integration (expected mid-2026): When DraftKings controls its own exchange, API access becomes a product decision rather than a technical limitation. If DraftKings opens even basic read APIs on Railbird, it becomes a first-class data source for prediction market agents.

Market-making division launch: DraftKings’ internal market maker will generate proprietary pricing that diverges from peer-to-peer exchange prices. This creates a persistent arbitrage surface that agents can monitor.

Super App cross-selling: As DraftKings funnels sportsbook users into Predictions (and vice versa), trading volume and liquidity should increase. Higher liquidity means tighter spreads and more reliable pricing signals for agent consumption.

Regulatory battles: Multiple states are challenging whether CFTC-regulated prediction markets can legally offer sports-related contracts within their borders. If state regulators succeed in restricting sports contracts, DraftKings Predictions’ market breadth narrows — but its financial and crypto categories remain unaffected.

Mapping to the Agent Betting Stack

The most constrained layer for agents is Layer 3. Without programmatic execution APIs, DraftKings Predictions cannot serve as an autonomous execution venue. Agent builders should treat it as a pricing signal source and execute on platforms with open trading APIs.

For the full Agent Betting Stack architecture, see how each layer connects across platforms. For sport-specific agent strategies on DraftKings, see the NFL bot, NBA bot, and MLB bot guides.

Related DraftKings Content

• FanDuel Predicts: Event Contracts Guide — the other CME Group-powered prediction market platform
• DraftKings Prop Bets Guide — market depth, SGP+, Flash Bets, vig analysis, and betting limits on the sportsbook side
• DraftKings Live Odds Comparison — real-time DraftKings sportsbook odds vs. other books
• DraftKings Sportsbook Review — full sportsbook review and feature breakdown
• DraftKings vs. FanDuel vs. BetMGM Odds Comparison — head-to-head sportsbook odds and vig analysis