← Back to blog
·12 min read·SettleRisk Team

What Is Resolution Risk in Prediction Markets?

resolution-riskprediction-marketsfundamentals

Every prediction market trade carries two types of risk: the obvious directional risk (will the event happen?) and the often-overlooked resolution risk — the probability that the contract will not settle cleanly, on time, or as traders expect.

Most traders spend 100% of their analysis on directional conviction and 0% on whether the contract they are trading will actually pay out the way they think it will. That asymmetry is where alpha lives.

Why Resolution Risk Matters

When you buy a prediction market contract at $0.65, you're implicitly pricing two things:

  1. A 65% probability the event occurs
  2. An assumption that resolution will be straightforward

But what happens when resolution isn't straightforward? Capital gets locked. Disputes freeze positions. Settlement delays destroy the time value of your trade.

Consider a concrete example. You buy 10,000 shares of a Polymarket contract at $0.65, expecting resolution in 48 hours. Your expected profit is $3,500 (10,000 x $0.35). But a dispute is filed, and resolution takes 30 days instead of 2. Your $6,500 in capital is frozen for an entire month. At a 12% annual cost of capital, that lockup alone costs you $65 in pure opportunity cost — before accounting for the possibility that the dispute changes the outcome entirely.

Now multiply that across a portfolio of 40-50 positions, and resolution risk becomes the dominant factor in your realized returns versus your modeled returns.

The Three Sources of Resolution Risk

1. Ambiguous Resolution Criteria

The most common source of disputes. When a market asks "Will Bitcoin reach approximately $100K?" — what does "approximately" mean? Is $99,800 close enough? What about $99,500?

This is not hypothetical. In early 2025, a major prediction market ran a contract on whether a specific economic indicator would "significantly exceed" consensus expectations. The indicator came in at +0.3% above consensus. Was that "significant"? The resulting dispute froze $2.1M in open interest for 18 days.

SettleRisk's AMBIGUOUS_DEFINITIONS driver (base points: 22) detects these patterns and quantifies the risk they introduce. A related driver, SUBJECTIVE_OUTCOME_LANGUAGE (base points: 24 — the highest-weighted driver in our taxonomy), flags rules that use discretionary language like "significant," "material," "substantial," or "in our opinion." When we detect these tokens, we assign a minimum strength of 0.6 even in fallback mode, because subjective language is consistently the strongest predictor of post-resolution disputes.

2. Oracle Dependencies

Many markets depend on a single data source for resolution. If that source goes down, changes its methodology, or delays publication, the market can't resolve. We've seen government API outages delay resolution by 10+ days.

Real examples that have caused settlement delays:

  • BLS employment data — the Bureau of Labor Statistics revised preliminary jobs numbers by 818,000 in August 2024, retroactively changing what many traders thought was the "true" outcome for employment-based markets.
  • CoinGecko API — a 14-hour outage in 2024 made it impossible to confirm crypto price thresholds for dozens of active markets.
  • Election results — the AP and Edison Research occasionally differ on race calls; markets that specify one but not the other as their oracle create resolution ambiguity when the two disagree on timing.

The SINGLE_SOURCE_URL_DEPENDENCY driver (base points: 20) catches these single points of failure. When a market's rules reference exactly one resolution-critical URL and no redundancy group exists, we assign maximum strength. The complementary DATA_REVISION_RISK driver (base points: 12) detects when the cited source explicitly publishes preliminary data "subject to revision" — a common pattern with government statistics.

3. Jurisdictional and Temporal Ambiguity

"End of day" without a timezone. "Regulatory classification" without specifying which regulator. These gaps seem minor until they trigger multi-week committee reviews and freeze millions in capital.

Our TEMPORAL_AMBIGUITY driver (base points: 18) catches the temporal cases: cutoff times without timezone markers, "close of business" without specifying which business day calendar, and deadline references that could be interpreted as either inclusive or exclusive. The MULTI_JURISDICTION_INTERPRETATION driver (base points: 10) handles the regulatory side — markets that reference overlapping regulatory frameworks where the SEC, CFTC, or foreign regulators might reach conflicting conclusions.

For a deeper look at how each driver is detected and scored, see our full methodology documentation.

Resolution Risk by Market Type

Not all market structures carry equal resolution risk. The contract type itself shapes the attack surface for ambiguity.

Binary Markets

Binary markets ("Will X happen before date Y?") are the simplest structure, but they concentrate all resolution risk into a single yes/no judgment call. When that judgment is clean — "Will the Fed raise rates at the March meeting?" resolved by the FOMC statement — risk is low. When it is not — "Will AI cause a major cybersecurity incident in 2026?" — the entire contract value hinges on interpreting a single ambiguous criterion.

Typical risk score range: 8-45, depending on criteria specificity.

Categorical Markets

Categorical markets ("Which candidate will win?" or "Which team will win the championship?") introduce resolution risk at the boundary conditions. What happens to a "Who will be the next CEO?" market if co-CEOs are appointed? What if the answer is "none of the listed options"? The AMBIGUOUS_DEFINITIONS driver fires frequently on categorical markets because the set of valid outcomes is often under-specified.

Typical risk score range: 15-55, with tail risk from unlisted outcomes.

Scalar / Range Markets

Scalar markets ("What will GDP growth be in Q3?") are the most resolution-risk-dense structure. They depend on a specific numerical value from a specific source at a specific time — three potential failure modes stacked together. Preliminary vs. revised data, rounding conventions, and seasonal adjustment methodology can all shift the outcome across a range boundary.

Typical risk score range: 20-70, driven heavily by DATA_REVISION_RISK and SINGLE_SOURCE_URL_DEPENDENCY.

You can explore scored examples of each market type in our case studies.

Hidden Resolution Risk: What Most Traders Miss

Beyond the three primary sources, several edge cases catch even experienced traders off guard.

Retroactive Rule Changes

Some platforms reserve the right to modify resolution criteria after trading has begun. This is sometimes buried in platform-wide terms of service rather than stated in the market-specific rules. When a platform retroactively narrows the definition of an outcome — for example, changing "U.S. inflation" from headline CPI to core CPI after contracts are already trading — positions taken under the original interpretation can become worthless overnight. SettleRisk flags markets on platforms with histories of mid-market rule amendments and weights the PLATFORM_SPECIFIC_ORACLE_COMPLEXITY driver (base points: 8) accordingly.

Regulatory Intervention

Markets on regulated activities face the risk that a regulator halts or redefines the underlying event. Consider a market on whether a specific drug receives FDA approval by a deadline. If the FDA issues a Complete Response Letter (neither approval nor rejection), does the market resolve "No"? What if the company withdraws its application the day before the deadline to refile — is that a "No" or does the market void? These scenarios are common in biotech and financial regulation markets, and they are almost never addressed in the rules text.

Information Asymmetry in Resolution

Perhaps the most insidious edge case: resolution information that is technically public but practically inaccessible. Markets resolved by data behind paywalls (SOURCE_PAYWALL_OR_AUTH_REQUIRED, base points: 14), by non-archived dynamic dashboards (NON_ARCHIVED_SOURCE, base points: 12), or by government databases that require specialized query access create a two-tier system where some participants can verify outcomes and others cannot. This asymmetry can delay challenges and suppress legitimate disputes, or alternatively, enable bad-faith disputes by participants who know the data is hard to verify.

Cascading Dependencies

A market on "Will Company X's stock price exceed $200 by December 31?" seems simple, but consider: what if the company undergoes a stock split, a reverse merger, or is acquired for a mix of cash and stock? The rules may not address these corporate actions, and the resulting ambiguity can cascade through a portfolio of correlated positions. Our driver system detects these through a combination of AMBIGUOUS_DEFINITIONS and HISTORICAL_CATEGORY_DISPUTE_RATE_HIGH when the market falls into a historically contentious domain.

Learn more about edge case detection on our features page.

Quantifying Resolution Risk

At SettleRisk, we score resolution risk on a 0-100 scale across four tiers:

  • LOW (0-19): Clean criteria, reliable oracles, unambiguous timing
  • MEDIUM (20-49): Minor ambiguities, some oracle concentration
  • HIGH (50-74): Significant ambiguity, dispute probable
  • CRITICAL (75-100): Extreme ambiguity, high dispute probability, expect delays

The score is computed deterministically from a closed-form formula:

raw_points = BASE_POINTS(platform) + sum(driver_points) - mitigation + complexity
aggregate_risk_score = clamp(round(raw_points), 0, 100)

Platform base points reflect structural differences: Polymarket starts at 12 (on-chain oracle complexity) while Kalshi starts at 8 (centralized resolution). Each detected driver contributes base_points * strength * confidence, capped at the driver's maximum. A redundancy mitigation of 6 points applies when multiple resolution-critical URLs share a redundancy group, and a complexity adjustment adds up to 8 points when resolution criteria exceed 3 items.

Each score comes with up to 15 explainable risk drivers, a dispute probability estimate, and a lognormal settlement delay distribution. The dispute probability follows the formula:

p_dispute = clamp(0.01 + 0.003 * score + 0.05 * I(polymarket) + 0.06 * I(CRITICAL), 0, 0.90)

A Polymarket contract scoring 60 (HIGH tier) would have a dispute probability of 0.01 + 0.003 * 60 + 0.05 = 0.24 — roughly a 1-in-4 chance of a formal dispute.

From Risk Score to Fair Price

The real power of resolution risk quantification is pricing. If a market has a 24% dispute probability and an expected 18-hour settlement delay, those factors should be priced into your positions.

Our pricing engine computes:

  • Adjusted fair price — the input price (mid-price or p_event) used as baseline
  • Risk premium — the discount for bearing resolution risk, in basis points
  • Capital lockup cost — opportunity cost of frozen capital, in basis points
  • Fair spread — minimum spread to compensate for resolution risk

Worked Example

Let's walk through a concrete pricing calculation for a Polymarket contract with:

  • mid_price = 0.65
  • aggregate_risk_score = 60 (HIGH tier)
  • p_dispute = 0.24
  • expected_settlement_delay_hours = 18.0
  • base_spread_bps = 100 (market maker's base spread)
  • annual_capital_cost_apr = 0.12 (default)

Step 1: Capital lockup cost

expected_lockup_years = 18.0 / (24 * 365) = 0.002055
expected_capital_lockup_cost_bps = 0.12 * 0.002055 * 10000 = 2 bps (rounded)

Step 2: Resolution risk premium

resolution_risk_premium_bps = 5 + 40 * 0.24 + 20 * 1 + 0
                            = 5 + 9.6 + 20
                            = 35 bps (rounded)

The 20 * I(risk_tier == HIGH) term adds 20 bps because the score falls in the HIGH tier. The CRITICAL surcharge (60 bps) does not apply.

Step 3: Recommended spread

recommended_spread_bps = 100 + 35 + 2 = 137 bps

So a market maker who would normally quote a 100 bps spread should widen to 137 bps to compensate for the resolution risk on this contract. That is a 37% increase in required spread — material for any high-frequency market making strategy.

Step 4: Do-not-quote check

The engine also evaluates whether you should quote at all. If the risk tier were CRITICAL, the p99 delay exceeded your configured maximum, or any resolution-critical URL were currently offline, the API would return do_not_quote: true — an explicit signal to step away from the market entirely.

For details on integrating pricing into your trading systems, see our pricing page.

API Integration

You can evaluate any market's resolution risk with a single API call:

curl -X POST https://api.settlerisk.com/v1/evaluate-rules \
  -H "Content-Type: application/json" \
  -H "X-Vellox-Key-Id: your_key_id" \
  -H "X-Vellox-Timestamp: 1705334400" \
  -H "X-Vellox-Nonce: abc123" \
  -H "X-Vellox-Content-SHA256: sha256_of_body" \
  -H "X-Vellox-Signature: computed_hmac_signature" \
  -d '{
    "platform": "polymarket",
    "rules_text": "This market resolves YES if Bitcoin reaches $100,000 as reported by CoinGecko on or before March 31, 2026 at 11:59 PM."
  }'

For cached scores on markets we already track, the GET endpoints are even simpler:

curl https://api.settlerisk.com/v1/markets/polymarket/btc-100k-march/risk-score \
  -H "X-Vellox-Key-Id: your_key_id" \
  -H "X-Vellox-Timestamp: 1705334400" \
  -H "X-Vellox-Nonce: def456" \
  -H "X-Vellox-Content-SHA256: e3b0c44..." \
  -H "X-Vellox-Signature: computed_hmac_signature"

For batch operations, you can score up to 1,000 markets in a single request:

curl -X POST https://api.settlerisk.com/v1/risk-scores:batch \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: batch-20260115-001" \
  # ... auth headers ...
  -d '{
    "items": [
      {"platform": "polymarket", "platform_market_id": "btc-100k-march"},
      {"platform": "kalshi", "platform_market_id": "FED-RATE-MAR26"}
    ]
  }'

Batch responses use partial success semantics — the overall response is HTTP 200, and each item in the response carries either a result or an individual error, so a single bad market ID does not fail your entire batch.

Start Pricing Resolution Risk Today

Resolution risk is the largest unpriced factor in prediction markets. Every disputed contract, every delayed settlement, and every ambiguous ruleset represents alpha that flows to the traders who quantify it and away from those who ignore it.

Whether you are a market maker widening spreads on high-risk contracts, a directional trader avoiding capital lockup traps, or a quant researcher building resolution-adjusted fair value models, SettleRisk gives you the data to make better decisions.

Three ways to get started:

  1. Try the live demo — paste any market's rules text and see a scored breakdown instantly, no account required.
  2. Read the full methodology — understand exactly how every driver, score, and pricing output is computed. Our scoring is fully deterministic and reproducible.
  3. Integrate the API — production-ready REST and gRPC endpoints with HMAC-SHA256 authentication, batch support, and webhook alerts for score changes.

Check out our pricing plans for details on usage tiers and compute unit billing, or explore case studies to see how other teams have integrated resolution risk into their trading workflows.