# Changelog of Polymarket + Kalshi Arbitrage Opportunity Radar (`nexascout/prediction-market-opportunity-risk-radar`) Actor - **URL**: https://apify.com/nexascout/prediction-market-opportunity-risk-radar/changelog.md - **Full Actor documentation**: https://apify.com/nexascout/prediction-market-opportunity-risk-radar.md ## Changelog ### 2.8.2 — Output Console hotfix - Removed `clean=true` from Actor output URL templates. Apify Console parses output query parameters and expects `clean` as a boolean, while URL templates supply strings, causing the Output tab to fail with `Expected property clean to be of type boolean`. - Dataset views remain unchanged: Opportunity overview, Execution economics, and Semantic & resolution risk. - Matching engine, Tasks, pricing, and signal logic are unchanged from the validated V2.8 core. ### 2.8.1 — Store-ready polish - Reworked Store README around customer value, supported use cases, signal semantics, economics, monitoring, and limitations. - Added dedicated dataset views for opportunity overview, execution economics, and semantic/resolution-risk audit. - Expanded Actor output links to the three dataset views plus run diagnostics. - Reduced the default maximum output rows from 20 to 10 for safer exploratory cost. - Made cross-run monitoring opt-in by default. - Updated Actor title/description for clearer Store positioning. - Reworded the fee-enabled warning so it no longer implies fee math is unknown when economics can determine it. - Matching and contract-scope gates remain the validated V2.8 engine. ### 2.8.0 - Added a hard sports contract-scope gate: cross-venue sports economics is allowed only for `FULL_GAME` moneyline ↔ `FULL_GAME` moneyline. - Fixed quarter/half scoped props leaking through as full-game winner contracts even when their `contractScope` differed. - Added prop detection for longest field goal and both-teams-to-score style NFL derivatives. - Contract-type inference now prioritizes the contract-facing title instead of allowing the shared event title (`Team A vs Team B`) to force `SPORTS_MONEYLINE`. - Runtime markers bumped to `engineVersion=2.8.0` / `contract-semantics-v7`. ### 2.7.0 - Hardened sports contract scope so `1H/2H/quarter/inning moneyline` is never equivalent to a full-game winner market. - Added explicit `SPORTS_PERIOD_MONEYLINE` and `SPORTS_EXACT_MARGIN` semantics plus `canonical.contractScope`. - Added guards for winning-margin buckets and event props such as safety/overtime/coin-toss style contracts. - Runtime semantic guard now calls the same canonical sports detector instead of maintaining a second drifting regex implementation. - Pure league queries (for example `NFL`) skip noisy Polymarket keyword search and use league-native sports discovery. - Reduced pure-league discovery fan-out while preserving league event coverage; this also prevents hundreds of derivative contracts from reaching order-book analysis. - Added regression coverage for the false `1H Moneyline ↔ full-game winner` ACTIONABLE signal observed in live V2.6 testing. ### 2.6.0 - Added league-native Polymarket discovery via `/sports` metadata and `/events?series_id=...` with tag fallbacks for league-only queries such as NFL. - Focused sports queries no longer depend on noisy `public-search` to discover game-level markets. - Added league-targeted discovery counts/hints to run diagnostics. - Removed stale “fee math must be known” warnings after the fee engine has successfully resolved the Polymarket fee curve. ### 2.5.0 - Team-specific queries no longer widen to every market in the inferred league; `Yankees` filters to Yankees-related contracts instead of all MLB. - Macro vertical detection now uses contract-facing fields, preventing unrelated search-result/rule text from misclassifying IPO/general markets as `FED_DECISION`. - Focused query runs cap broad fallback ingestion while preserving targeted Polymarket public-search and Kalshi series discovery, reducing latency and API load. - Expanded sports prop detection for extra innings, first/last score and common stat/player props. - Improved two-team side inference when a platform duplicates the YES label into the secondary outcome field. - Added Polymarket AUTO fee-curve economics using category coefficients and per-fill `contracts × feeRate × price × (1-price)` calculations. - Added runtime marker `engineVersion=2.5.0` / `contract-semantics-v5`. - Added regression tests for team-query precision, macro search contamination, extra-innings props, and sports fee-curve economics. ### 2.4.0 - Added runtime engine/semantics version markers to every row and run summary. - Added redundant raw-title semantic gates so spreads, totals, inning/period props can never match moneyline contracts even if canonical fields are stale. - Added annual Fed cut-count and hike-count horizons distinct from single-meeting Fed decisions. - Added league inference from two explicit team aliases when Polymarket omits a league token. - Added runtime post-canonicalization semantic guard in the main pipeline. ### 2.3.0 - Added contract-type semantics for sports: MONEYLINE vs SPREAD vs TOTAL. Spread/total markets can no longer pass as winner contracts. - Added annual Fed cut-count canonicalization so cumulative yearly cut-count markets cannot match single-meeting 25bps decisions. - Removed the exact-day mismatch penalty for monthly macro release/meeting identities when the canonical period and contract semantics match. - Added support for Kalshi current `orderbook_fp.yes_dollars/no_dollars` fixed-point payloads, preserving legacy orderbook support. - Targeted query runs now keep a 3x base discovery pool so exact game markets are not crowded out by props/futures returned from native search. - Treat Polymarket HTTP 422 at the public pagination ceiling as end-of-list rather than a source outage. ### 2.2.0 — targeted discovery and cross-platform recall hardening - Added Polymarket `public-search` discovery for user queries instead of relying only on arbitrary market-list pages. - Added Kalshi series-targeted discovery for NFL, MLB, NBA, NHL, Fed, CPI, unemployment, GDP, payrolls, BTC and ETH families. - Broad scans now fetch a larger discovery pool and reserve capacity for overlap-heavy Kalshi series before matching. - Polymarket list discovery is ordered by volume before local canonical matching. - Bare `Fed` query intent now maps to `FED_DECISION`; short-form Fed rate markets canonicalize as macro decisions. - Added canonical coverage, targeted-discovery counts, series hints and shared-event-key counts to run diagnostics. - Preserved exact sports event identity and the semantic/resolution gate before order-book economics. - Added live-source regression fixtures for Polymarket nested search results and Kalshi `series_ticker` discovery. ### 2.1.0 — 2026-09-23 - Fixed Polymarket Gamma pagination so `maxMarketsPerPlatform` can actually load beyond the first 100 markets. - Added canonical league matching to focused queries such as `NFL`. - Prefer explicit event dates from titles/slugs/tickers over exchange close timestamps. - Replaced the O(N×M) GENERAL/POLITICS fuzzy fallback with date/period-blocked comparisons. - Prevents broad 1000×1000 scans from timing out solely because of fallback matching. - Added regression coverage for pagination, league query matching, slug event dates, and blocked fallback behavior. ### Product V2 / source 2.0.0 — 2026-09-23 Major architecture rebuild of the existing `nexascout/prediction-market-opportunity-risk-radar` Actor. #### Matching - Replaced title-first cross-platform matching with vertical-specific canonical event fingerprints. - Added sports league/team/date normalization with common MLB, NFL, NBA and NHL aliases. - Added semantic outcome-side mapping for equivalent vs inverted contracts. - Added macro metric/period/threshold normalization and separated CPI MoM from CPI YoY. - Added crypto asset/date/threshold and weather metric/location/date canonicalization. - Added conservative GENERAL/POLITICS fallback instead of allowing weak generic text matches into economics. - Added alias-aware search filtering (`Yankees` ↔ `New York Y` / `NYY`). #### Resolution gate - Contract-threshold mismatches are rejected before order-book requests. - Event-date and period mismatches are hard risk signals. - Resolution-source mismatches are surfaced and scored. - Cancellation/postponement policy differences are compared where detectable. - Rejected semantic pairs no longer receive fake arbitrage-profit calculations. #### Market data - Reduced Kalshi pagination pressure by requesting larger pages. - Added HTTP 429/backoff retry behavior and public-host fallback. - Added explicit `PARTIAL_SOURCE_FAILURE` diagnostics when a platform source fails rather than implying no market opportunity exists. - Preserved a 2,000-market Polymarket safety cap from the latest pre-V2 source line. #### Execution economics - Added multi-level depth-aware fills on both venues. - Correctly derives Kalshi executable asks from opposite-side bids. - Added conservative Polymarket fee handling. Fee-enabled markets with unknown fee math cannot become ACTIONABLE by default. - Added manual fee override mode for controlled research. #### Monitoring and schemas - Replaced run-scoped history with a named persistent KVS. - Added complete dataset/output schemas for human and agent consumers. - Preserved existing Store slug and actor-level monetization. #### Tests - Added regression tests for the previously observed false sports match (Rays/Yankees vs Baltimore/Yankees). - Added tests for shortened team aliases, macro threshold matching, CPI MoM vs YoY separation, fee safety, order-book depth, and schemas.