Predge Whale Data API
unreachableDispute and refund verdicts for AI-agent payments: POST the job terms, the delivery evidence and the payment reference, get a signed delivered / notdelivered / inconclusive verdict with a refund recommendation ($0.05 when the paid call is under $1, $0.50 at $1 or more; free with a 30-day pilot key from https://predge.io/start). Also: free signed Polymarket settlement-risk checks, the paid…
Settled via PayAI.
- Transactions · 30d
- 0
- Volume · 30d
- $0.00
- Unique buyers · 30d
- 0
- Uptime · 30d
- 56.4%
- Latency p50
- 82ms
Endpoints (28 live)
GET/v1/onchain/pm-flows— 24h aggregate ON-CHAIN Polymarket flow by market, as ONE ed25519-SIGNED attestation (kind "onchain-flows"): per-market YES/NO volume split, net flow, trade counts and unique wallets over tx_hash-anchored Polygon trades only (platform='polymarket' AND tx_hash present — Kalshi's anonymous prints are excluded by construction). Honest scope: Polymarket-whale on-chain activity, NOT general crypto flows. The signed window_start/window_end are the exact bounds the aggregation ran over (window_end honours the 15-min delay); USD volumes are MODELED notionals, labeled in data_basis. Top-20 markets by volume; market titles ride outside the envelope (mutable metadata, never signed). Verify FREE offline at GET /verify. (0.02 USDC on Base)GET/v1/signals/consensus— PREMIUM smart-money consensus. "Smart money" is EDGE-based, not the opaque scorer: a wallet qualifies when its luck-damped realized edge over the last 30d is positive — edge_lower_bound = Wilson-95%-lower-bound(live win rate) − avg entry price > 0, from RESOLVED outcomes (same machinery as the wallets leaderboard), with skin-in-the-game floors (>= 10 trades, >= $1,000 volume). For every active Polymarket market with whale trades (>= $10,000) in the last 24h from those wallets: YES/NO volume split, net flow (USD), distinct smart-wallet count, and a direction verdict — "yes"/"no" when one side holds >= 55% of smart volume (>=10% imbalance), else "split". Sorted by smart volume, top-50. Honest degradation via {"insufficient_data": true, "reason": ...} while no wallet has a positive edge yet. 15 min delay. (0.03 USDC on Base)GET/v1/signals/daily— Aggregated 24h whale digest for Polymarket (15 min delay): top-10 markets by whale volume with YES/NO net flow, top-10 largest single bets, and totals (whale trade count, volume, unique wallets). One call = full daily picture. (0.02 USDC on Base)GET/v1/wallets/leaderboard— Top Polymarket wallets over a 7d or 30d window, ranked by EDGE, not by raw win rate. Win rate is computed from RESOLVED markets only (a trade wins when its side matches the final resolution), but ranking a wallet by win rate alone rewards backing favourites — winning a bet the market prices at 0.99 shows no skill. So wallets are ordered by edge_lower_bound = (Wilson-95%-lower-bound of win_rate) − avg_entry_price: how far the wallet beat the price it paid, damped so a tiny perfect record can't top a large one. Final ranking is by conviction_score = edge_lower_bound × ln(1 + total_volume_usd), scaling proven edge by the capital behind it (a real edge on $200k outranks the same edge on $6k), with log diminishing returns so size alone can't buy the top. Rankable wallets need >= 10 trades AND >= $1000 total volume in the window (a skin-in-the-game floor — cent-sized longshot bets otherwise manufacture a real-looking edge) plus at least one decided trade. Each entry: address, Predge score (0-100 or null), win_rate, avg_entry_price, edge, win_rate_lower_bound, edge_lower_bound, conviction_score, decided_trades, trades_count, total_volume_usd, avg_position_usd. Honest degradation: {"insufficient_data": true, "reason": ...} when the dataset is too young. Query params: window (7d|30d, default 30d), limit (1-100, default 50). (0.01 USDC on Base)GET/v1/markets/movers— Largest YES-price moves across active Polymarket markets over a 1h/6h/24h window (15 min delay). Prices are derived from stored trade prints normalized to the YES side (NO prints count as 1 - price) — trade-print series, not orderbook quotes; markets need >= 2 prints in the window. Each entry: condition_id, title, price_before, price_now, delta, prints_in_window, volume_in_window_usd, whale_volume_in_window_usd (trades >= $10,000). Top-20 by |delta|. Query param: window (1h|6h|24h, default 6h). (5000 on Solana)GET/v1/whales/latest— Latest Polymarket whale trades (15 min delay). Trades >= $10,000 notional with market, side, size, price, wallet address and Predge wallet score (0-100, 30d win rate). Source: Predge whale-tracking pipeline (@predge_alerts_bot). Query param: limit (1-100, default 50). (5000 on Solana)POST/v1/verdicts— Delivery verdict, standard tier ($0.05, for a paid call under $1): did the service an AI agent paid for deliver as promised? POST the job spec (endpoint, expected content type, optional JSON Schema, required fields, max latency, expected sha256), the delivery evidence (HTTP status, headers, body or its sha256, timing) and the payment reference (amount_usd, tx, payer, payee). Deterministic checks, no LLM. Returns a verdict (delivered | not_delivered | inconclusive), per-check results, reason codes and a refund RECOMMENDATION (the merchant honours it; no automatic or on-chain refund yet), ed25519-signed (predge-attest-v1 + x-predge-body-signature) and stored: free re-read at GET /v1/verdicts/:id. A payment.amount_usd of $1 or more gets 400 wrong_tier (use /v1/verdicts/large), not charged. A pilot key (x-predge-key) is free in its 30-day pilot window. (0.05 USDC on Base)GET/v1/wallets/compare— Side-by-side comparison of 2-10 wallets. Query param: addresses (comma-separated 0x… addresses, 2-10, deduplicated). Each wallet: Predge score, live win_rate_7d/win_rate_30d (computed from RESOLVED markets only — outcome_verified: {win_rate: true, pnl: false}), modeled_pnl_30d_usd/modeled_pnl_all_time_usd (a MODELED binary-resolution estimate over RESOLVED markets — $1 per winning share, shares = size_usd ÷ entry price; the same model every wallet route uses, not realized/settled P&L), and top-5 favorite categories; plus market_overlap: markets traded by >= 2 of the compared wallets. Wallets unknown to the dataset are returned with found:false rather than dropped. (20000 on Solana)GET/v1/bets/biggest— Largest single trades by notional across ALL Polymarket markets in the window (15 min delay). Same trade shape as /v1/whales/latest, sorted by size_usd descending, top-20; live (unresolved) flow — outcome_verified: false. Query param: window (24h|7d, default 24h). (5000 on Solana)GET/v1/sports/attest/nba/1— Cryptographically SIGNED settled-SPORTS-outcome ATTESTATION for one game (path params: league — one of nba|wnba|nfl|college-football|mlb|nhl|mls|epl|champions-league|laliga — and eventId, an ESPN event id). Returns the SETTLED result (winner, final score, completed) as an ed25519-signed envelope verifiable OFFLINE against the SAME published key as /v1/attest. Optional query param: winner (team abbreviation) — adds correct (true iff it matches the settled winner, null while unsettled). Data source: ESPN's public scoreboard (no key, no account). Outcome-verified truth, not a prediction — the tamper-evident settled outcome a betting/trading agent needs for settlement, dispute, or a verifiable track record. 404 (not charged) for an unknown league or event id. (0.02 USDC on Solana)GET/v1/signals/consensus/1— Smart-money consensus for ONE market category (path param, e.g. economics, politics, sports — case-insensitive). Same EDGE-based smart-money definition as /v1/signals/consensus (luck-damped realized edge over the last 30d > 0, NOT the scorer score), scoped: active Polymarket markets of that category with whale trades (>= $10,000) in the last 24h from edge-positive wallets; YES/NO smart volume, net flow, direction verdict. Category coverage is partial — markets without a category are excluded (see coverage_note in the response). Unknown category => 404 (not charged) listing known_categories. Honest degradation via {"insufficient_data": true, "reason": ...}. 15 min delay. (0.02 USDC on Base)GET/v1/signals/daily/1— 24h whale digest for ONE market category (path param, e.g. economics, politics, sports — case-insensitive): top-10 category markets by whale volume with YES/NO net flow, top-10 largest single bets, and totals — same shape as /v1/signals/daily, category-scoped. Category coverage is partial — markets without a category are excluded (see coverage_note in the response). Unknown category => 404 (not charged) listing known_categories. 15 min delay. (0.01 USDC on Solana)GET/v1/settlement-risk/1/estimate— SIGNED settlement-risk record WITH settlement_estimate for one Polymarket market (path param: condition id, numeric market id or slug). Same record as the free GET /v1/settlement-risk/:market (UMA state, current proposal, every dispute, UMA-vote status, bulletin-board updates, on-chain resolution, risk_level) plus settlement_estimate: for a market disputed once and not yet resolved, the calibrated probability that it settles differently from the disputed proposal (logistic model, settlement-estimate-v1.1, trained on disputes raised January to June 2026, tested on 754 disputes raised August to September 2026: AUC 0.674, precision in the 100 highest estimates 0.65 averaged over tied scores); for a twice-disputed market, the 2026 reference rate for markets sent to a UMA vote; null when there is nothing to estimate or on Protocol V2 markets. Also segment_dispute_rate. ed25519-signed (predge-attest-v1), the payment that bought it named in the signed payload, verifiable at /verify. A statistic, not advice. 400/404/422/502 are not charged. (0.02 USDC on Solana)GET/v1/attest/1— Cryptographically SIGNED resolved-outcome ATTESTATION for one Polymarket market (path param: 0x… condition id). Returns the SETTLED truth straight off the market row: resolved (bool), resolution (yes|no|null), resolved_at, and outcome_verified: {resolution: true}. Optional query param: side (yes|no) — when given, adds queried_side and correct (true iff side matches the resolution, null while unresolved). Every response carries an "attestation" envelope: an ed25519 signature over a canonical JSON payload plus the issuer's public_key, so any consumer can verify the outcome offline without trusting this endpoint. This is on-chain resolved outcome truth, NOT a prediction or model estimate — the "was this signal actually right", independently-verifiable primitive no other x402 source sells. 404 (not charged) for a condition id unknown to the Predge dataset. (0.02 USDC on Base)GET/v1/track-record/1— SIGNED portable track record of a predictor's live calls: resolved-only decided/correct counts, Wilson 95% lower bound, and a sha256 merkle root content-addressing the exact corpus of signed calls it summarizes (recomputable offline). Records under the min-decided floor carry clears_bar=false. Param: pubkey (64-hex ed25519), query window (30d|90d|all). (0.05 USDC on Base)GET/v1/wallets/1/history— Historical trades for one wallet (path param: 0x… proxy-wallet address) plus a running win-rate & MODELED-PnL curve, computed ONLY from RESOLVED markets (modeled_pnl_usd is a MODELED estimate, NOT realized/settled P&L — it assumes a $1 binary payout per winning share, shares = size_usd ÷ entry price; not a substitute for on-chain settled P&L). Unresolved trades still appear (won: null, modeled_pnl_usd: null) but do not move the curve. Query param: window (30d|90d|all, default 30d — "all" is UNBOUNDED, window_days: null). The trades array is capped at 500 most-recent-first, but summary is computed over EVERY trade in the window (see summary.trades_returned / trades_capped), so the total matches this wallet's profile. 15 min delay. 404 (not charged) for wallets unknown to the Predge dataset. outcome_verified: {win_rate: true, pnl: false}. (0.02 USDC on Base)GET/v1/signals/market/1— Single-market smart-money VERDICT for one Polymarket market (path param: conditionId, 0x… Polymarket condition id) — the "should you trust this move?" decision, built to be looped once per market per cycle. This is /v1/signals/consensus scoped to one market: SAME edge-based smart-money population (a wallet qualifies when its luck-damped realized edge over the last 30d — Wilson-95%-lower-bound(live win rate) − avg entry price — exceeds 0, from RESOLVED outcomes, with the same >= 10 trades / >= $1,000 floors), and the SAME direction rule (>= 55% of smart volume / >= 10% imbalance => "yes"/"no", else "split"). Returns market metadata (title, category, yes_price, status, and resolution/resolved_at when settled), the edge-proven smart-money read for whale trades (>= $10,000) in the last 24h — distinct smart wallets, YES/NO split, net flow (USD), direction verdict — plus total-vs-smart whale volume and a plain confidence (none|low|medium|high) + basis an agent can act on. Honest degradation via {"insufficient_data": true, "reason": ...} when too few smart wallets have touched the market. 404 (not charged) for a condition id unknown to the Predge dataset. 15 min delay. (0.01 USDC on Base)GET/v1/markets/1/edge— Sub-cent per-market smart-money EDGE PULSE for one Polymarket market (path param: conditionId, 0x… condition id), built for CONTINUOUS POLLING by trading bots — a lean projection of /v1/signals/market meant to be hit every loop across many markets, not looked up once. Returns just the decision numbers: direction (yes|no|split) and net smart-money flow (USD) over edge-proven wallets' whale trades (>= $10,000) in the last 24h, distinct smart_wallets, smart_volume_usd, a confidence (none|low|medium|high), the current yes_price for context, and a "stale" flag (true once the market resolves or too few smart wallets have touched it — the bot's signal to stop acting). SAME edge-based smart-money definition as /signals/consensus. 30s server cache; 15 min delay. 404 (not charged) for a condition id unknown to the Predge dataset. (0.002 USDC on Base)GET/v1/wallets/1/attest— Cryptographically SIGNED WALLET TRACK-RECORD attestation (path param: 0x… wallet address; optional query param: window = 7d|30d, default 30d). Returns the wallet's RESOLVED-market-only record — win_rate, Wilson 95% lower bound, edge over entry price, conviction_score, volume, and a verdict (verified_smart_money | positive_edge | no_proven_edge) — as an ed25519-signed envelope verifiable OFFLINE against the SAME published key as /v1/attest. Every other tool shows a self-reported or scraped win rate you must trust; this SIGNS it, so a copy-trading platform, a fund proving a track record to an allocator, or an agent sizing a copy can verify the record without trusting this endpoint. Win rate is computed from resolved markets only (a trade wins when its side equals the market's eventual resolution). 404 (not charged) when the wallet has no resolved trades in the window. (0.02 USDC on Solana)GET/v1/onchain/pm-activity/1— Recent ON-CHAIN Polymarket trades of one wallet (path param: 0x… proxy-wallet address), EACH wrapped in its own ed25519-SIGNED attestation (kind "onchain-activity"). Honest scope: Polymarket-whale on-chain activity — real Polygon transactions of prediction-market wallets — NOT general ERC-20/DEX crypto intel. Every attested row has a real tx_hash (Kalshi's anonymous prints have no on-chain identity and are excluded by construction). Signed facts per trade: tx_hash + block_time (no block number is stored, none is claimed — derive it from the tx receipt), wallet, condition_id, side, price; size_usd is included but labeled a MODELED USD notional in data_basis. Attestations are minted ONCE per tx_hash and reused (byte-stable signature per transaction), and verify FREE offline at GET /verify against the same published key as every Predge envelope. 15 min delay. Query param: limit (1-50, default 20). 404 (not charged) for a wallet with no on-chain Polymarket trades in the Predge dataset. (0.02 USDC on Base)
+8 more endpoints.
First seen · last seen · last active