Oracles and Pricing
Understanding how pricing will work and how to prevent manipulation
Goals
track a fair “spot” for each skin (per market-relevant variant)
make it hard to spoof (wash trades, thin-book spikes, stale quotes)
keep the feed cheap + fast enough for funding, margin, and liquidations
1) Data sources
Aggregators & APIs
Pricempire: multi-market aggregator with a real API (current + historical, volume stats, wear/sticker meta). Use as your primary consolidation layer and to quickly fan out to per-market quotes.
Skinport: stable, well-documented REST API; good for executable spot quotes.
Steam Community Market: official pricehistory endpoint; useful for baseline + long tails, but beware fees and different liquidity regime.
BUFF163: no official API but will fetch these prices from Pricempire when available considering it has the deepest liquidity of any market available at the moment
2) What’s a contract and what’s an item?
Skins come in many variants (wear/float, stickers, pattern index). We will list:
Specific-item perps (e.g., “AK-47 | Redline (Field-Tested)”), or
Index perps (a “collection floor” for X class of items).
3) Oracle architecture
A. Fast path (real-time mark/reference)
Poll Pricempire (and direct markets) every few seconds for best bid/ask + last trade per venue.
Compute a venue-weighted median of midprices, with liquidity weights (recent volume × active listings) and latency decay (quotes older than N seconds down-weighted).
Publish this “Oracle Spot” every ~3s (same cadence as hl). Use it in the mark price and funding basis.
B. Robust path (safety backstop)
Maintain a rolling VWAP/TWAP over 15–60 min per venue; compute a trimmed-median index across venues (drop top/bottom outliers by %).
If the fast path trips guardrails, fall back to the robust index until conditions normalize.
4) Normalization
Before aggregation, normalize quotes:
Fees & currency: convert all venues to net proceeds (buyer/seller fees, FX).
Liquidity screen: require minimum active listings and recent trades per venue to be eligible for weighting.
5) Manipulation resistance
Borrowing from Hyperliquid “robust price indices” plus NFT floor pricing practice: Outliers & stale data
Trimmed set: per tick, drop quotes beyond k·MAD (median absolute deviation) from the venue-weighted median.
Stale cutoff: any quote older than, say, 15–30s excluded or heavily down-weighted.
Depth check: ignore prints whose size would not clear at least dollars X or Y items at the venue’s current book protects against micro-size wash spikes.
Venue trust & dilution
Dynamic venue weights: function of 24h volume, current depth, uptime, and historical deviation from cross-venue medians.
Steam vs. OTC markets: account for Steam fee regime; treat Steam as signal but not sole anchor.
Wash-trade filters
Per venue, require distinct counterparties, minimum notional, and time-separation; de-prioritize self-matched or ping-pong patterns
Circuit breakers & fallbacks
If fast-path deviates > X% from robust TWAP (or if eligible venues < M), freeze to robust index and widen funding caps until recovery
6) What the DEX actually uses
Mark price (for PnL, margin, liquidations):
Median of: {fast-path Oracle Spot, robust TWAP index, last on-chain mid (if orderbook), and a safety “reference” (e.g., Steam 24h VWAP net of fees)} with caps on pairwise divergence.
Funding rate (keep perp in line with spot):
Hourly:
funding = clamp( k * (PerpIndex - OracleSpot) / OracleSpot , ±cap ). Use the Oracle Spot as the external anchor like Hyperliquid; cap funding during oracle fallbacks.
Index governance
List only items whose venues pass liquidity/uptime SLOs; add/remove venues by governance and auto-scoring (volume, quote quality, downtime).
7) On-chain availability
Publish the oracle on-chain every few seconds via a validator set or keepers.
Expose a precompile or lightweight view (if on your own L2/appchain) or a simple Oracle contract with:
getPrice(id),getTWAP(id, window),isFallback(id)and round IDs for auditability
TL;DR blueprint
Aggregate many venues (via Pricempire + direct APIs). Normalize for fees/variants.
Compute a liquidity-weighted median fast price, updated every ~3s. Backstop with a trimmed, multi-venue TWAP.
Use a median of robust components as your mark; anchor funding to the oracle spot; enforce outlier, staleness, and depth guards; and fall back cleanly under stress.
Last updated