Quant.
Mid-price is not execution price.
Microstructure features for event contracts: top-5 depth, spread, depth ratio, flow toxicity proxy, regime label. Executable edge after slippage. Depth-adjusted Kelly. Same shape across Kalshi and Polymarket, streaming over WebSocket, persisted as time series, drop-in for your signal pipeline. JSON everywhere.
本間宗久 · 堂島米会所 1750s — Munehisa Homma, the rice trader who invented candlestick charts, the world's first quant signal.
Live edges with orderbook depth
Updated every 15 minutes · data from Kalshi and Polymarket
Microstructure features per contract
Seven fields. On every tick. On both venues.
Event-contract microstructure is not equity microstructure. Lot sizes, tick sizes, settlement, payout asymmetry — all of it shows up in the features. The runtime ships them all in one shape.
bidAskTopNTop-5 bid + ask levels with size, refreshed on every venue tick.
executableEdgethesisPrice − askPrice, walked through your size, debited for worst-case slippage.
depthRatioBid-side depth ÷ ask-side depth. Imbalance ≠ noise — it predicts mid revisions.
regimeLabelTrending / ranging / dispersing — derived from spread persistence + cross-venue divergence.
flowToxicityVPIN-style proxy on recent prints; modulates adverse-selection threshold.
kellySizeHalf-Kelly on post-slippage edge, capped by 25% of cumulative depth at fill price.
whatIfDeltaOverride any tree leaf, cascade through, get edge-shift vector — ~10ms, no LLM hop.
Why a normalized quant surface
Venue SDKs (raw)
Tightest possible quotes per venue
No cross-venue normalization, no thesis loop, no edge math, no regime label
Generic algo platform
Familiar tooling (FIX, drop-copy)
Has zero understanding of event-contract structure or settlement
Notebook + cron
Total flexibility for ad-hoc research
No real-time depth, no kill chain, no audit trail, lifetime maintenance burden
SimpleFunctions Quant
Cross-venue depth + edge + regime + intent
Surfaces only when post-slippage edge clears floor and depth covers size
Who runs the quant loop
Six recurring quant shapes — same microstructure surface, different cadence and risk envelope.
Systematic shops
Microstructure features fed straight into your signal pipeline; depth + spread + regime + flow.
Stat-arb desks
Cross-venue divergence + matched-pair execution; same shape across Kalshi and Polymarket.
Discretionary quants
What-if engine for IC packs; override a leaf, see edge cascade across the tree.
Microstructure researchers
Per-tick orderbook history + flow toxicity proxy; backtest on parquet dumps.
Algo agents
MCP-driven thesis tree, intent declarations, kill conditions — runtime executes on heartbeat.
Risk teams
Audit trail per intent: edge-at-entry, kill chain, slippage realized, regime label at fill.
Quant endpoints
JSON everywhere. Pipe to your model in shell, Python, R, or Julia. Same shape across CLI, REST/API, and MCP adapter.
Full orderbook ladder
GET /v1/orderbook/{ticker}openEdges with depth + regime
GET /api/public/edges?minEdge=8&depth=high®ime=trendingopenTick-level OHLC
GET /v1/candles/{ticker}?tf=1mopenWhat-if cascade
POST /api/whatif { thesisId, overrides }openMCP — sf.book (agent)
mcp call simplefunctions.book { ticker: "KXFEDDEC..." }openA quant tick, end-to-end
From orderbook pull to slippage-debited fill — five steps, JSON at every boundary.
$ sf book KXFEDDEC-25DEC31 --json | jq .
{
"ticker": "KXFEDDEC-25DEC31",
"mid": 38.5, "spread": 1,
"bids": [[38, 2400], [37, 1800], [36, 800]],
"asks": [[39, 1500], [40, 1100], [41, 700]],
"depthRatio": 1.78, "regime": "trending",
"flowToxicity": 0.21, "venue": "kalshi"
}$ sf edges --json --depth high \
| jq '.[] | select(.execEdge > 8)'
{
"ticker": "KXFEDDEC-25DEC31",
"thesisPrice": 52, "ask": 39,
"execEdge": 13, "slippage": 0.4,
"kelly": 350, "regime": "trending",
"depthAtFill": 24500
}$ sf whatif --thesis fed-cuts \
--override 'recession_prob:0.45'
CASCADE 11ms · 6 markets touched
KXFEDDEC-25DEC31 +13¢ → +18¢
KXFED-26-T2 -3¢ → 0¢
KXRECESSION-26 +20¢ → +27¢
KXGDP-26 +5¢ → +12¢
KXCPI-MAY-T4 -1¢ → -2¢
KXTREASURY-10Y +4¢ → +7¢$ sf intent buy KXFEDDEC-25DEC31 350 \
--price 39 --kelly 0.5 \
--regime-gate trending \
--kill above:48 --kill flow:>0.4
✓ armed · trail intents/i_8a2f4c
fired 11:14 UTC, fill 350 @ 38¢
exec edge realized +14¢, slippage 0.2¢
regime at fill: trendingClient + pipelines
Python client
pip install simplefunctions-ai
import simplefunctions as sf
book = sf.book("KXFEDDEC-25DEC31", depth=5)
edges = sf.edges(min_edge=8, depth="high")
delta = sf.whatif("fed-cuts",
override={"recession_prob": 0.45})
sf.intent.buy("KXFEDDEC-25DEC31", 350,
price=39, kelly=0.5)Shell pipeline
# regime-gated edges, sized by depth
sf edges --json --depth high \
| jq '.[] | select(.regime=="trending")' \
| jq '.[] | select(.execEdge > 10)' \
| sf size --kelly 0.5 \
| sf intent.bulk
# stream tick-level depth
sf ws subscribe orderbook KXFEDDEC-25DEC31FAQ
What microstructure data does SimpleFunctions expose?
Top-5 bid/ask levels, mid, spread, total depth (both sides), depth ratio (imbalance), trade-flow toxicity proxy, recent print VWAP, time-since-last-trade, and per-tick spread + depth time series. Persisted on every heartbeat snapshot for both venues; queryable via /v1/orderbook, /v1/candles, /api/public/markets, or sf book --json.
How is executable edge computed?
YES buyer: thesisPrice − askPrice; NO buyer: bidPrice − thesisPrice. Then walked through the orderbook for your size to get post-slippage realized edge. Then debited for the worst-case price-impact scenario. The model edge minus half-spread shortcut is also exposed for sanity-checking, but intent triggers gate on the post-slippage number.
What about regime detection?
Three regimes are tracked per topic group: trending (depth ratio persistently > 1.5x one side), ranging (mid stable within 2¢ over 24h), and dispersing (cross-venue divergence > 5¢ for 1h+). Regime labels feed into per-topic Kelly multipliers and adverse-selection thresholds. Exposed as a derived field on every edge.
Depth-adjusted Kelly sizing — what does that mean?
Half-Kelly on the post-slippage edge, capped by a fraction of cumulative orderbook depth at your fill price (default 25%, configurable). The runtime returns a recommended size with each surfaced edge; per-thesis caps and per-venue caps stack on top.
What-if engine, in concrete terms?
Override any leaf in your causal tree (e.g., "Hormuz reopens with prob 30% instead of 12%"); the runtime cascades the change through the tree, recomputes thesis-implied prices for every matched contract, and emits a delta vector showing edge shifts. Pure computation — no LLM hop, ~10ms per scenario.
Python client / R / Julia support?
pip install simplefunctions-ai for the preview Python client over the HTTP/Data API. R can use httr2 around /api/public/*; Julia can use HTTP.jl. The CLI supports --json for shell pipelines. WebSocket streaming for sub-second tick data is exposed through /v1/ws.
How is this different from a Kalshi or Polymarket SDK?
A venue SDK ships endpoints; SimpleFunctions ships cross-venue normalized data, thesis loops, and an intent layer. Same shape across both venues, with depth/spread/regime/edge/kill computed on top. The CLI and HTTP/Data API are the source of truth; SDKs are wrappers around those contracts.
Latency profile?
Heartbeat 15-min default for the curated context; orderbook snapshots refresh on every venue tick (sub-second on Kalshi, ~1s Polymarket via Gamma). WebSocket /v1/ws streams tick-level updates. Out-of-band ticks fire on FOMC / CPI / NFP releases and major exchange events.
Backtest pipeline?
Daily Hugging Face dump under SimpleFunctions/markets and SimpleFunctions/edges (CC-BY-4.0). Per-tick orderbook history via /v1/candles?tf=1m. Both expose ticker, side, depth, spread, mid, edge, regime label — drop straight into pandas / vectorbt / Polars.
Cost?
Public endpoints free, no auth, CC-BY-4.0. Authenticated execution endpoints (intents, orders, BYOK) are metered by call. WebSocket /v1/ws is free for read; authenticated writes are metered. Heartbeat ticks are bundled into the metered quota.
Related surfaces
Edge discovery
How thesis-vs-market edges are surfaced and ranked.
For funds
Cross-venue tools, audit trail, BYOK execution for fund desks.
Maker strategy
Quote inside the spread, manage adverse selection.
Taker strategy
Cross the spread on news-driven directional moves.
Portfolio Autopilot
LLM agent that runs the loop on a schedule.
Crypto traders
BTC, ETH, SOL ladders + ETF flow + halving contracts.
Prediction market API
CLI, REST/Data API, real-time WebSocket streams, MCP adapter.