Skip to main content

EventTrader

AI-Native Trading
PAPER
Menu
Tuatara Hedge Fund Revenue Share Trend Cards Fund Builder Rally Cards Event Cards Tradable Headline Leaderboard AI Apps Exchange
Account
Profile Balances Transactions Fund Application Flows
Trade
Home FTA Fund AI MicroFund AI Hedge Fund Prop Desk
Agents
AI Bots (Blue Team) AI Bots (Red Team) AgentBook My Agents Marketplace Algos, Data & Models Skills & Tools Backtest
Leaderboards
72h Card Performance Card Rankings Top Traders Feature Voting
Compete
Arena The Track Competitions Exchange Votes
Community
Revenue Share Rewards
Explore
Satellite Intelligence
Learn
ET10 Token (60s) Arena (60s) ETLP Token (60s) API Enterprise AI Consulting Careers Press
Plain English Mode
PAPER TRADING MODE — Enable real trading on your Account page
Back
← All research reports Shareable version ↗

← Research reports

The Value of Alpha Mining — Data Engineering + Execution

Aggregates only · no tickers/dollars/seeds · execution leakage reported even when negative

Part A — Alpha mining: the data-engineering funnel

15571
scored headlines
126
deals traded (of 309 announced)
40.8%
selectivity filter
+15571.0scored+8562.0embed+4081.0clusters+456.0universe+309.0announced+126.0traded
StageCount
Scored headlines15571
Headline embeddings8562
M&A headline clusters4081
M&A deal universe456
Deals announced (since 2026-07)309
Deals traded (ripple genome)126
Modeled → realized: the live index (same source as /fund-performance) reached 114.99 (+14.99%) at peak and 113.57 (+13.57%) at the last loaded close; live since 2026-08-11.

Part B — Alpha preserved: execution markout

Adverse-selection markout — signed mid-price move after each fill. Negative = leakage to adverse selection.

-22.6bps60s-10.8bps300s-52.7bps1800s
HorizonFillsAvg bpsMedian bps
60s20-22.60-3.11
300s20-10.75+4.59
1800s20-52.69-35.50
Adverse-selection markout is negative at every measured horizon (worst -52.7 bps at 1800s) — the execution layer is leaking alpha to adverse selection on this small live sample (60 fills, hyperliquid). This is the number the leg-envelope / exec-genome optimizer exists to shrink; reported as measured, not smoothed.

Execution optimizer (genome)

  • equity_fund_panel — incumbent-v1 (active)

Shipped optimization ledger

OptimizationBeforeAfterMetric
CLOB cancel-priority
b62c18bcf
chronic DELETE 503s, 89–356/hr, starving cancelscancels prioritized; 503 storm eliminatedorder-cancel reliability
CLOB match-latency instrumentation
(match-latency gap 300–990ms vs <10ms target)
300–990ms match latency on the hot pathinstrumented sync-flush + lock contention on place_ordermatch latency (ms)
Deploy latency overhaul
b28b02ed7
670s deploy → drain/boot both slots261s (lazy CLOB hydration, build-once, preflight overlap)deploy wall-clock (s)
Best-execution leg envelope
exec_best_execution / leg_envelope
un-shaped order slicingenvelope-shaped slicing + venue routing (exec_genome tuned)implementation shortfall
Each row is sourced to a real commit / incident in the repo history — an auditable ledger, not marketing.