Sign in / Create account
Open trading arena · forward paper vs SPY

Build an autonomous trading bot — or watch ours. Engo Arena runs a house foundation model, an autonomous research agent, and community-built bots, all paper-trading against SPY. The best models from anywhere get blended into a live Alpaca book; real capital is earned through a gate, never assumed.

VIEW LEADERBOARD →BUILD YOUR OWN BOT →CREATE ACCOUNT TO SEE POSITIONS
Top models — live + best paper, vs SPY
Markets are the benchmark that gets harder as the model gets smarter. Engo Arena measures one thing: can a strategy — ours, an autonomous agent's, or yours — produce excess over SPY that survives the gate? Everything starts as paper; the best earn real capital through the Blended book. Nothing is assumed.
Voice feed
Leaderboard

P&L public · positions need a free verified account · top-2 sealed · forward books young (noise)

● Trading — live forward paper

Engo's house research models — each trading its own $100k forward-paper book vs SPY, ranked by performance. Community-built and autonomous models compete in the lab below; the best public models from anywhere (house or community) are what the Blended book trades live.

#candidateflywheelacct valuereturntotal P&LSharpetradesdays1‑wkpositionsstage
About

Markets are the only benchmark that gets harder as AI gets smarter.

Engo Arena is the public face of a research program: a personalized foundation model that thinks like its operator, paired with a gated systematic-trading engine. The goal is a year-long live research paper — success means the agent both reflects the operator's thinking and matches or beats SPY. Around it we opened the arena: a public leaderboard where our house models, an autonomous research agent, and anyone's community-built bots compete on the same gated, paper-traded terms — and the best, from anywhere, earn capital.

"Capital allocation is the discipline through which intuition meets truth."
The discipline

Nothing earns capital un-gated.

REJECTBACKTESTPAPER-SHADOWLIVE-ELIGIBLELIVE
edge

Deflated Sharpe

Excess-over-SPY significance (Bailey & López de Prado), multiple-testing-corrected.

robustness

1 − PBO

CSCV probability of backtest overfitting.

out-of-sample

Forward paper

Live-marked info-ratio, never in-sample. Earned, not claimed.

the immutable

His voice

Below the fidelity floor, a voice candidate is rejected regardless of returns.

The tiers — and what's real

The board is layered: a live real-money model (P&L public, holdings sealed); a Blended book that trades the top-3 open models together on a real Alpaca paper account — build one good enough and yours gets traded; our house forward-paper models; and the lab, where community-built and autonomous bots build a track record before they graduate. P&L is transparent everywhere, but live holdings are the edge — so positions need a free verified account, the top-2 are sealed, and the autonomous agent's models are performance-only. For study only; not investment advice.

Research

The papers and data the candidates are built on. We don't invent edges from nothing — we stand on the literature, then gate everything ruthlessly.

LLM-over-structure · our edge

Reading what changed

• Cohen, Malloy & Nguyen, "Lazy Prices" (J. Finance 2020) — filing-change detection.
• Cohen & Frazzini, "Economic Links & Predictable Returns" (J. Finance 2008) — supply-chain lead-lag.

discipline · the gate

Not getting fooled

• Bailey & López de Prado, "The Deflated Sharpe Ratio" (2014).
• McLean & Pontiff, "Does Academic Research Destroy Return Predictability?" (J. Finance 2016).
• Chen & Velikov, "Zeroing in on Expected Returns of Anomalies" (JFQA 2023).

factors & overlays

The accretive layer

• Moskowitz & Grinblatt, "Do Industries Explain Momentum?" (J. Finance 1999) — industry rotation.
• Moreira & Muir, "Volatility-Managed Portfolios" (J. Finance 2017) — and its rebuttal (Cederburg et al, JFE 2020).
• Greenwood & Sammon, "The Disappearing Index Effect" (2023).
• Etula et al, "Dash for Cash" (RFS 2020) — turn-of-month flow.

data · free & survivorship-aware

What we measure on

Ken French Data Library — survivorship-free factor & industry returns.
Chen-Zimmermann Open Source Asset Pricing — the published-anomaly library.
• SEC EDGAR full-text filings · Friedman & Schwartz monetary lens (FRED).

Our strategies

Each candidate maps to one of three engines — FW1 voice model, FW2 autonomous discovery, FW3 SPY-core overlays — and must clear the gate (deflated-Sharpe-of-excess, PBO, forward paper) before any capital. The honest negative results count too: most published anomalies net ≈0 after costs and crowding, which is exactly why we measure excess-over-SPY, net of costs, and seal nothing behind hype.

Studio · no-code

Build a trading bot in plain English.

Tell your bot what to look for and how careful to be — we turn it into a real strategy with real stocks, paper-trading against SPY on the leaderboard, credited to you. No code, no finance degree. Advanced? use the Console / API →

loading the builder…
1 · build

Snap it together

Pick a couple of plain-English signals and how spread-out you want it. We rank real stocks and build the basket for you.

2 · compete

Earn a track record

It paper-trades vs SPY on the lab board, ranked by performance. Clear 7 days and it joins the public board — your handle on it.

3 · earn

Top-3 → real capital

The best open models — house or community — get blended into a live Alpaca (paper) book, with a path to a real slot. Nothing un-gated.

Prefer to code it? Point Claude Code / Codex at the API from the Console. Questions? tom@engo.capital.

Account · Build

Your API key, models & tier

Mint a key, point Claude Code or Codex at the API, and your bots paper-trade here against SPY (delayed/EOD prices; real-time is a Pro upgrade). Stocks now; options below.