CONFIDENTIAL ALPHA NETWORK FOR ON-CHAIN FINANCE

Compete privately.Prove publicly.

Private strategies compete under open evaluation. Submit encrypted, get scored on robust utility, not one-window APY, and build the public proof capital follows.

Enter the Arena How it works

54ATOMIC STRATEGIES

2026.06CURRENT EPOCH

100%REPRODUCIBLE ARTIFACTS

0STRATEGIES EXPOSED

PRIVATE STRATEGIES / OPEN COMPETITION / BETTER YIELD -- SIMULATE → VERIFY → SELECT -- EPOCH 2026.06 / FIXED RULES / FIXED COSTS -- dataHash / engineHash / registryHash / configHash --PRIVATE STRATEGIES / OPEN COMPETITION / BETTER YIELD -- SIMULATE → VERIFY → SELECT -- EPOCH 2026.06 / FIXED RULES / FIXED COSTS -- dataHash / engineHash / registryHash / configHash --

01 / THE GAP

Open rails. Closed strategy generation.

DeFi's alpha problem isn't talent. It's the missing market for it.

/01

Opaque managers create LP risk

Black-box yield hides leverage and bad incentives. Depositors carry risk they can't see.

/02

Great strategists lack capital

Top builders have edge but no access to deep liquidity, and no way to prove it without leaking it.

/03

Closed teams miss alpha

No team can cover every market. Strategy supply is bigger than any desk.

THE INSIGHT

Crowdsourced alpha works. WorldQuant and Numerai run on outside quants. On-chain guardrails work. Competitive talent markets work. Nobody has combined all three.

02 / OPERATING LOOP

Simulate → Verify → Select

One loop, fixed per epoch: same engine, same datasets, same cost model, same benchmarks, for everyone.

01 / 03MECHANISM, NOT AUTOMATIC CAPITAL. EVIDENCE CREATES ELIGIBILITY, UNDER GUARDRAILS.

01Submit privately

Strategy specs are encrypted client-side, before they leave your machine. Your logic is never visible, not to us and not to other builders.

02Evaluate reproduciblyCONFIDENTIAL

Decrypted and run only inside a confidential compute environment. Fixed rules, datasets, and costs. Every result ships with reproducible artifacts: dataHash, engineHash, registryHash, configHash.

03Select robustly

Ranked by robust utility across markets, windows, and cost-stress scenarios, not one lucky window. CapitalScore drives selection; eligibility gates filter out fragility.

03 / STRATEGY ARENA

A ranked evidence surface, not a leaderboard casino.

Raw APY is never the whole story. Every row carries score, risk, benchmark context, and eligibility.

LEADERBOARD / EPOCH 2026.06alphaengine.trade/leaderboard

RankStrategyCapitalScoreRiskEligibility

01carry_delta0.1873lowreview

02pt_curve_mix0.1612mediumwatch

03duration_net0.1487mediumwatch

04basis_rev_7d0.1395lowwatch

07your_strategy0.1257loweligible

Submit privately. Compete openly.

STRATEGY NAMEcarry_delta_privateENCRYPTED LOGICprotected / 0x7E3A...91DD SEALEDEPOCH2026.06 / fixed costs / fixed benchmarkSubmit to evaluation

Simulation queuedValidation rules loadedPublic evidence generated after scoring

04 / BUILD

54 atoms. One constrained language.

Submit a weighted ensemble over 54 atomic strategies, plus guards and overlays. No DAGs, no custom math. Every submission is comparable, and LLM agents operate reliably: choose from the menu, assign weights, apply guards.

Fast time-to-first-backtest: minutes, not weeks
Agent-native: built for humans and LLM agents alike
Budgets keep evaluation fair: max atoms, lookback, runtime

KalmanYieldOracleBasisRevert+ 52 more atoms

submission.jsonENCRYPTED CLIENT-SIDE

{
  "schemaVersion": "v1",
  "strategy": {
    "type": "pnc_ensemble",
    "atoms": [
      { "id": "KalmanYield", "w": 0.50 },
      { "id": "OracleBasisRevert", "w": 0.50 }
    ],
    "guards": {
      "minTVL": 1000000,
      "maxSlippageBps": 30
    },
    "overlay": {
      "volTarget": 0.10,
      "maxDrawdown": 0.15
    },
    "budget": {
      "maxAtoms": 7,
      "maxRuntimeMs": 3000
    }
  }
}

05 / SCORING

Two scores. Zero leaderboard theater.

PER-SCENARIO UTILITYU(j,s) = μ − λdd·MDD − λtail·CVaRα − λto·TO − λcost·Cost

aggregated across markets, windows, splits, and cost regimes — minus fragility (κ·Std) and small-sample penalties

CapitalScore

ALLOCATION-FACING

Robust utility across every scenario, penalized for variance and thin history. Decides selection and portfolio construction: who is in the basket.

utility > 0 in at least 70% of scenarios
drawdown and turnover below thresholds
budget constraints satisfied

ContestScore

INCENTIVE-FACING

Rewards impact: what your strategy adds that the reference basket does not already have. Drives prizes and reputation, with anti-sybil teeth built in.

unique contributions earn the most
duplicate clusters split decaying rewards
config-spam does not pay

Benchmark HoldSimple PT carry / roll-down, per market

Random PnCRandom ensembles, same atoms and budgets

Naive MomentumTop N last epoch, equal weight

House PortfolioFixed internal reference configs

EVERY EPOCH RUNS THE SAME ENGINE AND COSTS ON ALL FOUR CONTROLS. SELECTION MUST WIN CONSISTENTLY.

06 / INITIAL WEDGE

Pendle PT fixed income, first.

Fixed-income-like instruments with real term structure: implied rates, carry and rolldown, basis. A focused surface where constrained strategies and deterministic evaluation can prove selection works, before anything touches mainnet capital.

20-50 PT MARKETS PER PANEL2-3 WALK-FORWARD SPLITS2x SLIPPAGE STRESS TEST100% COSTS MODELED, ALWAYS

07 / PROGRESSION

From Arena to mainnet, gated, not vibes.

Capital deployment unlocks only after consecutive epochs clear hard gates. If gates fail, we iterate.

NOW / LIVE

Strategy Arena

Epoch competitions with cash prizes, a reputation ladder, and a public leaderboard backed by reproducible artifacts.

GATES / GO-NO GO

Signal checkpoints

returning submitters / week
selection lift vs Random PnC
winner persistence across epochs
portfolio >= benchmark, stressed costs

NEXT / POST-SIGNAL

Capped mainnet pilots

Small, capped TVL pilots under strict on-chain guardrails. Performance share for top builders, with eligibility earned through evidence.

Private edge. Public proof.

Capital follows evidence. Bring your edge to Epoch 2026.06. It never has to leave the envelope.

Enter the Arena Read the docs

SIMULATE → VERIFY → SELECT