The pnl whitepaper
Systematic trading strategies on pnl are authored as programs, admitted by a statistical gate, and published as immutable listings. Users take signed positions in them from their own perpetual-DEX accounts; the platform never custodies funds.
The design rests on one property: a strategy emits a target position series, and target position series superpose linearly. This supports portfolio netting and a linear decomposition of target exposure. Realized accounting follows reconciled execution records.
The paper uses the vocabulary of a layer-one blockchain throughout — an algo VM with declared gas and storage, a fixed block clock, immutable deployment, and an admissibility rule we call Proof-of-Backtest. The paper states where that correspondence fails, because a metaphor that is not bounded is a claim that cannot be checked.
What is in it
Section titled “What is in it”The canonical PDF covers the strategy VM, deterministic resource bounds, position netting and attribution, the tradeability gate (drift OR fee-aware persistence, 3 of 4 quarters, 420-day minimum) and the team-operated services: router-service, data-node, signal-node, execution-node, agent-service, backtest-node workers and admin, built on @pnl/core and @pnl/engine. Passing full evaluation produces an owner-bound verdict; listing metadata is supplied afterward. It also describes the platform fee: 5 bps per fill up to one turn of capital a day, then 5/T bps down to a 0.1 bps minimum, with the referral discount applied separately.
Language and surface claims refer to the pinned release. Private policy thresholds are omitted; deployment status and Trial/Proven qualification criteria remain separate from this architecture description.