VALHALLA / METHODOLOGY

Evidence you can interpret.

How to interpret forecast records, paper simulations, live fills and model updates without confusing direction with profit.

Separate the three records

Prediction results measure whether the recorded direction matched settlement. Paper results measure a simulation under its fill assumptions. Live results measure actual execution. These are different records and must be labelled separately.

Use the correct contract

Interpret each result against its asset, market identifier, target, horizon and official settlement rules. Do not substitute a convenient chart price for the contract’s settlement reference.

Preserve the decision context

The useful context includes when a forecast or trade decision was made, what data was available then, the engine version and its eventual outcome. Historical analysis must not use later observations as though they were known at the decision time.

Treat fills as evidence

Quoted prices and model values are not fill receipts. Assess execution using order acknowledgements, fills, available depth, cancellations and reconciled positions. Paper assumptions should account for delay, liquidity and partial fills.

Measure profit after costs

Account for entry cost, exit proceeds or settlement payout, and applicable fees. A high win rate alone does not establish profitability. Report sample size, losses and drawdown alongside returns.

Keep research distinct

Historical experiments and shadow results are research evidence. Repeatedly selecting a rule on the same history can overstate its reliability. Later evaluation and live execution evidence are needed to assess how an update performs.

Keep decision and execution authority clear

Prediction dashboards provide forecasts and market information. HUSCARL is the optional execution product, enabled through eligible members’ account and asset controls. A dashboard forecast does not itself authorize an order.

No universal performance promise

Markets, liquidity and behavior change. Neither a historical result nor a self-learning process guarantees a future win or profit. Review the current engine’s record and trading configuration rather than relying on a claim from an older release.