Backtesting
A backtest replays a strategy over historical daily closes. Oreno's engine is deterministic: same strategy and data, same result. Every backtest is hypothetical and labelled with its period.
Rules of the simulation
- Trading days are Monday to Friday (UTC). Each day is valued at its close.
- Trades happen at the close of the day they're triggered.
- No look-ahead. Each day only sees closes up to that day. If an asset has no history at the requested start, the start moves to its first real close — and the result says so.
- Trend filters need a full moving-average window of real closes before day one, which can move the start later.
- Money is fixed-point: cash in micro-dollars and quantities at 10⁻¹², so totals carry no rounding drift.
Costs
Every trade pays 5 bps venue fee + 30 bps spread and impact (0.35%), applied against you on buys and sells. The figure comes from comparing Robinhood quotes, Arcus and Uniswap pools with the Chainlink price on weekdays.
Schedules
| Rule | When it runs |
|---|---|
| Initial deposit | First trading day |
| Recurring deposit | First trading day of each week or month |
| Rebalance | First trading day of each week, month or quarter |
| Drift band | On rebalance days, trades only if a position is further than the band from target |
| Trend filter | Checked on rebalance days; switching branches forces a full rebalance |
Metrics
| Metric | How it's computed |
|---|---|
| Return | Time-weighted: deposits are excluded, so putting money in never counts as a gain |
| Max drawdown | Largest peak-to-trough fall of the time-weighted index |
| Volatility | Annualised standard deviation of daily returns; shown from 20 days |
| Annualised return | Only shown for periods of a year or more |
| Benchmark | SPY bought once over the same period with the same costs |
| Costs, trades | Total fees and spread paid, and number of trades |
Every result also lists its period, number of trading days, put-in amount, final value and costs.
On-chain history only goes back to mid-2026, so backtests cover a few months. Short backtests say little about how a strategy behaves over years.