Backtesting & results

Portfolios

Backtest several strategies and pairs together as one account — fixed weight or a shared pool, the account settings every leg shares, each leg's own settings, the book's risk policy, risk-balanced weights, and checking whether a new strategy would make the book better.

Updated 2026-10-16·Rev. 2026.10

A portfolio runs several strategies together as one account. Each leg is one strategy on one pair and timeframe, and the same strategy can sit on several legs with different pairs. The backtest returns one result for the whole book (one equity curve, one drawdown, combined metrics), and each leg keeps its own numbers.

Open it from Portfolio in the workspace switcher at the top of the app.

A premium feature

Portfolios are included on paid plans. On a plan without them, creating one opens an upgrade prompt. A portfolio backtest counts as one deep backtest per leg against your daily allowance.

Capital model

The Capital model decides how the book's capital reaches its legs.

ModelHow it works
Fixed weight"Each strategy runs on its own slice of the pool. Weights must total 100%." Every leg is its own account with its own slice.
Shared pool"All strategies draw from the same capital pool; their orders compete for buying power." There's one account, and every leg trades from it.

Fixed weight can Rebalance (Never, Weekly, Monthly, Quarterly, Annual). On each date, capital moves from the winning legs to the laggards, so later trades are sized from the target weights. You can also limit it to legs that have drifted far enough from their target (Every date, or Off by 2 / 5 / 10 pts). Money tied up in an open trade moves once that trade closes.

Shared pool has no weights. Every leg sees the whole pool, and a fill by one leg uses buying power the others then can't. Two consequences follow:

  • An entry can be refused because another leg holds the capital.
  • It's one margin account. Every leg's open loss counts against it, so a leg can be liquidated because another leg's loss sank the pool. See margin, liquidation, and funding.

The account

The editor's Account section holds what belongs to the wallet, not to any one strategy. Everything here applies to every leg:

  • Currency: what the book's cash and results are counted in.
  • Position mode: One-way (one net position per symbol) or Hedge (a long and a short on the same symbol at once). On Binance this is an account-wide setting.
  • Book rails: the portfolio's risk policy. With a shared pool it's judged on the pool every leg trades from. With fixed weight it's judged on the combined sleeves (every leg's equity added up).

Each leg's settings

Each leg row has a settings chip showing its leverage, sizing, and fee. It opens Leg settings, the same dialog as a strategy's backtest settings, covering fees, slippage, margin, sizing, and the leg's own rails. Currency, position mode, and capital come from the portfolio.

A leg copies its strategy's settings when you add it. From then on it owns them: editing a leg changes that leg only, and editing the strategy later doesn't silently change the portfolio. The chip tells you when the two have drifted apart:

BadgeMeaning
EditedYou changed this leg's settings in the portfolio. The strategy is untouched.
Strategy updatedThe strategy's settings changed after this leg copied them.

To catch up, open the leg and choose Sync (or Copy from strategy). The leg takes the strategy's current settings, and the leg's own edits are replaced.

A strategy's own risk policy doesn't come along

A leg never copies its strategy's risk policy. Rails belong to the account, so a portfolio uses its Book rails plus any Leg rails you set on each leg.

Leg rails

In Leg settings, Leg rails add limits to that leg alone, on top of the book's:

  • Soft drawdown and Hard drawdown, measured from the leg's peak.
  • Max margin share: the leg's initial margin ÷ the book's equity.

A leg follows whichever is stricter, the book's state or its own. When the book halts, every leg is flattened. When a leg halts, only that leg closes and stops, and the other legs keep trading.

Reading the report

Run the book with Run backtest. The results column opens alongside the editor, and next to the combined equity curve and metrics, two panels matter most for the account model:

  • Risk shows what the book's and legs' rails did: the state the book ended in, time in each state, exposure, the peak margin ratio, downsized and refused entries, every state change, and a row per leg. Without a policy it says so.

  • Settings is the receipt: what each leg actually ran with. It stays true however the portfolio or its strategies are edited afterwards. Its columns are Leverage, Size, Fee, Slippage, and Pyramiding, plus:

    • Liquidation: the price that would have liquidated the leg at the end of the run, with the rest of its account frozen. In a shared pool this price moves with the other legs. The column also shows how many of the leg's closes were liquidations.
    • Funding: what the leg paid (−) or received (+), open positions included. It reads not charged when the pair's funding history was unavailable.

    A line above the table names the account's currency and position mode, and what liquidation was judged on (for example, the mark price and Binance's leverage brackets, with the date they were fetched).

Better weights: the Allocation panel

An even split of capital is almost never an even split of risk: one jumpy strategy can carry most of a book's ups and downs while the allocation chart shows equal slices. On a Fixed weight book, the report's Allocation panel proposes weights that balance risk instead:

WeightingWhat it does
Your weightsThe split you set.
Equal riskEvery strategy carries the same share of the book's day-to-day risk, counting how they move together (risk parity).
Inverse volatilityCapital in proportion to 1 / each strategy's volatility. Simpler: it ignores how they move together.

The weights are fitted on the first 70% of the run and judged on the last 30%, which they never saw. The panel leads with one sentence: either a weighting that would have done better on that held-back stretch, or your weights are already as good. It only uses how much each strategy swings and how they move together, never their returns.

Use … weights puts the proposal into the editor (it also turns on monthly rebalancing when the book had none, because risk-balanced weights only hold if the book is pulled back to them). Nothing is saved or re-run, and Undo restores your split. Run the backtest again to see the new split for real.

The panel isn't shown on a Shared pool book, which has no weights. A strategy sized by a fixed quantity or cash amount can't be reweighted on a book that never rebalances, and the panel names it.

Would it help? Checking a candidate

A strategy that looks great on its own can add nothing to a book that already holds the same bet, and a modest one can steady the whole book. The Would it help? panel answers the question that matters: pick a finished run of any Strategy, Portfolio, or Universe, choose Check fit, and see whether holding it next to this book would have made the book better.

  • The verdict: Improves your book, Mixed evidence, Does not improve your book, or Not enough shared history, with how sure it is (Strong evidence, Suggestive, Could be luck).
  • What changes: the book's Sharpe, annual return, max drawdown, and volatility, alone and with the candidate.
  • Did it hold up: the result on the last 30% of the shared history, which was kept out of choosing the weight, and how the candidate did on your book's worst days.
  • Caveats such as it moves with your book or it lost money on its own.

The candidate's weight defaults to Steadiest: the share that makes the combined book as steady as it can be (at most 50%). Choose Set to test your own. The check compares runs that already finished, so it's free and runs no backtest. For the cleanest answer, run the candidate over the same dates as your book first.

For a strategy run, Add to book puts it into the editor as a new leg at that weight, and every other leg keeps its share of the rest. A portfolio or universe runs as its own account, so hold it by deploying the two side by side.

The same check is under every finished deep backtest of a strategy: open Would it help a book? and pick one of your portfolios.

Deploying a portfolio

A tested portfolio deploys in one action, as one paper bot per leg. Each bot runs with the settings its leg ran with in the backtest. It carries the book's rails and its own leg rails, enforced the way the backtest enforced them. If the book reaches its hard drawdown, every bot in the portfolio stops. See monitor the Fleet.

Related