Paper trading now follows the backtest exactly: weights, timing, costs, dividends and cash interest

A paper portfolio that follows a strategy should do what that strategy's backtest does, day by day. We traced the whole path from the result card to paper fills and found several places where it didn't. As of 28 September it does, and we replayed nine strategy types and two allocations day by day through the real paper system to check.

The engine's own target. From 27 September, paper portfolios stopped re-sizing a strategy's picks to equal weights. They now hold the exact weights the backtest would hold on that day, including cash, so a 60/40 strategy holds 60/40 rather than 50/50, and a timing rule that is out of the market holds cash. They trade only on the strategy's own rebalance days and drift with prices in between, as in the backtest, instead of every 28 days.

Only the exact strategy is deployed. Deploying from a result card now sends only the exact engine request behind that card. Before, a card without one could deploy a request guessed from the card's text. That is now refused with the reason, and nothing is created.

The backtest's costs. Strategy trades on paper were charged the paper order model, 5 basis points plus $1 a trade, while the backtest charged 1 basis point of slippage and 0.5 of commission. On a real SPY deployment on 28 September that meant $49.97 of slippage plus $1.00 commission, where the backtest's model would have charged $10.00 plus $5.00. Strategy trades now pay exactly the backtest's cost model. Orders you place by hand still use the paper order model, and portfolios behind published track records keep the cost setting sealed in their commitment.

Dividends and cash interest. Paper portfolios now credit dividends the way the backtest does: on the ex-date, to the shares held at the previous close, after withholding tax, reinvested at that close. Uninvested cash earns the same risk-free rate the backtest credits, once per trading session, and each credit appears in a new "Interest on cash" tab.

One portfolio for an allocation. Running the allocator used to create a separate paper portfolio for each strategy, every time you clicked Run. Now "Deploy as one paper portfolio" creates a single portfolio: each strategy's current target times its weight, netted into one set of holdings, with cash where a strategy is out of the market. Each strategy's share trades when that strategy rebalances, and all shares are brought back to their weights if any drifts further than the band (5 percentage points by default).

No surprise selling. A drawdown rule paused every paper portfolio at a 10% loss and sold everything at 15%, including portfolios following a strategy, which no backtest does. For strategy portfolios it now only alerts you, and you decide whether to pause.

Checked day by day. We replayed 186 trading sessions, October 2024 to June 2025, through the real paper services against real engine answers. For every single-strategy type tested, the paper portfolio stayed within 0.19 basis points of the engine's own backtest curve and made every rebalance on the same day with the same target. The replay also caught an extra rebalance the day after an exchange holiday, which is fixed.

Strategy at work. A new default tab on the paper page shows what the portfolio holds and why, in the words of the rule (for example "Holding SPY because the 50-day average is above the 200-day average"), its holdings against the engine's target, the next scheduled rebalance, and every decision, trade, dividend and interest payment.

Differences we still state. Paper fills use live prices between 15:00 and 16:00 New York time, not the official close. Paper trading holds listings priced in US dollars only. Exchange holidays aren't modelled, and per-position stops and take-profits aren't reproduced; a portfolio following such a strategy warns you.