TestingView logo TestingView.com

Live account

Live results when the weekly backtest stops

1 November 2026 ยท TestingView

Goal

Live trades should use the same written rules as the backtest. When those rules stop fitting new prices, you should see the new test fail while you can still cut the size.

The chain

A London breakout on EURUSD is the example. You write the rules, then replay the past year with the next candle hidden. Every trade that the rules name goes in the first journal. Expectancy on that year is above 0, and the spread and the slippage that you pay are already in it.

You trade those rules live, at the same risk on each trade. While the London session still shows that kind of breakout, a down week on the account is a down week in the first journal too.

Then that year ends. Prices that weren't in the first journal start to arrive, and you don't run another replay. The account is the only record of those prices. When the breakout loses more often, the loss comes out of the account, and you still take the next trade. The first journal shows expectancy above 0.

So the same rules go back on a schedule, and they stay frozen. You replay prices that you didn't use while you wrote the rules. You read that journal before you decide to keep the live size. Those prices were outside the first test. This is an out-of-sample test.

But the dates are yours to choose, so you pick a stretch that you remember as clean. The out-of-sample test comes back looking like the first year. A bad month doesn't get replayed. The live size stays on.

So the out-of-sample test starts at a date that you don't choose, inside a range that you wrote down first.

A short green run can still come first. You take the next live trade at a larger size. A year-long journal can hold green weeks inside a later drawdown, and a short out-of-sample journal can do the same.

The live size stays at the risk from the first journal. The manual backtesting post says a sample should include a losing streak, a real drawdown, and more than one kind of market. The out-of-sample journal should get there before the size changes.

If you add the new trades to the first journal, one expectancy covers the old year and the new prices. It can stay above 0. The bad month doesn't show up on its own.

The first journal stays as it was. You copy its expectancy and its drawdown onto a note, and the out-of-sample trades go in a separate journal.

Later that separate journal's drawdown is deeper than the note, or its expectancy isn't above 0 anymore. The live size stays on because you call the stretch temporary. The account pays for prices that the first journal didn't hold.

The note is a rule, written when you copy the 2 figures, before you open the separate journal. It cuts the live size if the new drawdown is deeper, or if the new expectancy isn't above 0. This is a loss limit.

The breakout still looks fixable. You change a filter until that same journal is green again, and the next live trades follow rules that were fitted to the weeks that just failed. Later prices don't have to match those weeks. This is overfitting.

Once the size is cut, the old rules stay frozen. A change starts a new replay on a date range that you didn't use while you changed the rules. The size stays cut until that range passes the same loss limit.

A month can pass, and you don't replay anything in it. The weekly backtest post says that on a quiet stretch the winners stayed in memory and the losers faded. The account can hold trades that the written rules don't name. You don't have a list of the trades that the rules would have taken, so a down month doesn't tell you whether the breakout failed or you added trades.

You replay those dates with the next candle hidden, and you take only the written breakout. That list goes beside the fills that the account took. A fill that isn't on the list is a trade that the rules didn't ask for.

The month already happened, so on the replay you know the breakouts that paid. You skip the ones that you remember as losers. The replay list starts to match the account, and the extra trades don't stand out.

You write the decision before the next candle is shown. A setup that you skip gets a row too. A loser that you drop only after the candle is visible doesn't enter the journal.

The trades on the 2 lists can match, but the account can still finish worse. The replay filled at the chart price, and the account paid the spread and the slippage. If the replay's statistics leave those costs out, the replay can show a gain for a week that lost money on the account.

You put the spread and the slippage that you actually pay into the statistics. You read expectancy again. The loss limit uses the figure from after those costs.

Both journals are TestingView sessions. The unchosen start is ChronoShuffle. Slippage and spread in the statistics are on the Trader plan.

The whole design

  1. Written rules, traded live at the same risk: the account follows the first journal.
  2. First journal, left as it was: expectancy and drawdown are copied before the out-of-sample journal is read.
  3. Out-of-sample journal, started on a date that you don't choose: new prices get their own record.
  4. Length of that journal: a short green run doesn't change the live risk.
  5. Spread and slippage in its statistics: the loss limit reads expectancy after the costs that the account pays.
  6. Loss limit: cuts the live size if the out-of-sample drawdown is deeper than the copied one, or if the out-of-sample expectancy after costs isn't above 0.
  7. Frozen rules: a change starts on a date range that you didn't use for the change, and the size stays cut until that range passes the same limit.
  8. Replay of the dates that the account traded: a fill that the rules didn't name stays out of the comparison.

What the story leaves out

You wouldn't build the routine by waiting for each of these failures. Costs come last in these links, because the spread and the slippage show up once the 2 lists match. When you run the check, those costs go in before you read the loss limit.

A forward test uses prices that arrive after you freeze the rules, on paper or at a small live size. The manual backtesting post separates that from a replay of prices that are already on the chart.

A funded account can require a smaller loss than the first journal's drawdown.

Starter caps a session at 20 trades. Explorer caps it at 50. An out-of-sample journal that should include a losing streak may not fit in that cap. Trader doesn't cap the trades in a session.

TestingView records the replay. It doesn't pull fills from a broker. You bring the account's fills yourself.

Next month's prices aren't in the out-of-sample journal.

If you change timeframe during the replay, turn Prevent Candle Previews on. The hindsight post covers what that hides.