Building in Horizon

Portfolio Optimizer

Keep the line-up fixed. Compare what happens when it trades steadier or faster.

The Optimizer combines the pinned backtests in a portfolio and reads them as one account. The strategy line-up stays the same while you compare two sizing profiles, adjust trade size, inspect the evidence and save the fit you want to export.

The Optimizer uses simulated past results. Its target odds, target time, drawdown and prop-firm economics are estimates from backtests, not a forecast or a promise.

Open it

  1. Go to My Scripts.
  2. Click a portfolio in the left sidebar.
  3. Switch from Scripts to Optimizer.

When you use Build best portfolio, HorizonAI opens this tab for you as soon as the line-up is ready.

Pin a backtest first

Every strategy needs one pinned Horizon backtest: the run it should be known for. A strategy can have many tests across markets and timeframes; the pin tells the Optimizer which result represents it.

Open a strategy, find the right run in its backtest history, and pin it. A strategy without a usable pinned run is left out of the calculations rather than being guessed at.

The balanced line-up

At the top, the Optimizer lists the portfolio's strategies and says Same in both options. The automatic builder has already chosen the strategy count and combination using result quality, drawdown and how well the strategies' daily returns complement one another.

The sizing comparison does not swap strategies to make one option look better. Both options use the same line-up. If you created the portfolio manually, the current approved members are the line-up.

Compare Steadier and Faster

Each profile shows the size multiplier and four consequences from the same backtests:

  • Target odds: how often the replay reached the target.
  • Target time: the typical number of trading days it took.
  • Limit breaches: how often max drawdown or daily loss was broken.
  • Backtest drop: the worst historical fall at that size.

Steadier uses less exposure and leaves more room inside the loss limits. On a prop-firm account, that also means fewer failed attempts.

Faster uses more exposure when the extra size earns a shorter path to the target. For personal capital, limit breaches remain a hard failure rather than a cheap reset. For a prop-firm account, Faster weighs the extra failed attempts and evaluation costs against the expected net payout per trading month.

Faster needs a profit or evaluation target. If no target is set, add one by editing the portfolio.

Adjust trade size

Choosing Steadier or Faster moves the Trade size lever to that profile's suggestion. You can still drag the lever yourself from 0.25× to 10× normal size.

The readout updates the worst historical drop against your limit as you move it. Trading larger multiplies gains and losses together; a target reached sooner can also mean more failed attempts or more limit breaches.

For decision numbers, HorizonAI uses the shared stretch where all included strategies were live at once. This avoids judging today's full portfolio mainly from an older period when only one member had started.

Act on a removal recommendation

Sometimes one strategy's drawdown forces the whole portfolio to trade so much smaller that the target outcome materially worsens. When the evidence is strong, the Optimizer shows One decision would improve this portfolio, names the strategy, and compares the result before and after removing it.

Nothing is removed silently. Click Remove from portfolio only if you accept the recommendation. This changes the portfolio membership; it is not a temporary bench. The strategy itself remains in My Scripts.

Review the evidence

Open Review backtest details for the supporting facts:

  • Worst drop: the largest peak-to-trough fall for the whole portfolio at the current size.
  • How alike: whether the strategies tend to win and lose together, including the closest pair.
  • Teamwork: whether combining them made the ride steadier than the individual parts.
  • Target result: typical time, success rate, and whether failures came from drawdown, one bad day or running out of time.

The chart shows the combined account as a bold blue line and each strategy at normal size as a faint line. Money is split evenly across the strategies and evened up again each day for the analysis.

Save sizing and export

Open Deployment and export to see what the account would actually place. HorizonAI splits the starting balance across the portfolio and translates the selected size into quantities for each instrument. When a futures contract is too large, it can use a micro contract; when an instrument cannot be sized reliably, the row explains why and leaves it out of the export.

Click Save sizing to remember the selected profile and trade-size multiplier. The Optimizer opens with that plan next time, and portfolio exports use it.

Click Export & save for the portfolio's destination platform. HorizonAI downloads a ZIP with README.txt and the files needed for that platform:

  • TradingView: one text file per strategy. Add each file to the market and timeframe listed in the README. See Install your TradingView export.
  • MetaTrader 5: one complete folder containing Horizon Controller and its support modules. Copy the whole folder, compile it and attach only the controller. See Install your MT5 export.
  • NinjaTrader 8: one complete folder containing the portfolio controller and its support modules. Copy the whole folder, compile it and enable only the controller. See Install your NinjaTrader export.

The portfolio controller already contains its markets, timeframes and saved sizing. Test the complete export on a demo or simulation account before considering live use.

If the Optimizer has no numbers

  • "This portfolio is empty": add Approved strategies to it.
  • "None of these strategies has a pinned backtest yet": open a strategy and pin the representative Horizon run.
  • Very little shared history: re-run the members over a common period before relying on the combined decision numbers.

Fine print

  • Everything here comes from backtests: hypothetical trades worked out against past prices, not trades actually placed.
  • The target figures replay historical trading days in many different orders. That helps expose fragile sizing, but it does not predict what happens next.
  • Optimizer suggestions do not place trades, connect to a broker or enforce live account rules.
  • Removing a member does not delete its script; deleting a portfolio does not delete any scripts.

Next

Quant Mode: research and approve more candidates for your next portfolio →