TL;DR. The Portfolio Builder combines multiple validated strategies into one system whose equity curve is the weighted sum of theirs — but whose drawdown is usually much smaller than any one of theirs alone. The catch is that the gain comes from uncorrelated strategies, not just multiple strategies. The whole point of this page is recognizing that difference.
What you'll be able to do
- Stack 2–8 strategies into a portfolio and read its combined equity, drawdown, and correlation matrix.
- Decide between equal-weight, volatility-parity, and Sharpe-weighted allocation, and know which fits which situation.
- Spot two strategies that look diversified but are quietly the same trade — and fix that before it shows up live.
Portfolio Builder — combined equity curve and correlation matrix
What it lets you do
A portfolio of strategies is not just a longer list. The whole reason to combine systems is that when their returns don't move together, their combined drawdown is smaller than any individual drawdown. That's the math of diversification — and it's why a small basket of mediocre-but-uncorrelated strategies often beats a single great one over the long run.
The Portfolio Builder lets you:
- Stack strategies — drop in backtests you've already validated in Analytics.
- See the combined system — equity curve, drawdown curve, KPIs, all summed (or weighted) across components.
- Audit correlation — the heatmap shows which strategies are quietly duplicating each other.
- Set allocation weights — equal, volatility-targeted, or Sharpe-weighted.
- Detect conflicts — when two strategies want to be long and short the same instrument at the same time, the engine nets the positions and shows your true exposure.
Where to find it
From the dashboard: Portfolios → New Portfolio. From inside any strategy: Add to Portfolio in the toolbar. Both routes land in the same view.
Entry points to the Portfolio Builder
Step-by-step
1. Pick the components
Drop the strategies you want to combine into the portfolio. Each component must already have a finished backtest in Analytics on overlapping date ranges. The portfolio's evaluation window is the intersection of all components' windows — strategies tested on different dates can be combined, but you'll lose the early or late periods.
Reasonable starting basket: 3–6 strategies. Fewer than 3 doesn't smooth the equity curve enough to be worth the bookkeeping; more than ~8 starts adding components whose marginal correlation contribution is small and whose execution complexity is real.
2. Read the correlation matrix
The portfolio view immediately shows a correlation heatmap of the components' daily return streams. Each cell is the Pearson correlation between two strategies' daily returns over the shared window.
What to look for:
- Correlation under +0.3 between any pair — that's where diversification benefit lives.
- Correlation above +0.7 between any pair — those two strategies are largely the same trade. Drop one, or trace why they overlap (same indicator, same symbol, same regime).
- Negative correlation is the holy grail, and rare. Strategies that genuinely hedge each other are usually intentional (a long-bias system paired with a short-bias system on the same instrument, or trend-following paired with mean-reversion).
Correlation matrix — green low, red high
Two strategies with a +0.85 daily-return correlation aren't a portfolio — they're one strategy you're paying double the slippage to run. The fastest portfolio improvement is usually deleting a redundant component, not adding another one.
3. Choose allocation weights
Four built-in weighting schemes, plus a manual override:
| Scheme | What it does | Use when |
|---|---|---|
| Equal weight | Each strategy gets 1/N of the capital | You don't trust the lifetime metrics enough to over-weight on them. Often the right default. |
| Volatility parity | Each strategy's position size is scaled so every strategy contributes the same daily standard deviation to the portfolio | One strategy is much higher-volatility than the others and would otherwise dominate the result. |
| Sharpe-weighted | Weights proportional to each strategy's lifetime Sharpe | You're comfortable assuming the past Sharpe predicts the future. Often optimistic. |
| Inverse-DD weighted | Weights inversely proportional to max drawdown | You care more about smooth equity than peak return — typical prop-firm sizing. |
| Manual | You set the weights | You have a thesis about which strategies are most likely to keep working. |
Volatility parity is the most common professional choice. Equal weight is the most common honest choice.
4. Read the combined output
The combined equity, drawdown, and KPI panels are the Analytics view applied to the synthetic portfolio. The four numbers worth checking first:
- Combined max drawdown — is it materially lower than the worst individual component's? If yes, the diversification is working. If no, you're carrying components that don't actually diversify.
- Combined Sharpe — usually higher than the average individual Sharpe, because of correlation smoothing.
- Worst month — what's the deepest single-month return? This is the number that determines whether you stay in the seat.
- Component contribution — pie chart of how much each strategy contributed to total P&L. If 80% comes from one component, the portfolio is a single strategy in a costume.
Combined equity curve and per-component contribution
5. Resolve conflicts
When two strategies want to be in opposing directions on the same instrument, the engine doesn't double-charge you — it nets the positions. The conflict log shows every time this happened and what the resulting net exposure was.
Some users want strict netting (one position per instrument, ever). Others want each strategy to manage its own position independently in a separate sub-account. Both modes are supported; the trade-off is execution cost (separate positions pay separate commissions) vs. accuracy of attribution (separate positions make it crystal-clear which strategy made which dollar).
Portfolio optimization
The optimizer here does for portfolio weights what the parameter optimizer does for indicator settings. Pick a target — Maximum Sharpe, Minimum Volatility, Maximum Calmar, or Maximum Sortino — and the engine runs a constrained search to find the weight set that maximizes it on the in-sample window.
The exact same caveats apply: weights that are optimal on the in-sample period frequently don't survive the out-of-sample period. Always reserve a held-out window for verification, and consider walk-forward optimization over a one-shot fit.
Configuration reference
| Setting | Default | What it controls |
|---|---|---|
| Components | (your selection) | The list of strategies in the portfolio. |
| Evaluation window | Intersection of all components' windows | Date range used to compute the combined metrics. |
| Allocation scheme | Equal weight | How capital is split. See table above. |
| Conflict mode | Net | Net (combine opposing positions) or Sleeve (each strategy in its own sub-account). |
| Rebalance frequency | Monthly | How often the engine snaps weights back to their targets. |
| Total capital | Sum of component starting equity | The portfolio's notional bankroll, used to convert weights into position sizes. |
Tips and pitfalls
- Don't add a strategy that doesn't improve the combined metric. A new component might have a great individual Sharpe but if its correlation to the existing basket is +0.9, it's not actually adding diversification. The portfolio's combined Sharpe is what counts.
- Be skeptical of "free lunch" combinations. If the combined Sharpe jumps from 1.2 to 2.5 by adding one component, look hard at correlation stability — historic correlation often understates correlation in crises, when you most need diversification. The strategy correlation page covers crisis-correlation modeling.
- Watch the worst week, not just the worst month. Monthly returns can hide a week where everything blew up simultaneously and recovered. Open the daily return series.
- Cap any single component at ~40%. Even if Sharpe-weighting says one strategy deserves 70% of the capital, concentration risk usually outweighs the math. A 40% cap rarely costs much performance and prevents most disasters.
Pro pattern: build for the worst regime, not the average
A common professional habit: pick the worst single year in your shared window and audit how each strategy performed in that year. A "portfolio" where four of five components were in deep drawdown simultaneously in 2022 is not a portfolio — it's one trade you've made five times. Rebuild the basket around components that survived that year, even if their lifetime Sharpe is lower.
How this fits the workflow
You design strategies in the Visual Builder, validate each in the Backtest Engine and Analytics, then assemble them here. From here you typically run forward testing on the portfolio as a whole before exporting the executable for live trading.
The portfolio construction page covers the theory in more depth; the position sizing page covers how each component's sizing interacts with the portfolio's total bankroll.
A portfolio's max drawdown is almost always smaller than the worst individual component's. That's the only structural reason to run one. If your combined drawdown is the same as your worst component's, the basket isn't a portfolio — fix the correlation, not the components.