Correlation-of-symbols-in-portfolio

QUOTE:
... want to select the equity curves/drawdowns of the symbols that are the least correlated in order to have a diversified portfolio ...

Correlated against what exactly? What's the regression function you plan to use to define the correlation? In other words, what function defines that two stocks are highly correlated?

Let me add, you can define your own ScoreCard functions with your own metrics. https://www.wealth-lab.com/Support/ExtensionApi/ScoreCard And one of those metrics can be an R-squared of some functional relationship you create.

---
If you want to diversify by sector and industry (Do you?), IQFeed carries the NAICS classification codes and Norgate carries the GICS classification codes. I trade the Fidelity Select Sector funds, so I'm an anti-diversification investor trading the sector funds I think are doing the best at the moment. But there's some high turnover in trading sector funds because you are always chasing the highest performing sectors.

The correlation could be among themselves or compared to a benchmark like SP500.
See charts below. The yellow line is SP500 equity curve and the others are other industries/stocks.

Tradestation has a correlation indicator (See link below)
I wonder if there is one in WL* or could be coded?

https://help.tradestation.com/10_00/eng/tradestationhelp/elanalysis/indicator/correlation_indicator_.htm

Did you try typing "Corr" in the indicator list filter?

Thank you I didn't but I do now.
Is there a similar one for equity curves?

You want to obtain the correlation of the equity curve with what else? I guess from Post #1 you're looking for the equity curves of each of the symbols in a backtest?

To correlate you need two things. What two things do you want to correlate?

If you looked at the individual symbol equity curves in my best datasets, they all correlate exactly the same way. They all have positive slopes telling me they are all making money over time. And if they don't correlate with a positive slope, those symbols are removed from my top datasets because they aren't steadily making money. That's the only correlation I look for in the equity curves--a steady positive slope. I hope all your equity curves look like mine.

Check out this discussion about the equity curve at https://www.wealth-lab.com/Discussion/What-is-the-most-important-backtest-result-you-look-for-in-a-strategy-10787

I'm too under impression that topic starter is after correlation of the individual symbol's equity in a portfolio backtest, i.e. symbol to symbol. We don't seem to offer it. On a related note, the PowerPack > MetaCorrelations visualizer shows a correlation matrix of child strategies in a MetaStrategy. It helps determine the ones that have a low or negative correlation in the quest for building a robust portfolio of strategies.

QUOTE:
... shows a correlation matrix of child strategies in a MetaStrategy ... [to] determine the ones that have a low or negative correlation in a quest for building a robust stable portfolio of strategies [within the MetaStrategy] ...

because you don't want trades needlessly bouncing around between like (i.e. highly correlated) child strategies.

Or maybe the intent would be to enhance the By Symbol Visualizer, to add a new column showing the correlation between each symbol's equity curve and the strategy overall equity curve.

QUOTE:
... enhance the By Symbol Visualizer, to add a new column showing the correlation between the symbol's equity curve ...

And why would you want to do that? It's not the intra-symbol correlation that matters here. It's only the steady positive slope of the equity curve that matters, which can be measured with an R-squared of the simple linear regression line of the equity curve.

And if a By Symbol visualizer is the goal, he should be voting for the https://www.wealth-lab.com/Discussion/Symbol-Rankings-tool-9116 feature request.

Why is correlation important. If the individual components within a portfolio are
correlated you won't get any benefit from diversifying across them in fact they're likely to increase your risk
because when one of them experiences a loss the other correlated components are
likely to incur a similar loss so being able to measure the correlation between components
before we include them in our portfolio is essential.
If you don't and you include overly correlated components
this could increase risk and drawdowns
instead of reducing them.

QUOTE:
Or maybe the intent would be to enhance the By Symbol Visualizer, to add a new column showing the correlation between each symbol's equity curve and the strategy overall equity curve.

Could we add this as a feature?

Done!

QUOTE:
Why is correlation important. If the individual components within a portfolio are correlated you won't get any benefit from diversifying across them ...

Agreed for NAV Price action. But for the ideal equity curve, you never want to get into a trade that's going to produce a loss in equity. What's the point of that?

In all cases, you always want a winning trade, so the equity should always be going positive. It's the positive slope, not the correlation (or loosing trade events) that matters with the equity curve.

QUOTE:
But for the ideal equity curve, you never want to get into a trade that's going to produce a loss in equity. What's the point of that?

So you're saying you do not need diversification only pick your best strategy and only trade one if the others are not giving you the highest results?

QUOTE:
In all cases, you always want a winning trade, so the equity should always be going positive. It's the positive slope, not the correlation (or loosing trade events) that matters with the equity curve.

This is true even if you include correlation in your factors. The correlation factor serves to minimize drawdowns in the portfolio,

QUOTE:
On a related note, the PowerPack > MetaCorrelations visualizer shows a correlation matrix of child strategies in a MetaStrategy.

If I do a MetaStrategies backtest of a few strategies I get what I was looking for. See images below. Thank you.

QUOTE:
So you're saying you do not need diversification. Only pick your best strategy and only trade one if the others are not giving you the highest results?

You need diversification for sector, industry, NAV Prices, and between child strategies; but you want all your symbol equity curves to ideally look alike (highly correlated) with a steady positive-going slope. You want to entirely avoid trades that are going to give you losses and reduce the slope of your equity curve.

And yes, you do want diversification between child strategies for the reason I outlined in Post #9. But we don't want diversification in the equity curves; we never want to loose equity. Diversification is not beneficial for all situations.

QUOTE:
you want all your symbol equity curves to ideally look alike (highly correlated) with a steady positive-going slope.

Of course that would be the intent. The correlations factor is only to avoid drawdowns that occur in the same periods. You will pick an choose the strategies whose drawdowns occur at different times in history so you do not incur losses at the same time. It is a more balanced portfolio.
In the diagram below I would select the Yellow and the Red equity curves since the drawdowns occur at different times rather than selecting the Red and the Blue wich drawdowns happen at same times.

QUOTE:
The correlations factor is only to avoid drawdowns that occur in the same periods.

Now that makes more sense. In that case, I would be modeling the correlations of the drawdown curves instead of the equity curves.

QUOTE:
It is a more balanced portfolio.

So have your considered a Modern Portfolio Theory (MPT) approach that's more Buy-and-Hold to avoid turnover but balance your portfolio so you maintain a steady, positive portfolio equity curve without any trading? The two MPT methods I'm familiar with are Efficient Frontier and Zero Beta, but there are others. We should start a new topic to discuss MPT methods. And these MPT methods aren't something WL really does out-of-the-box. WL is more of a backtesting and trading tool.

QUOTE:
I Now that makes more sense. In that case, I would be modeling the correlations of the drawdown curves instead of the equity curves.

That would be nice. BTW that was always my intention if you re read my first post in this thread. Maybe Glitch can add it as a feature in WL8 sometime in the future.