Hello,

So there are a ton of different analyses available in WL regarding the backtest of a strategy. I am curious what people prefer to look at when evaluating their strategies. Do you look at profit factor, WL score, Sharpe Ratio, Sortino Ratio, or something else?

I have found myself drawn to the Recovery Factor of a WFO. I feel that this can tell me the risk adjusted return of the strategy I might experience in live trading. However, it almost feels too simple.

So what is your favorite backtest result to look at?

Thanks,

Dandude

So there are a ton of different analyses available in WL regarding the backtest of a strategy. I am curious what people prefer to look at when evaluating their strategies. Do you look at profit factor, WL score, Sharpe Ratio, Sortino Ratio, or something else?

I have found myself drawn to the Recovery Factor of a WFO. I feel that this can tell me the risk adjusted return of the strategy I might experience in live trading. However, it almost feels too simple.

So what is your favorite backtest result to look at?

Thanks,

Dandude

Rename

This is an excellent question. And I do have one production buy-high strategy that optimizes against Recovery Factor with pretty good results.

But the one metric I look for is the

Moreover, if the equity curve has a recent negative slope, I remove that symbol from the "active" datasets. (I wouldn't throw the symbol away totally because it might have a rebirth someday.)

WL8 does have a ScoreCard metric that looks at the slope of the equity curve using a simple (i.e. first-degree) linear regression, but it fits the entire equity curve, not just the most recent years. You could reduce the Data Range to 3 years, then run that equity-curve slope metric though. Be sure you get at least 3 to 5 trades into the equity curve slope calculation; less than that is meaningless.

But the one metric I look for is the

__recent__slope of the equity curve (say for the last 3 to 5 trades). It needs to be__positive__so you know your strategy is making money over time; otherwise, what's the point?Moreover, if the equity curve has a recent negative slope, I remove that symbol from the "active" datasets. (I wouldn't throw the symbol away totally because it might have a rebirth someday.)

WL8 does have a ScoreCard metric that looks at the slope of the equity curve using a simple (i.e. first-degree) linear regression, but it fits the entire equity curve, not just the most recent years. You could reduce the Data Range to 3 years, then run that equity-curve slope metric though. Be sure you get at least 3 to 5 trades into the equity curve slope calculation; less than that is meaningless.

Thanks for that, but what is that WL metric called which you are referring to?

WLScore?

WLScore?

It's called "Slope of Equity Curve" in the Metrics Report.

Let me add, one shouldn't exclusively rely on ScoreCard metrics. I typically step through poorly behaving symbols

However, if you are really sold on using ScoreCard metrics for measuring performance, then you should vote for the feature request https://www.wealth-lab.com/Discussion/Symbol-Rankings-tool-9116 so you can have a listing of symbols and how they rank with each ScoreCard metric individually.

__eyeballing__each of their equity curves without employing a ScoreCard metric. That's a good way to get the feel about how a symbol is performing.However, if you are really sold on using ScoreCard metrics for measuring performance, then you should vote for the feature request https://www.wealth-lab.com/Discussion/Symbol-Rankings-tool-9116 so you can have a listing of symbols and how they rank with each ScoreCard metric individually.

If I'm backtesting a new strategy (vs. live trading) to compare various alternative datasets or parameters, or reviewing the results of an optimizer run, I like to start with the Wealth-Lab score. It's a thoughtfully designed metric that takes a stab at comparative risk / reward with a single weighted score of the most common high-level objectives for a strategy: APR (higher is better), drawdown % (lower is better), and exposure (lower is better). It's a good starting point although I usually drill down to compare additional metrics such as recovery factor (higher is better), which correlates with the TTR calculation (lower is better) that @superticker mentioned in Post #4.

Of course, high APR is desirable, but you might have your own limit for how much drawdown you want to risk and choose a strategy with less drawdown and lower APR and WL scores based on your risk profile. You might also choose to sacrifice a higher WL score for more exposure and a higher APR.

Of course, high APR is desirable, but you might have your own limit for how much drawdown you want to risk and choose a strategy with less drawdown and lower APR and WL scores based on your risk profile. You might also choose to sacrifice a higher WL score for more exposure and a higher APR.

Your Response
Post

Edit Post

Login is required