Can I See an Equity Curve Please?

A client of mine recently got in touch because he wanted me to test a strategy that he’d read about in a trader magazine.

The strategy in question was a mean-reversion strategy that traded liquid US stocks and the sample period used to produce the performance statistics of the strategy was 01/01/2000 – 31/12/15.

The performance metrics of the strategy were provided by the author as…

  • Trades =  3410
  • Average P/L % per trade = 1.26%
  • Win-Rate = 54.66%
  • Average Trade Duration = 3 Days

Alongside the above performance metrics, the author of the article had suggested that this equated to roughly 1 trade per day during the sample period. Seems good enough, right?

Not so fast!

Whilst the above statistics were accurate, what they fail to convey is that nearly 440 of the 3410 trades from the sample period were produced over the course of just 1 day!

The exact date of the 440 trades was October 10th 2008.

The average profit/loss % of the signals traded on Oct 10th 2008 was over 15%, and the win-rate % of all signals traded on Oct 10th 2008 was 96%.

But if we remove Oct 10th 2008 from the sample period, and test the strategy again…we can see how grossly misleading the original performance metrics really are:

  • Average P/L % per trade = -0.03%

The rest of the performance metrics aren’t even worth looking at…The average P/L % per trade over the entire 10 year sample period is negative if omitting just one single day from the data.

The above should illustrate the importance of analysing more than just raw performance metrics.

For example, some questions to ask might be:

  • How evenly are the trades distributed?
  • What is the highest number of signals produced on any given day?
  • If I traded this strategy with actual money, what would a typical equity curve look like?

There’s nothing wrong per se with publishing the performance metrics of a particular pattern or signal, but the above points are clearly worth thinking about.

Lets take the above strategy and apply some realistic assumptions about how me might actually trade the signals.

We definitely don’t want to buy 440 stocks on a single day, so we”ll include a rule which states that we can hold 20 positions at a time and our position size will be $5000 per trade.

When there are multiple signals, we’ll buy the most oversold stocks first…

Here is the equity curve produced by the strategy (to be fair to the strategy author, I’ve reintroduced the data from Oct 10th 2008):

EquityCurveP

As you can see, asking to see a typical equity curve produced by a strategy can help you to avoid being duped by raw performance metrics.

 

Leave A Response

* Denotes Required Field