What defines a successful strategy?

[taowave]

How did you approximate a 15 percent return with 20 percent max drawdown equating to a Sharpe of 1.5?

Agree with the earlier points about referencing a benchmark. The benchmark should be appropriate. If you're just trading US stocks, then the S&P 500 probably makes sense (you'd do much better with a wider set of assets, but thats another story).

15% return on 20% maximum drawdown works out at a sharpe ratio of 1.5. That strikes me as rather optimistic. Many highly diversified sophisticated hedge funds get a sharpe of 1.0 or less. I'd expect my return to be significantly less for that drawdown.

FWIW I expect to make between 10% and 20% a year over the very long run, with a maximum drawdown of 40%.

Yes, you should be compensated for the time you put in, over and above getting a decent return on capital*. Let's say you expect to make 5% a year from buy and hold investing (reasonable if you're limiting yourself to just US stocks). If you make 15% spending all day trading; then you're being paid 10% of your capital** to work say 10 hours a day. If you've got a $250K account you're being paid $10 an hour. That's less than minimum wage in the UK.

* And all of this of course is after commissions and other costs of doing business.
** This is ignoring the extra risk you're probably taking on that makes that extra 1% less valuable than the other 5%.

As I trade systematically with a fully automated system my hourly rate is somewhat better.

But you need to be careful about not over leveraging to get the return you need to properly compensate for your time, if you've got a small account size.


GAT

 

[globalarbtrader]

How did you approximate a 15 percent return with 20 percent max drawdown equating to a Sharpe of 1.5?

If you bootstrap random data you find the max drawdown is roughly twice the annualised standard deviation of returns (see my post here).

So we've got:

max drawdown=20%
implies ann. std. dev= 20% * .5 = 10%
SR = 15% / 10% = 1.5

GAT

 

[OddTrader]

If you bootstrap random data you find the max drawdown is roughly twice the annualised standard deviation of returns (see my post here).

So we've got:

max drawdown=20%
implies ann. std. dev= 20% * .5 = 10%
SR = 15% / 10% = 1.5

GAT

My guess is in realily, it would be very unusual for a 15% return system to get a SR 1.5 with 20% MaxDD. Just 2 cents!

 

[globalarbtrader]

My guess is in realily, it would be very unusual for a 15% return system to get a SR 1.5 with 20% MaxDD. Just 2 cents!

Well if the returns were drawn from a gaussian distribution (which is an assumption you can argue about, if you like) then on average that's exactly what you'd see. Unless you're saying that the SR of 1.5 is unusual; I'd agree this is highly optimistic also in my opinion.

http://www.risk.net/data/Pay_per_view/risk/technical/2004/1004_tech_atiya.pdf

GAT

 

[nonlinear5]

Quantitatively, I mean.

Perhaps an easier thing to quantify is not the success, but the reciprocal of it, the failure, and more specifically, the probability of the failure to produce return above 0, or above some other reference return. The lower that probability, the lower the risk of failure, and thus the higher the success.

From this perspective, I like the Stutzer index. It has some desirable properties:

-- for normally distributed returns, it equals to Sharpe's ratio
-- unlike the Sharpe's ratio, it does not penalize upside volatility (i.e. abnormally high returns)
-- unlike the Sharpe's ratio, it is appropriate for non-normally distributed returns
-- unlike the Sharpe's ratio, it does not violate the stochastic dominance (or in plain English, it does not produce nonsensical results)

Here is Stutzer's original paper:
http://www.yats.com/downloads/StutzerIndex.pdf