Sharpe Ratio

Adamus

Experienced member
Messages
1,898
Likes
97
I'm having trouble with this Sharpe Ratio thing.

In NinjaTrader, it spits out the Sharpe Ratio for every backtest or optimization run, and so far the value for my strategy are always about 0.10 or below.

Yet NinjaTrader Help says this:

Sharpe Ratio
This statistic returns a ratio that measures the risk premium per unit of risk of your strategy. It can help you make decisions based on the excess risk of your strategies. You may have a high-return strategy, but the high returns may come at a cost of excess risk. The Sharpe ratio will help you determine if it is an appropriate increase in risk for the higher return or not. Generally, a ratio of 1 or greater is good, 2 or greater is very good, and 3 and up is great.

(Profit per Month – risk free Rate of Return) / standard deviation of monthly profits

So by their scale all my backtests so far are absolute rubbish!

Admittedly the best backtest I've got so far has $30K profits and $20K drawdown, so I was expecting poor stats, but this is pretty sad.

Or do I understand it wrong ?
 
Account size should be irrelevant to the Sharpe Ratio, surely?
 
The low values that you get mean that your system does not profit from clever timing of the market but by assuming excess risk. This further means that it may at some point in the future generate even greater drawdown.
 
Basically I'm wondering what sort of graphical representation would best display the Sharpe Ratio of a particular trade history, i.e. what sort of graph would I have to look at to be able to best guess what the Sharpe Ratio is.

I thought it might relate quite closely to equity curves and how close the equity curve is to a straight line from bottom left to top right, but I don't think it's that.

My guess now is that I'd need both the equity curve, and the graph of the distribution of returns.
 
Check the system result against your own calculation. Tradestation does this if you run a system on intraday data the figure it spits out has no relatioship to monthly sharpe.
 
I prefer to look at CAGR/max drawdown, you should aim to get that 0.7-1.0 or higher. I think the issue with Sharpe ratio is if (for example) you trade a breakout system, you will have a large number of small losses with the occasional big winner .... your Sharpe ratio might look poor because the volatility in the P/L is on the upside, not the downside. In other words, your system might make money but the Sharpe ratio is low due to volatile returns. Doesn't mean you shouldn't trade it tho!
 
Aha. That looks like a useful performance indicator to optimise on but I'm not sure NinjaTrader will ever do that. NinjaTrader 7 is currently in beta, so I might be able to persuade them to add it.

Thanks!

Looking at some of the other stats that NinjaTrader pumps out, there are some really useful looking performance indicators. Maximum adverse excursion, maximum favourable excursion and end-trade-drawdown which is drawdown from maximum favourable excursion.

My stats for those are rubbish too - I have MFE of $600 and ETD of $590.

Time to look at my exits I think.
 
The main impediment to successful mechanical trading is the trader him/herself. Analysis and backtesting is all good and well, and you can use whichever stats you like, but the crunch comes when you've had sustained drawdown and get yet another signal which looks like it'll turn out to be a loser.. so you skip it.. and it turns out to be one of the top trades of the year....
 
That is illogical, Captain. (Need a Mr Spock emoticon)

In terms of operating systems, what you describe there is known as a 'kernel panic' and it leads to a core dump and shutdown. In human terms, it is also a type of panic. You are second guessing your system. It's a totally illogical thing to do.

It's easy for me to say this while out of the market, but if you think you know what the market is going to do, then open a position in a seperate account, but leave your system alone.

I can see that it's an easy trap to fall into if you are entering your own orders. Fortunately the software I use will enter the orders, so the temptation to meddle will be that much less.

I have no idea about this, but I suspect that the performance statistics of live trading compared to backtesting will quickly show results up to 50% worse than what you experience in your backtest. Any more than that and my plan would say - "shut down the system". Hopefully I would stick to that, and hopefully the future won't take me there.

With fixed fractional money management, it would happen really quickly if it did happen. Hopefully though I will be able to put together non-correlated systems to buffer it before it does happen.
 
Last edited:
Yeah, if you fully automate, then shouldn't be an issue. My trade frequency is fairly low (5-10 a month) so I haven't bothered to go down that route.
 
The low values that you get mean that your system does not profit from clever timing of the market but by assuming excess risk. This further means that it may at some point in the future generate even greater drawdown.

This guy was able to articulate what I was trying to get at with my draw downs: account size post lol.

How's your system coming along btw?
 
Maybe it's because I'm not familiar with the maths terms, or maybe because his mothertongue's not English, but I just didn't get much value out of that message. As far as I can tell, he's saying that I need to improve my system's timing - which I would love to do -, that I'm assuming too much risk - which is too subjective a comment to be valuable - and that drawdowns in the future are going to be greater than in backtesting, which I already assumed.

So what are you saying re account size? You are talking about account size in regard to the calculation of the Sharpe Ratio, but account size is not explicitly mentioned in the formula. I didn't work out the Sharpe Ratios on my system that I am complaining about - that was NinjaTrader software.

Let me try to work out this: you mean the Expected Return? I guess you can't measure the return if you don't know how much you started with and therefore NinjaTrader must have assumed something to calculate it.

I just checked. For the expected return, NinjaTrader uses the avg profit per month. I thought they had to use a percentage. And they use a risk free rate of return of 0.
 
Look at this hedge fund's Sharpe Ratio:

John W Henry

They're worse than me. I can't see the value of this. It looks like just another measure that doesn't actually tell me anything.

I think you got it right, meanreversion, when you said it's probably the volatility of the returns on the upside. The system I'm tweaking with these Sharpe values has winner:loser ratios of 1:2 and the average winner is roughly twice the size of the avg loser.

I tried to get NinjaTrader to put CAGR into their performance statistics, but they've now frozen development until the next release, so I'll have to stick with what's there, which isn't much.
 
Sharpe ratio was originally designed to compare performance of mutual funds, I think it's a bit flaky with hedge funds. Nonetheles, that's what people seem to look at.

(The famous case of excellent Sharpe ratio was with LTCM which was simply astonishing, 40 pct a year returns with almost no volatility, until such time as they blew up.)
 
(The famous case of excellent Sharpe ratio was with LTCM which was simply astonishing, 40 pct a year returns with almost no volatility, until such time as they blew up.)


Sharpe Ratio -----> drag drag drag -----> garbage can ----> scrunch!
 
Top