V11 – Sharpe Ratio, Meaningful or Not?
- This topic has 3 replies, 2 voices, and was last updated 3 years ago by
.
-
Hoping to get a discussion going on Sharpe Ratio as it is a new feature available in V11.
Judging from the information below, we need to make sure we are not mislead by high Sharpe Ratios?
The extract below is the 4th comment on this link …
https://www.myfxbook.com/en/community/general/sharpe-ratio-highest-myfxbook/266265,1
1. The Sharpe Ratio is claimed to be used more as a performance comparator rather than as a standalone risk indicator. In fact, it does not measure risk at all. But even as a standard measure across portfolios/accounts, we would be committing a logic fallacy. I elaborate after the next point.
2. The nature of using standard deviation anywhere in its calculation implies a conformity to normal distribution, which is not the case in account performance. A martingale system with consistent gains, but with no stoploss, will have a large Sharpe Ratio, even though it is exposed to complete destruction every time it trades. A profitable system like the Turtle trading system (multiple small losses, with big and infrequent large gains) will have a tiny Sharpe Ratio compared to a system with consistent but small gains.
3. Using the Sharpe Ratio to compare systems is like trying to determine which swims better, a spoon or a mountain? It is simply not a significant qualifier.
4. Okay, now we look at the mathematics, to get a Sharpe Ratio greater than 1, you would need an average of the returns to be greater than the standard deviation of the returns.
You would achieve a small standard deviation if the absolute $ returns per period is similar, resulting in a large Sharpe Ratio.This is achievable if each trade/period yields the same $ amount regardless of the growth or loss in the balance(via organic P&L or deposit/withdrawals).
This is silly because it means the system does not increase or decrease the trade sizes in accordance with the balance of the account, i.e. zero money management, and yet it has a large Sharpe Ratio AND lower % yield.
This is illustrated by comparing ‘Fixed $10 Return’ and ‘10% growth/Period’ tables. The SR for ‘Fixed $10 Return’ is a whopping 316.54 with net profit of 100%, while the SR for ‘10% growth/Period’ is 3.49 while it’s net profit is greater at 159%.Comparing ‘Fixed $10 Return’ and ‘Fixed $1 Return’, they both have the same Sharpe Ratio, when the former has 10 times more profit.
Comparing ‘10% growth/Period’ and ‘1% growth/Period’, the former has 15.9 times more profit but the SR is 3.49, while the latter has a higher SR at 33.20.
Comparing ‘Irregular Return’ and the rest, it has the second smallest SR 0.60, but the highest net profit at 938%.
Comparing ‘Turtle-like Return and ‘1% growth/Period’, the former’s SR 0.15 is but a fraction of the latter’s SR 316.54, but it has 5.1 times the profit.5. Also, the longer an account is open, the lower the Sharpe Ratio will go simply because the probability of different returns for each period increases.
All said, why is a statistical measure for particle physics being used to measure fund performance? It’s so absolutely ridiculous.
Part of another comment on the same page …
And there’s the flaw. Sharpe’s ratio does not take floating DD into account. And more importantly, the SR only ‘works’ on normally distributed returns (profits). And very few forex systems can be classified as such. Finally, if you have a very short track record, SR might show absolutely unreliable values as it’s independent of track period.
1 user thanked author for this post.
For me Sharpe Ratio is all about checking long term investment ideas compared to just putting your money somewhere safe. It was created for this purpose and I’m not sure it has much use for testing 1 second time frame scalping strategies.
1 user thanked author for this post.
All these ‘new’ toys are there to try to persuade traders that what they have discovered will work in to the future. Personally I think that most of them are worthless. What use is a Sharpe ratio on a curve fitted strategy? Making sure that your strategy is not curve fitted is the most important thing and this can only be done by testing on in sample and out of sample in my opinion. Also checking that your variables work over a broad range is also very important.
The other day I created a strategy on a 1 hour chart with just one variable and one exit condition. The variable was profitable over a broad range of values. Sometimes trades were open for quite some time so I was obviously missing many possible trades whilst those positions were open – these would be my out of sample trades. So I then tested the most profitable variable value by using my SEB technique to create 24 strategies a day and for every day of the week. So that’s 144 strategies. The original single back test returned just a few hundred trades and the 144 strategies returned thousands of trades – so that is a very small in sample and a very large out of sample. The out of sample results were only slightly worse than the in sample and very profitable. To me that sort of test is worth 10000% more than any Z score will ever tell me.
Just my humble opinion – of which I have many! (Especially when it comes to tea!)
Find exclusive trading pro-tools on 
