Please enlighten me more about Sharpe ratio, which is said to measure systems with the best reward to risk ratio. I was looking at the top systems by Sharpe ratio and two on the list (Wait and Ambush & 2000v Euro Trading System) had a negative net profit yet were in the top 12! How can that be? Please explain.
One important thing to understand about the Sharpe ratio is that math calculations behind the ratio really require at least 9 months of performance data. C2 tries its best to calculate a Sharpre ratio – even when there’s only a few week’s of system performance – but the numbers will often be a little nonsensical.
I’ll try doing a google search for the definition of the sharpe ratio a bit later.
MK
I think I figured why those two systems (Wait & Ambush & 2000v Euro Trading) have high Sharpe values, though having negative net profit. The Sharpe ratio must be based on closed trades only; both of these systems as of today have a large negative open equity, though the closed trades balance is positive. However the negative open equity exceeds the closed trades balnce and both have negative balances overall. As you said, time (9 months or so) will prove if a Sharpe value is valid, but I think it is still very helpful that you include this factor.
I am also a little confused by the sharpe ratio. My system CTS CLC System, had 7 closed profitable trades with no losing trades so far and no open trades currently, but the sharpe ratio is below -1 which is not a good value.
If 9 months are needed to display correct values for the sharpe ratio to be accurate, maybe the ratio should only be displayed for systems with 9 months of history? Currently the value seems to be misleading by giving bad systems a good ratio and potentially good systems a bad ratio. I think no value displayed would be better than an incorrect value.
Regards
- Fanus
Fanus,
Yes it can be confusing, because you have good systems.
However–whether fair or not-- I have noticed Sharpe rating is affected by average size of win, not just win-loss percentage. The fact CTS CLC has 7 winning trades, but is up only 5.5% I believe is part of the negative Sharpe rating. My guess is a winning trade must add at least 1% per trade for Sharpe and within a certain amount of time in order to be positive. It does make sense in that a person can get a great winning percentage by setting a profit target of 0.5% and a stop loss of 20%. Maybe he wins 20 in a row, but that doesn’t make it a great system. One loss and it becomes a losing system. Also I guess the fact it took 18 weeks to go up 5.5% is a contributing factor. If you had gained +5.5% in 1 week, it would definitely be higher. Sharpe ratings always seem to decrease for systems when they generate few trades or becomes inactive.
I know the formula of sharpe ratio from some website.
The definition of the Sharpe Ratio is:
S(x) = ( rx - Rf ) / StdDev(x)
where
x is some investment
rx is the average annual rate of return of x
Rf is the best available rate of return of a “risk-free” security (i.e. cash)
StdDev(x) is the standard deviation of rx
I think the confusing problem may be resolved if we calculate the return of systems every trading day using market price no matter whether the transactions are closed. Then we change the formula to:
S(x) = ( rx - Rf ) / StdDev(x)
where
x is some investment
rx is the average DAILY RATE of return of x
Rf is the best available daily rate of return of a “risk-free” security (i.e. cash)
StdDev(x) is the standard deviation of rx
I uses this modified formula to calculate sharpe rato of the follow performance. (Rf is assumed to be zero)
The sharpe ratio is 0.309
Date Balance (including floating P/L)
day1:100
day2:90
day3:80
day4:70
day5:60
day6:50
day7:40
day8:30
day9:20
day10:20
day11:100
day12:110
The sharpe ratio is 1.535
Date Balance (including floating P/L)
day1:100
day2:90
day3:90
day4:90
day5:90
day6:90
day7:90
day8:90
day9:90
day10:90
day11:100
day12:110
The sharpe ratio is 0.959
Date Balance (including floating P/L)
day1:100
day2:90
day3:95
day4:96
day5:97
day6:98
day7:99
day8:100
day9:102
day10:104
day11:105
day12:110
Comparing the above three cases, we should agree that the first case is the worst case though they all get the same result at the end. Sharpe ratio give the first case the lowest value, as we expect.
Robert
Yes, 5.5 percent in 18 weeks maybe is not much… but this is also then where position sizing come in. By only trading two e-mini contracts per signal, returns will be lower. If I have traded 4, then the return would have been 11. If I have traded 10 the return would have been 27.5. The same system then would have a much higher sharpe ratio. This doesn’t sound right.
You have mentioned that there are systems which are currently unprofitable which have high sharpe ratios. Shouldn’t any system with a postive return, however small, get a better rating than one with a negative return?
I agree that a system which only make 0.5% on a series of trades and lose 20% on one should not get a good sharpe ratio, or any other kind of rating. But if this didn’t happen yet, a system shouldn’t get penalized. The sharpe ratio should not penalize systems for something which might happen. It should only look at what happened already.
Regards
- Fanus
PS: I don’t have 20% stop losses on any of my systems… 
Fanus,
I agree! Anyway - Microforex in his reply before you gave the formula for Sharpe. If a system has only a few trades or traded for a short period of time we probably should take the rating with a grain of salt. It is just one rating to consider, while considering other factors that are important to us.
