"Buy low, sell high". Who doesn't use that?
System made 124% in 1/2 a year but ...
- Thread starter zanek
- Start date
"Buy low, sell high". Who doesn't use that?
Most systems with 49% winning trades will never have the same sharpratio as 90% winning trades. So you take a hypothetical situation that is almost non existing, but favorable for your explanation. But theoretically it is indeed possible to have the same results with 49% winning trades instead of 90%.
If you have a good system you will never have the same results with 49% winning trades as with 90% winning trades. Because of the fact that 90% of the time you know what the market will do you will close losing trades quickly. So your sharpratio will never come close to mine.
I have seen a short term system with 90% profitable trades. In the 10% losing trades the system gave back less than 10% of the total profits.
Can you show we an excel with figures that proof that you are right?
I can show you an excel where net profits are equal and profitable trades are 49% and 90%. Sharpe ratio for 49% was 0.460983 and 1.027615 for 90% profitable trades.
But I must admit you are the expert because you know how to write Sharpe ratio correctly.
Good points. Win percentage is way overrated.
Also, the OP mentioned 6 losing tradings in succession. That is to be expected. In fact, i think statistics will predict something like one run of 10 or more in ayear of trading a 50% system. I may have the numbers a little wrong, but the point is a long run of losers is virtually guaranteed.
Typical trend following systems used by commodity traders usually will have a win percentage around 30%. They will certainly have large drawdowns, meaning they can only be used with large accounts that can trade them across multiple markets.
I can't agree with some of the other posters that you need years or decades of backtesting. It depends on the timeframe you want to trade and the logic of your system. If you are trading off weekly or monthly bars, then yes, many years data is necessary. If you are trading 5 minute bars, i don't see the need.
Market conditions can obviously have a big effect on the results. That is the idea behind using long periods of test data. I think yu can accomplish the same thing by selecting perods to reflect different conditions. Or, you can just trade the system in the environment it likes best. of course, that raises the need to classify the market environment as a system filter, but that is pretty basic.
A far bigger risk is curve fitting. Too much optimization or tweaking can produce great numbers, but they will not be reproducible in real time or out of sample testing. Robust systems typically are simple.
I hope you realize that this does not prove anything. I never said win percentage is related to Sharpe. You can have every possible combination of win percentage / Sharpe ratio imaginable.
When I started, I also followed the win % metric very closely, now I don't pay attention to it at all. I suppose you do have bragging rights if you're right 90% of the time compared to say 55%.
Also, psychologically it's harder to trade a 90% system that fails quite badly 10% of the time, compared to a 60% one which has losers equal to winners.
Curve fitting is the thing that keeps me up at night as it's way too tempting to commit this deadly sin
I've never monitored streaks of losers, I suppose the size of the losers is also extremely important - an intraday system that has 10 losers in a row on July 3rd is mighty different compared to 10 losers on October OEF.