65% win ratio with 1:1 risk/reward?

I'm glad I found your thread cause I have just came to the same conclusion. I've been working for a month now on backtesting and forward testing a few systems.

I was trying for the classic 3:1 reward/risk ratio. The first system ended up with a 60% 1.5 ratio. It was the best I could do. Then I tested hundreds more ideas and none were even close. So I now think 3:1 ratio is really rare.

I tried to get my system up to 2.0 ratio. I even told myself 2.0 was my minimum to trade, but all I did was make things worse. The percentage would go down, or the avg gain per trade would go down. It just didn't give me more money.

 

To make matters worse cunparis, your 60% 1.5 ratio BACKTESTED system will more than likely fall apart in real life trading...I can almost guarantee it.

Why?

"Whenever you run an optimization on noisy data on a fixed time interval, the best performance will almost always be due to curve fitting the noise and signal. This curve fitting is also called “Data Mining” or “Data Snooping”. A fixed number of prices on a fixed time interval has any spurious movements which are also called noise. When we run many different combinations of the input parameters, the best performance will be from those cases that are able to capture profits from BOTH the spurious movements AND the price or signal dynamics. While the signal dynamics, if there, may repeat, the same spurious price movements will NOT repeat in the same fashion in the out-of-sample data. If the spurious movements that were captured by a certain set of input parameters were a large part of the net profits, then choosing these input parameters would produce losses when traded on future data. These losses occur because the spurious movements will not be repeated in the same way. The input parameters chosen from the test section performance are “cherry picked” to perform well on only those exact same spurious movements.
This is why curve fitted systems with no out-of-sample testing cause losses in the out-of-sample section [i.e., future] from something that looked great in the test section.[i.e., past]" - Dennis Myers

The above reality is the bane of trading, and the reason why countless newbs waste years spinning their wheels.

 

Who says the data is noisy or that it's on a time interval??? You're just quoting someone else's ideas about backtesting and turning it into a gross generalization.

 

I think the above two sentences was the key to that insightful quote. When looking at trading methods where reward exceeds risk considerably and with a low win rate, they are dependent on events which occur much less fequently, and therefore with much less consistency to past results (at least in looking at distributions of returns on a probability map).

Tying in what an earlier contributor on this thread said about a casino's edge in roulette, I have personally found much better trading performance in ES/ER2 when using methods with a high win rate but a risk/reward ratio >1.0. This is contrary to the trading mantra of keeping your reward larger than your risk.

For me (and I'm referring to daytrading), looking at the standard distribution (variance) of daily profitability is critical (more so than the size of the mean of daily profitability). I look at daily and weekly mean and SD of returns to smooth out trade by trade results. When the daily SD of profit is smaller than the mean of daily profit, then there's a 66% probability of positive daily results (assuming a "normal" bell curve distribution of returns....again which is more accurate for a larger sample size). I prefer to find methods where the daily mean profit is 2X or larger of the SD, because this gives a 92% probability of daily profitability. For me (and I think for most traders), consistency of real time performance is paramount to actually achieving something close to theoretical results. Get a larger variance in daily/weekly returns and you are much more likely to undermine your method's performance, no matter how good it looks in theory.

 

I guess I wasn't awake yet after a long July 4th night.

When I wrote "standard distribution" above, I of course meant the statistical term "standard deviation".

 

I agree with this. I can tell you what I've done to try and combat it.

I have been testing with YTD data because we have had a correction, a flat period, a bull rally, and now another correction. So 2008 is pretty representative (unlike 2005 & 2006 for example).

I then test each phase differently. I test a bull rally, I test a decline, I test a flat range.

My goal is to make my system keep out of bad trades. This is my goal more than trying to find the best trades.

Then finally I try to test against other data, for example from 2007. This way I test against data that I have not "curve fitted".

I have only been doing this for a month. My first system, the one 65% 1.5 ratio hasn't done so well. But then again in the past few weeks we've had a big decline.

The new system I'm working on is much more selective. Less trades. But I just started forward testing.

My biggest fear is that some of the stocks have low volume days and since my systems are daytrading, I'm worried about my participation affecting the price. I got burnt on that the other day when I went to exit and drove up the price. For me this is the biggest risk.

 

That's why you are a loser.

I just answered the question to your first post, but you didn't even take the time to read and THINK about what I posted.

I know how to get one to a high win %, but I dare not reveal it to you. I'll just let you keep spinning your wheels for years and years to come..

 

Come on guys, this is an interesting thread let's keep it civil.

I'm sure everyone would appreciate any hints to helping achieve a better system.