Effeciency of an Auto Strategy

Huh?

I think you're wrong.

But it depends on other factors.

If your average win/loss amount is breakeven and you have an 80% win rate, you've got the grail. High percentage win rate has some advantages.

m

Quote from AnonymousTrader:

If you are running a strategy above 80% winning you are not efficient.

You should either a) increase size or b) increase some other parameters to get the volume up and percentage down to lower 50s.

Efficient Frontier my friend.
More...
 
Would you rather have an 85% win rate?

or....

a higher average win to loss ratio?



It boils down to if you believe your edge is "over fitted"

Michael B.


PS. I look at drawdown mainly...but thats me...I would choose to make half the profits with lower drawdown.
 
Quote from ElectricSavant:

AnonymousTrader,

85% mean average win rate.

When a programmer or mathematician adds the words "mean average" in front of win rate...does this make the calculation different from just "win rate"

Sorry to be so stupid....just tell me to go buy a book, or if you would be kind enough to explain to me what "Mean Average" might mean? and if its different?

Michael B.
More...


i think it's simply inaccurate wording. you are perfectly right.
 
Quote from AnonymousTrader:

We've done some tests (forward with tiny test size) and saw an interesting thing: His strategy had a 85% mean average win rate.

...

It would be much better if the win rate (winning percentage) was in the low 50s.
More...

Some questions:

How did you calculate the averaging result for adjusting some spike performance (as the 85% could be mainly due to spikes, if sample size is small), how many parameters for the averaging performance (as too many parameters could make the 85% not very robust), how many combinations for each parameter were tested (as a small combination could be hard to evaluate robustness), what was the % of profitable combinations (as a small % would make the 85% questionable), and what do you mean (how small number?) by tiny test size?

How many markets have been tested for the system regarding the above questions. More data would be required.

:confused:

Quote from AnonymousTrader:

If you are running a strategy above 80% winning you are not efficient.

You should either a) increase size or b) increase some other parameters to get the volume up and percentage down to lower 50s.

Efficient Frontier my friend.
More...

I would suppose the increase of size can be covered by the combinations mentioned above, by adjusting the values of a parameter. I wonder any adding parameters (that would reduce robustness) would increase size with profitable effects, as that may as well posssibly reduce size (number of trades).

I'm in the process of learning the above. I don't quite understand clearly the relationship with efficient frontier (?). :confused:
 
Quote from tntneo:

trade frequency is important as well, DT-waw mentioned it.
I track profit factor in my context (high frequency trading) because it gives a comprehensive view of the efficiency and smoothness of systems.

if you have 5 trades with a profit factor of 3, it is meaningless. Number of trades is of course important to get statistical significance.

profit factor will give you an idea of the slope of your equity curve. the higher the number, the smoother the curve as well. which can be confirmed with max and average drawdowns.

I won't trade anything below PF=2, and I try to have PF higher than that of course.
More...
Hi tntneo,

Do you use 'Profit Factor' as a synonym for 'Sharpe ratio'?
 
Quote from nononsense:

Hi tntneo,

Do you use 'Profit Factor' as a synonym for 'Sharpe ratio'?
More...

Hi nononsense, I think tntneo was refering to profit factor in distinction to sharpe ratio.

Profit factor = (Total Profit)/(Total Losses)

while

Sharpe = (AvgReturns)/ Stddev(Returns)) ... modifiedSharpe

From my own observations, having a high profit factor doesn't always mean having a high sharpe (though usually it's the case)... this is due to the stddev factor in the sharpe calculation which penalizes the model for outliers. This isn't a factor with PF.

Anyway, don't want to speak for tnt, but I'm pretty sure that's what he was talking about... and I would agree with him as well. I don't think it too wise to trade anything below a PF of 2 either. Regards.
 
onelot is right, that's my definition of profit factor.

I also look at Sharpe ration nononsense btw.
which indeed takes into consideration the smoothness of equity curve.
I like Sharpe as high as possible obviously.

I admit using profit factor more easily, it talks more to traders while Sharpe talks more to investors :)
Both are fine by me.
yes, it's also true that Sharpe will also catch your drawdown while profit factor can't.

but like I said I usually focus heavily on profit factor, max drawdown and average drawdown, while number of trades is high enough to be statistically valid.

tntneo
 
You must log in or register to reply here.
Back
Top