What are good sharpe & PF values for a system?

[Adamus]

Sure, live trade results beat backtests hands down, no question about it, but when you don't have live results, all you can do is backtest.

And you should really backtest over several years or at least several different market types, to identify what market types your system thrives under, and what kills it.

 

[rafla]

Yep, I have used five years tick data at backtests. Still I don't know what would be "perfect" walk forward period (otimization/test) to my system, too little computer power even for genetic optimizer... and so little time.
So let the live results judge that. Target is to keep this live at least three months (and not as martingale system as those seems to be around too much).

 

[Adamus]

sounds good. the first thing i would look at is the equity curve over the backtest period, and then i'd compare the real results with the backtest results to check that they're all within expectations. i don't bother with sharpe and i hardly pay much attention to the profit factor.

net profit over drawdown is a good stat for comparing systems. but you need a long enough period to see some realistic drawdowns.

 

[meanreversion]

Intuitively I would think a Sharpe ratio in excess of 3 is not sustainable, although I don't have anything to back that up mathematically.

 

[Blaiserboy]

Intuitively I would think a Sharpe ratio in excess of 3 is not sustainable, although I don't have anything to back that up mathematically.

Do you have a list of favorite measurements that you would apply to a system that you would be building or evaluating.......?

Thanks

 

[Adamus]

Your backtest should show a nice equity curve.

 

[Blaiserboy]

Your backtest should show a nice equity curve.

you would rely on that more than the ratios like Sharpe..?

Thanks for your help....

 

[meanreversion]

Sorry, I don't think looking at an equity curve helps, because the eye sees what it wants to. Also, most people run backtests and look at equity on a linear scale.. log scale makes far more sense.

Return/drawdown (or MAR ratio) is a good one. I don't think I've ever seen anyone mention this on t2w.. the assumption is if that system A returns 100% and system B returns 50%, then without question system A is better.. what b0ll0cks.

 

[Blaiserboy]

Sorry, I don't think looking at an equity curve helps, because the eye sees what it wants to. Also, most people run backtests and look at equity on a linear scale.. log scale makes far more sense.

Return/drawdown (or MAR ratio) is a good one. I don't think I've ever seen anyone mention this on t2w.. the assumption is if that system A returns 100% and system B returns 50%, then without question system A is better.. what b0ll0cks.

Thanks for this....

I am looking at several methods for evaluating a system, so far it seems that a combination may offer better results. To know what the ideal combination would be is going to take some detailed research and testing.

 

[meanreversion]

Yes, work is always required. Those who don't trade systems assume it's easy, but it's not.

Furthermore, just because your system worked in the past doesn't mean it will work going forward. There are an infinite number of patterns; you're sure to find one.

 

[Adamus]

Sorry, I don't think looking at an equity curve helps, because the eye sees what it wants to. Also, most people run backtests and look at equity on a linear scale.. log scale makes far more sense.

Return/drawdown (or MAR ratio) is a good one. I don't think I've ever seen anyone mention this on t2w.. the assumption is if that system A returns 100% and system B returns 50%, then without question system A is better.. what b0ll0cks.

MAR? Is it any of these: http://acronyms.thefreedictionary.com/MAR

 

[horace]

Hi rafla,

I've been trading real automatic trading strategies for a long time now. I can't say I know all what there is but I have a pretty vast and sometimes painfull experience.

Based on it I can say two things regarding your post.

1) Sharpe ratio is pretty much useless in automatic trading systems. At least for me. I have non-optimized systems than have been profitable for more than 2000 trades and 14 years and show painfull sharpe ratios (besides most applications don't calculate it correctly). So I don't pay much attention to SR.

2) I have seen no long term profitable trading system with more tan 45% profitable trades. So I would say somethign is going wrong with yours.

Btw, in regards to your question, I'm making money in the long term with sharpes of 0.24, percent profitable of 42% and profit factor of 1.2. All this is (unfortunately) accounting huge expenses (commisions, slippage, renting, etc.) for a reliable platform.

Regards,

Horace

Hi,

What kind of results people are having with ProfitFactor and Sharpe ratio with their system?

Those are current ones from one of mine:

# Profitable 53 (80.3%)
Avg trade duration 2.2 hours
Annual return (compounded) 134.3%
Average win $99
Average loss $275
Profit factor 1.5:1
Correlation w/ S&P -0.226
Sharpe ratio 5.333

(dailysetup.collective2.com)

I think that Profit factor should be more to be succesful?
Back test has shown very high Profit Factor but how much slippage is expected to face at real trades ? Any ideas?

 

[meanreversion]

Hi rafla,

I've been trading real automatic trading strategies for a long time now. I can't say I know all what there is but I have a pretty vast and sometimes painfull experience.

Based on it I can say two things regarding your post.

1) Sharpe ratio is pretty much useless in automatic trading systems. At least for me. I have non-optimized systems than have been profitable for more than 2000 trades and 14 years and show painfull sharpe ratios (besides most applications don't calculate it correctly). So I don't pay much attention to SR.

2) I have seen no long term profitable trading system with more tan 45% profitable trades. So I would say somethign is going wrong with yours.

Btw, in regards to your question, I'm making money in the long term with sharpes of 0.24, percent profitable of 42% and profit factor of 1.2. All this is (unfortunately) accounting huge expenses (commisions, slippage, renting, etc.) for a reliable platform.

Regards,

Horace

Horace, interesting post. How many systems do you trade? Your Sharpe sounds on the low side, a fairly modest bog standard trend system should have Sharpe > 0.5 (provided it has been sufficiently profitable).

In the high frequency arena, Sharpe ratios in excess of 5 are not uncommon, but there is an issue of scalability there. Same applies to stat arb techniques, i.e. pair mean reversion strategies.

I think whatever the approach, it's important to have at least two systems with zero to negative correlation. This **should** provide for better risk adjusted returns, which is what everyone wants after all.

 

[horace]

Hi meanreversion,

True enough. The issue is real trading. I'm completely unable to get better ratios taking into account real world expenses as I said.

Of course I could paint an optimistic scenario with much lower cost, but unfortunately I'm unable to implement them successfully.

Example:

This numbers where extracted form a system that is currently trading profitably in the Eurex Bund market.

I'm "paying" 300€ per year for the renting of the system (simulated because the system is mine and hence I don't pay for it).

I'm paying 14€ per r/t, slippage & platform fees where the industry standard is way way under half of that.

If I accounted for 0€ renting and say 5€ per r/t all inclusive, probably the average trade would be a little bit more than double. I'd get a 22 instead of a 10. Everything would look far better in terms of paper trading, but I would not be honest with myself in terms of risk management (which is, btw, what keeps me in the business for this long)

Regards,

Horace

Horace, interesting post. How many systems do you trade? Your Sharpe sounds on the low side, a fairly modest bog standard trend system should have Sharpe > 0.5 (provided it has been sufficiently profitable).

In the high frequency arena, Sharpe ratios in excess of 5 are not uncommon, but there is an issue of scalability there. Same applies to stat arb techniques, i.e. pair mean reversion strategies.

I think whatever the approach, it's important to have at least two systems with zero to negative correlation. This **should** provide for better risk adjusted returns, which is what everyone wants after all.

 

[DashRiprock]

14 bloody euros???!!!

Horace get a better broker mate. Sheesh, you're GIVING money away!

 

[horace]

I know mate, I know. :-(

But... you know, this is the price for a hands up approach. I only see the accounts trade every once in a while. They take care of everything.

As I said I know it is expensive, but I have other accounts else where with the same approach and much cheaper and every week I need to send an email and ask for a rebate. :-(

Horace

14 bloody euros???!!!

Horace get a better broker mate. Sheesh, you're GIVING money away!

 

[meanreversion]

What's your style of trading, and how frequently do you trade?

 

[DashRiprock]

I know mate, I know. :-(

But... you know, this is the price for a hands up approach. I only see the accounts trade every once in a while. They take care of everything.

As I said I know it is expensive, but I have other accounts else where with the same approach and much cheaper and every week I need to send an email and ask for a rebate. :-(

Horace

it's a relatively high frequency strategy? if your standard trade is 10 (either euro's OR ticks), it does sound like the type of strat that needs either fully automating or to have a trader monitoring it.

Have you tried going down the automated route?

 

[horace]

Hi DashRiprock,

The AvTrade is +10€ (that is total net profit divided by number of trades). The strat is not HF, it trades once a day or even less in 120min time frame.

And, of course, it is fully automated since 2004 as I said before.

Thx for reading.

H

it's a relatively high frequency strategy? if your standard trade is 10 (either euro's OR ticks), it does sound like the type of strat that needs either fully automating or to have a trader monitoring it.

Have you tried going down the automated route?

 

[horace]

Hi meanreversion,

I trade fully automated trading systmes developed by myself. I trade almost once a day per market.

What's your style of trading, and how frequently do you trade?