Backtesting vs. Realtime results

Quote from kotika:

That very well may be true, but doesn't make it right ...
More...


there are reasonable arguments against sharpe ratio as such. yours is none among them. and though it could very well be that thousands of professionals are completely offTrack, when they use sharpe ratio as one of their tools to judge strategies - it is just unlikely that someone who has obviously not really heard about the concept as of lately has found out.

too little self confidence can be as dangerous as too much.

peace
 
Quote from Grob109:

... So why do the great traders not use computer sciences, commonly. Why is computer science not a necessary componet of trading well and with expertise. ...

... Who among those with high money velocity wealth building need things like sharpe ratios ...?

...So why don't most people in the profession of improving trading results have a set of tools for doing trading analysis ...?
More...


i think it is necessary to clarify that besides the universe of traders you are talking about, there seems to be another one as well, which apply a lot of computer "sciences" to trading. (you heard of rentec?)

in my opinion someone who does not intensively look at returns versus the risk at place, is in danger. question of time until he finds out that it makes huge difference whether to apply leverage 3 to a system with a sharpe of 1 or with a sharpe of 3. (i have a personal file of people who were sure they never had to look at their risk in detail - from top of my head i think no one survived financially).

most pros do use "trading analysis tools", including more sophisticated methods than has been talked about here.

i do respect that you seem to do well by applying your way of thinking. but for those new to the business it is necessary IMHO to put your comments into proper perspective.
 
Man are you saying only apply leverage to systems with high sharpe ratios?

The reason im wondering is I have a promising equity trading system with risk management in place , but it has a low sharpe ratio of .25 eventhough it returns 30% annually for 40years with no leverage. I wanted to bump it up to 40% annual with 50%margin.

What do you think?
 
hm. first i doubt that your sharpe is that low. in your case it would mean that your vola was 120%. you would not consider trading that i guess. that is simply too low. what i think is that you have the numbers flawed.

take your equity curve and load it into excel into column A. calc next to it your daily log returns. go into cell b3 and type "=ln(A1/A2)". now copy that cell down to the line which holds the last entry in column A. which is the last day of your equity curve. that was the part where you calculate your daily log returns.

now go into cell c2 and type "annualised return".
now go into cell c3 and type "annualised volatility".
now go into cell c4 and type "modified sharpe ratio".

now go into d3 and type "=average(B2:B20000)*252"
now go into d4 and type "=stdev(B2:B20000)*sqrt(252)"
now go into d5 and type "=d3/d4", this is your modified sharpe.

if the figure was below 1 i would reconsider trading it. if your system is trading single equities, be very, very, very careful that you do not fall into to the cruel trap of survivorship bias. if the system has a sharpe of 1, trades single equities and is long only, i would put it on paper trading with not too much hope. if it is indices, well, different thing.

strong advice: check, check and check. be sceptical about your results until your account grows. make sure you adjust for trading costs. be aware that trading costs have recently sunk a lot. if you test over forty years that might be an issue. put more emphasise on recent history, since competition in this kind of game has come up only during the last ten years. maybe your edge has already gone. look at logarithmic charts only - everything else flaws your eye.

best luck to you.

(do not get angry when the annualised return figure is below your percentage return. this is because of the logs used. their advantage is that they "know" that a trading curve of 100, 50 and 100 again has an average return of 0%, while perecentage thinking believes average return is 25%.)

peace
 
man, Im confused. I attached what I have. I dont have the equity curve, but I have cumulated profit. Is there a way to use the cumulated profit, though its listed per trade, not per day?

Thanks,

Eric
 

Attachments

hm. not really. you definitely miss mark to market while you are in a position. thus your profits or losses seem to occur on one day, namely when you are closing out.

it looks like a single stock system, being long only.
you are probably heavily exposed to survivorship bias. that means that all the stocks you bought after your pattern occured, when bust consequently are ignored. the simple fact that a stock is currently alive is your biggest bet. the oither thing is that you are buying very cheap stocks as well. my rule of thum is not to buy a stock that trades below 7, but that is personal choice.

flip side is that your approach handles the bear market very well. might be something to it. i would think the strategy can make money but not to the extent you would wish. but who am i to tell?

peace
 
Quote from man:

... that means that all the stocks you bought after your pattern occured, when bust consequently are ignored ...

peace
More...


sorry. mistake in this sentence. shall read " ... that means that all the stocks you bought after your pattern occured, WENT bust consequently are ignored ..." .
 
After a quick look it seems that your per trade ratio is 0.23, that is very good, in fact so suspiciously good ...

But it is very difficult to use this number for a few reasons,

1) there are overlapping trades, in other words the program is not waiting to get out of a trade to initiate more trades. That greatly complicates the issue of position sizing and may in fact get you killed if using margin. Another way to say the same thing: bad luck may spread itself for a few months over many positions.

2) there are many stocks whose prices are under a dollar, for some of them may be because of splits that occured since then. But there seem to be also some genuine penny stocks. I would exclude those because you wouldnt have been able to buy (esp. sell) them at nearly the price assumed in the backtesting program.

(By the time i write this others seemed to point out the same things already.)

Bottom line, this sems like a <b>hypothetical</b> strategy which is not in fact implementable exactly as tested.

Hope i'm wrong... and good luck,
K

PS The strategy i have experience myself has bad luck streaks of no more than a week. In real results it has so far behaved as expected. I believe this is because i have carefully analysed statistics myself, and not trust it to some computer program without knowing how it works.
 
You must log in or register to reply here.
Back
Top