trendisyourfriend 

What about if you realise that patience as a discretionary trader is one of your problem among a myriad of other problems that need to be addressed properly and in a systematic manner? Does it not look like a problem or series of problems that need to be solved?

hi trendisyourfriend - to me what you describe is more akin to a skill to be learned.

I too had the patience problem and I have become better at it. Still working on improving it further.

grausch 

Interesting thread - I spent a significant amount of time developing automated strategies, but currently am a discretionary trader. Why the change? I just could not find systems that gave the type of returns I wanted with acceptable drawdowns. Does discretionary trading solve that problem? No, but at least it gives me a chance to reach my goals. That being said, I agree with @kevinkdog - most people developing mechanical strategies fail to put in the required research in order to fully develop robust systems. Regarding the title of the post (What about luck), a lot of traders seem to think that they have a certain amount of skill and that their results reflect that skill. I was monitoring a lot of hedge funds prior to the 2008 crash and several of the long-short funds had results that looked like they were buying the dips. It was a great strategy until the crash came and several of these funds found that their minimal short positions did not cover their downside sufficiently. Thus, to what do we attribute their good returns prior to 2008? Skill, a problem that was solved or luck?

This thread delves quite deeply into the relationship between luck and skill

xplorer 

It's good to see that some people disagree, I think it can generate an interesting debate. It would also be useful to have one or more of the pro-mechanical forum users discuss a specific mechanical strategy that they themselves have adopted that works. Do you guys have any?

I'm not sure what the debate is here, because I'm not sure what there is to debate.

I have over 80 mechanical strategies I am currently trading, although I've also tried thousands that did not work.

Here's one I developed back in 2011 for Soybeans. I trade this live right now with real money. Tradestation code is below the chart, with walkforward stuff taken out.





rrange=high-low;


if rrange>2*stddev(rrange,xr)+average(rrange,xr) and close>close[daysback] then buy next bar at market;
if rrange>2*stddev(rrange,xr)+average(rrange,xr) and close<close[daysback] then sellshort next bar at market;

if barssinceentry>=100 or openpositionprofit<-5000 then begin
sell next bar at market;
buytocover next bar at market;
end;

kevinkdog 

I'm not sure what the debate is here, because I'm not sure what there is to debate. I have over 80 mechanical strategies I am currently trading, although I've also tried thousands that did not work. Here's one I developed back in 2011 for Soybeans. I trade this live right now with real money. Tradestation code is below the chart, with walkforward stuff taken out. rrange=high-low; if rrange>2*stddev(rrange,xr)+average(rrange,xr) and close>close[daysback] then buy next bar at market; if rrange>2*stddev(rrange,xr)+average(rrange,xr) and close<close[daysback] then sellshort next bar at market; if barssinceentry>=100 or openpositionprofit<-5000 then begin sell next bar at market; buytocover next bar at market; end;

Interesting chart. Is this equity curve an 'actual' record of somebody's (e.g. yours) trading activity? Or is it the profit curve generated by the conditions above?

xplorer 

Interesting chart. Is this equity curve an 'actual' record of somebody's (e.g. yours) trading activity? Or is it the profit curve generated by the conditions above?

Neither. This is a hypothetical walkforward backtest with an enhanced version of the code I published above. To get this actual curve, the parameters XR and daysback change with time, based on walkforward testing - and that has to be put into the strategy. The curve then is the actual strategy, with all the walkforward parameters. I only provided the baseline rules for entry and exit.

So the curve is hypothetical, and contains only out-of-sample results. It is for 1 contract always, and includes appropriate slippage and commission.

I will say that my actual results since trading this live are very close to this hypothetical curve.


I will also say the skill involved in developing this strategy is really knowing how to do the walkforward optimization correctly, knowing when to stop development (ie not add too many rules or filters), knowing what types of orders can be backtested that can give the same performance in the real world, knowing what is acceptable performance in the first place, etc.

Thanks Kevin. As I understand, the chart you attached provides an example similar to what I described earlier and what I struggled with.

Here is a graph that shows hypothetical returns over a 3 months period of time for EuroStoxx 50. It is risk-adjusted.



The one below shows again hypothetical returns over 3 months for Cable. This one is not risk-adjusted but the returns are similarly highly favorable.



The two charts above - correct me if I'm wrong - would appear to show a very favourable return: for example if the Cable strategy is sustainable over time that would generate a return of over 40k a year or ~400k over 10 years.

Does the equity chart tell the whole story?

xplorer 

Thanks Kevin. As I understand, the chart you attached provides an example similar to what I described earlier and what I struggled with. The two charts above - correct me if I'm wrong - would appear to show a very favourable return: for example if the Cable strategy is sustainable over time that would generate a return of over 40k a year or ~400k over 10 years. Does the equity chart tell the whole story?

That's the problem - the equity curve by itself NEVER tells the whole story. To be meaningful, you really need to understand HOW an equity curve was derived.

For example:

With the curves you show, was that your first attempt, or was that the end result after trying 10 or 20 rule variations with entry and exit?

Was this the best output from optimization?

Is this in-sample, or out of sample?

Does it include proper commissions and slippage?

The list of questions goes on and on - I just touched on a few...


Hopefully, you see the point. Just looking at a curve, it is impossible to really say if it is a good strategy.

Of course, I realize the irony here, since I just posted an equity curve. Should you trust it? Probably not, unless you have a reason to trust me, and to feel comfortable with how I develop equity curves. I trust it, and that is why I trade it live.

Putting too much faith in pretty equity curves is one big reason why traders get fleeced by scumbag vendors. Just about anyone can produce a fanciful, enticing equity curve - even bankrupt losers living in Mom and Dad's basement. By itself that curve means nada.


But, for the sake of argument, let's say you developed your strategy a proper way (I use one way I feel is proper, and there are other ways too).

My big problem with your equity curves is that it is only a 3 month test. In my opinion, that is nowhere near enough time to evaluate a strategy, regardless of number of trades you took. What does that same system look like over 3 years, or better yet 5-10 years? If you can, plot the equity curve for that amount of history. That should be revealing.

kevinkdog 

That's the problem - the equity curve by itself NEVER tells the whole story. To be meaningful, you really need to understand HOW an equity curve was derived. For example: With the curves you show, was that your first attempt, or was that the end result after trying 10 or 20 rule variations with entry and exit? Was this the best output from optimization? Is this in-sample, or out of sample? Does it include proper commissions and slippage? The list of questions goes on and on - I just touched on a few... Hopefully, you see the point. Just looking at a curve, it is impossible to really say if it is a good strategy. Of course, I realize the irony here, since I just posted an equity curve. Should you trust it? Probably not, unless you have a reason to trust me, and to feel comfortable with how I develop equity curves. I trust it, and that is why I trade it live. Putting too much faith in pretty equity curves is one big reason why traders get fleeced by scumbag vendors. Just about anyone can produce a fanciful, enticing equity curve - even bankrupt losers living in Mom and Dad's basement. By itself that curve means nada. But, for the sake of argument, let's say you developed your strategy a proper way (I use one way I feel is proper, and there are other ways too). My big problem with your equity curves is that it is only a 3 month test. In my opinion, that is nowhere near enough time to evaluate a strategy, regardless of number of trades you took. What does that same system look like over 3 years, or better yet 5-10 years? If you can, plot the equity curve for that amount of history. That should be revealing.

Understood. Let me reveal what that 'strategy' I used was firstly. It was a really, really simple "when price crosses below lower-bound Bollinger band then buy at next bar". Reverse for sells.
The results were based on a predefined number of bars (after 10 bars in one case, after 5 in the other).

I see your point about longer periods of time. Unfortunately my platform keeps only the last 3 months worth of data, that's why I could not extend it, although I applied it to 8 different asset classes and they all showed the same promising, theoretical results.

I suppose - just to dumb it down to the simplest level - the question I am asking is, if a system has shown to be reliable (say profitable 50-60% of the time or even higher), why not load up with as many contracts as your margin allows and let it run, and retire?

xplorer 

I suppose - just to dumb it down to the simplest level - the question I am asking is, if a system has shown to be reliable (say profitable 50-60% of the time or even higher), why not load up with as many contracts as your margin allows and let it run, and retire?

Risk of ruin. You could have a very profitable strategy, but if you overtrade it, your chances of wiping your account out are huge - in fact, virtually guaranteed.

Let's take an extreme example: You and I play a coin toss game. But you bring a loaded coin, where it comes up heads 90% of the time. You know this, but I don't. You get to call the coin.

Let's say you have $1000 to bet with, with a $100 minimum bet. Do you bet all $1000 on the first flip? Or $900? You have a 90% chance of winning if you do, but 10% of the time you'll get wiped out. And you have a really biased coin!

So, you could bet less, but no matter what you bet, there is some chance that after a number of tosses, you'll get wiped out.


For trading, you'll never have such good odds. So, you have to bet as if a string of losers is around the corner. You want to stay in the game, to get the long term expectancy (profit). But, if you play the game you must endure pain in the short term (drawdown). The trick is to not let the pain be fatal...