Breakout systems

[MichaelD]

Michael,

Are you trading Turtlesoup ??

No, but the thought process behind Turtle Soup - figure out what other traders are doing and work out how to trade against them - is robust and the way I predominantly think these days.

However I'd argue that certain breakout entries provide an edge because of the risk/reward r'ship they present. Essentially, the entry may as well be random but the pattern provides a risk/reward 'filter'.

Not neccesarily the 'best' R:R but the most robust protective stop which minimises the chances of getting 'whipped' out of the position, while also minimising the loss if the pattern does fail.

I'm just not sure if you could effectively program these types of entries.

I tend to agree with you here, and also acknowledge other points in this thread as to exactly what a breakout/breakdown is.

There are significant limitations in what you can code and thus mechanically backtest for an entry. I have to this date never seen mechanical code for a breakout which enhances system performance in a significant manner.

This limits any conclusions you can draw about the effectiveness of such entries as the evidence is mostly anecdotal and in the form of "look at this particular chart" - it is not particularly helpful to look at when the pattern works out. Rather, you want to know what happens when the pattern doesn't work out and whether the R:R is worth it overall.

One particular pattern that has caught my eye of late is a breakdown through a double bottom - a nice example has been posted in this thread. Coding a breakdown through a double bottom to test the hypothesis...now THAT'S hard.

 

[It's Snake Pliskin]

One particular pattern that has caught my eye of late is a breakdown through a double bottom - a nice example has been posted in this thread. Coding a breakdown through a double bottom to test the hypothesis...now THAT'S hard.

Herein lies the problem of thinking purely in testing to give the answers.

 

[tech/a]

I have to this date never seen mechanical code for a breakout which enhances system performance in a significant manner.

Techtrader?

Or is it the exit that enhances the system?
Or the Stop?

If I removed the entry and replaced it with Random I'd be no worse off ?

 

[theasxgorilla]

Therefore, if one wanted to see whether a system was worth trading, won;t it be a simplier exercise just to compare its performance to the index? I just don't see what random entry/exit offers above comparing to index performance.

I think that because most system traders trade in a non-market-cap-weighted fashion that any comparison measuring whether a system is an improvement on something else should try to keep as many money management factors consistent as possible.

Take one of the rawest forms of a random entry/exit test; one that not only randomises when it buys and sells, but also how much to invest per trade (up to say, 50% of total capital). In one random iteration it might continuously invest 50% x 2 in small-cap miners and make a killing. In another it might choose TLS and some other struggling stalwart and underperform.

Then add fixed position sizing (eg. 10% per trade) and test again. You are now exposing your capital to the market in a non-random, non-market-cap-weighted fashion. My testing showed that this bunches and shifts the distribution of CAGR and Max DD in favourable directions. Just doing this should help you beat the index.

If you then add some purposeful entry, or exit conditions it is that study that you should compare your results with to see if you are improving consistency and effectiveness.

 

[MichaelD]

Techtrader?

Or is it the exit that enhances the system?
Or the Stop?

If I removed the entry and replaced it with Random I'd be no worse off ?

If you replace the entry with random you'd be marginally worse off. Most of the filtering benefit of the T/T entry comes from selecting for stocks above a medium term moving average - that's robust and that works.

You'd find, though, if you went back and redid the T/T creation process that most of the entry filters wouldn't work anywhere near as effectively as they used to - most of the T/T entry is merely curve fitting to the data of the day.

But that's not the point. That's time-wasting on whether a 30 EMA is better than a 15 EMA and other such irrelevant details.

What works is the consistent following of the rules to limit the losses and let the winners run. Do that, and long term trend following will work no matter what the entry contributes or doesn't contribute.

T/T is profitable DESPITE all efforts to corrupt it by overanalysis because it has an exit which skews the R:R in the correct direction.

 

[theasxgorilla]

Techtrader?

Or is it the exit that enhances the system?
Or the Stop?

If I removed the entry and replaced it with Random I'd be no worse off ?

Actually, TechTrader is difficult to 'admit as evidence' because the entry isn't mechanical. If it were the equity curve would be very different to the publicly traded model.

One of the best aspects to the system is the manual override that has allows cherry-picking of entries.

 

[theasxgorilla]

But that's not the point. That's time-wasting on whether a 30 EMA is better than a 15 EMA and other such irrelevant details.

What works is the consistent following of the rules to limit the losses and let the winners run. Do that, and long term trend following will work no matter what the entry contributes or doesn't contribute.

It's not the specific point of this thread but you are touching on a very important point IMO. We have to be able to follow 'the rules', as you put it. This is a given and it's 100% psychological. I don't think I could psychologically follow a system where I haven't spent some time coming to the conclusion that this parameter value suits me because of this reason, and that one because of that etc. I don't think it's a waste of time at all to convince yourself that your system has been purposefully designed and to understand when it will work well and it's shortcomings.

So long as people realise that they need to take this into account I think they stand a better chance of eventually trading something that they will be able to follow.

 

[tech/a]

Actually, TechTrader is difficult to 'admit as evidence' because the entry isn't mechanical. If it were the equity curve would be very different to the publicly traded model.

One of the best aspects to the system is the manual override that has allows cherry-picking of entries.

Hahaha I gotta love this.
I'm that good I can cherry pick those trades which skew results----yeh right!

Thats the whole point of Montecarlo testing.
After testing 50000 combinations,
I know that no matter what trades I take that I'll have results fall within the parameters returned in the testing.
So far thats exactly the case.

Mind you I agree with Michael that IN THE END entry is irrelevant on longterm trades.
But to get it going.

 

[theasxgorilla]

Thats the whole point of Montecarlo testing.
After testing 50000 combinations,
I know that no matter what trades I take that I'll have results fall within the parameters returned in the testing.
So far thats exactly the case.

I thought we came to the conclusion that Monte Carlo testing only 'forked in the road' when there were many trades being presented and limited capital. Using a discretionary override to choose whether or not to take a trade would be better compared with randomly ignoring a proportion of buy signals. It's two different things from where I'm sitting.

 

[tech/a]

Using a discretionary override to choose whether or not to take a trade would be better compared with randomly ignoring a proportion of buy signals. It's two different things from where I'm sitting.

Not at all. ASX.Thats EXACTLY whats happening.

You'll note that T/T isnt fully filled.
There is only a partial portfolio.

Thats more due to my slackness than design.
There are trades presented every night.
I dont run a scan more than once or twice a week.
So 100s go by without ever being viewed!
Mainly due to not trading it myself as you know.Trading discretionary shorter term.

Im blown away that it can still out perform the market in less than good trading conditions and less than Ideal trading.

 

[nizar]

One of the best aspects to the system is the manual override that has allows cherry-picking of entries.

One of the best aspects?

We have absolutely no way of knowing whether the reason behind the outperformance of the public T/T is due to tech/a's cherry picking or whether in fact he was just lucky.

For me, any discretion that leads to better performance would be one of the WORST aspects of the system because you can never know if this added to the edge of the system and consequently whether or not to keep using this discretion.

The very best systems require no discretion at all, in my opinion.

 

[wayneL]

One of the best aspects?

We have absolutely no way of knowing whether the reason behind the outperformance of the public T/T is due to tech/a's cherry picking or whether in fact he was just lucky.

For me, any discretion that leads to better performance would be one of the WORST aspects of the system because you can never know if this added to the edge of the system and consequently whether or not to keep using this discretion.

The very best systems require no discretion at all, in my opinion.

Most breakout systems will have a preponderance of signals, far far more than capital, or the system is designed to trade. Therefore there MUST be discretion on which signals to actually trade.

Like any rule based t/a system, the success depends on the trades NOT taken.

You cam only be truly mechanical on a single, or limited number of tickers (such as is the case with The Turtles).

Show me one trader that, when faced with twenty stocks in their scan, won't try to discretionarily(?) pick the best prospect(s).

 

[theasxgorilla]

Im blown away that it can still out perform the market in less than good trading conditions and less than Ideal trading.

I suspect that has more to do with not understanding WHY you system is working : As tongue in cheek as that might have sounded, try starting up TechTrader from the 1/1/07, using today's ASX300, leverage and other rules kept the same, no discretionary filter...then see if you can explain the divergence in results between the 'slack' version and the 'tested' one.

 

[stevo]

Most breakout systems will have a preponderance of signals, far far more than capital, or the system is designed to trade. Therefore there MUST be discretion on which signals to actually trade.

Like any rule based t/a system, the success depends on the trades NOT taken.

You cam only be truly mechanical on a single, or limited number of tickers (such as is the case with The Turtles).

Show me one trader that, when faced with twenty stocks in their scan, won't try to discretionarily(?) pick the best prospect(s).

A mechanical system can rank stock selections using objective rules and the trader executes the trades in ranking order. This is used on 500 plus possible stock tickers in mechanical systems being traded today.

It's arguable if it's truly mechanical even then since a lot can happen between getting the signal and actually getting filled (or not). It's probably the trader that has to be "mechanical" as well!

 

[MichaelD]

Like any rule based t/a system, the success depends on the trades NOT taken.

Show me one trader that, when faced with twenty stocks in their scan, won't try to discretionarily(?) pick the best prospect(s).

I'm not sure I understand what you mean by the first sentence, but on the second point - it interests me that when I do have a choice of trades, I am now tending to pick the worst looking charts rather than the "sure bet" charts.

 

[theasxgorilla]

One of the best aspects?

We have absolutely no way of knowing whether the reason behind the outperformance of the public T/T is due to tech/a's cherry picking or whether in fact he was just lucky.

For me, any discretion that leads to better performance would be one of the WORST aspects of the system because you can never know if this added to the edge of the system and consequently whether or not to keep using this discretion.

The very best systems require no discretion at all, in my opinion.

Yep, fair call. Although I suggest you compare the results of TechTrader using various recent start dates (within the last couple of years) with the public results of the system as traded with discretion. You will clearly see shortcomings of the system if left to it's own devices, and the importance of employing discretion to prevent it from digging itself down into rather perilous drawdown situations.

As for testing the impact of discretion on results, of course that is difficult. You can't Monte Carlo that, so it's not feasible to compare the distribution of two sets of results. And as you point out, to compare the position of a single traded path within the distribution of Monte Carlo'd results from testing without discretion doesn't answer the questions: was it luck? can it be consistently repeated?

IMO, with TechTrader the discretion and/or slackness has inadvertently done what some kind of switch like an index filter might have done mechanically. The current mechanical entry is not sufficient to keep the system out of an inappropriate market.

 

[tech/a]

Thought Id run that for interest.
Not a lot of difference.
From one simulation and the same capital as the Traded one $392,000
loss was around $20k over that period.
In realtime its a profit of $25,000.
But if we look at 20000 simulations its a little above average.
Dont think it proves anything.

 

[nizar]

Most breakout systems will have a preponderance of signals, far far more than capital, or the system is designed to trade. Therefore there MUST be discretion on which signals to actually trade.

Like any rule based t/a system, the success depends on the trades NOT taken.

You cam only be truly mechanical on a single, or limited number of tickers (such as is the case with The Turtles).

Show me one trader that, when faced with twenty stocks in their scan, won't try to discretionarily(?) pick the best prospect(s).

Wayne.

I dont disagree with what you are saying. My point was that the use of discretion in a mechanical system is a weakness.

My system often has more stocks triggering buys that what i can afford -- But I wish it didn't!

With this in mind, the next system I will design will be structured in such a way so that I can take every trade that triggers (only 2-3 pairs).

As for your last comment, ideally you shouldn't have to look at charts/fundamentals to decide which stock to buy, though I still do.

Rank the candidates according to price or ROC or something that has been backtested and can be traded systematically.

Michael.

I can relate to your comments 100%.
I thought LGL would go to the moon. REX looked very promising as well. And then there was BKN....

 

[tech/a]

IMO, with TechTrader the discretion and/or slackness has inadvertently done what some kind of switch like an index filter might have done mechanically. The current mechanical entry is not sufficient to keep the system out of an inappropriate market.

Whilst not entirely agreeing there is some merit in what you say.

 

[weird]

I agree with Stevo concerning the importance of entry. I would use a fixed bar length for the exit though to compare, instead of random exit to test the effectiveness of entries.

A few short-term future systems, pretty much rely on entry alone, including the stop (perhaps used more for account continued survival, which could also eventually in some circumstances cause the trader to give up in frustration if the stop is not dynamic or the risk perceived is no longer comfortable), with the exit being fixed at the close of the day.

I know of systems which are based purely on entry, and using a fixed exit, and if ignoring a tight stop the system may even perform fantastic during the most volatile periods (although nicely accepted in hindsight), however if actually employing a tight initial stop during those type of periods, it would most likely have caused frustration in following such a system. An adaptive stop may not always suit the risk profile of the trader. The very tight stop can sometimes cause frustration.

The only benefit of the random entry theory I can see, is helping a trader perhaps focus more on the rest of the system, which predominately looks at trade management.

But dismissing the importance of entry (which could also be used to define where to place the initial stop for determining a low risk entry (leading into the important area of trade management) is not practical in my opinion.