Technical Trading Exercise (Pavilion103 and tech/a) Discussion

[tech/a]

I'd call it performance review/improvement rather than hindsight evaluation.

Ok

You mentioned some statistics in the first period of you and Pav's trading... and they were quite different to what has transpired in this exercise. So it seems reasonable to wonder why - and the reason could be the market, trade management, stock selection, general randomness or some combination of the above.

SKC I'm sure you have been involved in Montecarlo simulations as part of your analysis.
But for those who haven't --- results of say 5000 possible portfolios will give a range of results ranging from the mean to low to high ranges.

This method has many many trades it COULD take but not enough capital to take them all and as PAV has shown the screening itself can also be different where I miss prospects and so can he.

The point is that the first exercise could well be at the very top of the possible returns using this method.
It certainly had many many more selections take off right from the get go.

This method hasn't been tested at all so I have no figures to work with other than the walk forward trading.
I have no doubt that over time there will be excellent profit and the equity curve is likely to be showing spikes up with flattish periods.

So it is highly likely that this portfolio is currently trading at the low end of possible results. Those who have worked with Monte carlo analysis like yourself---will know what I mean.

I was just hoping to get you to show in this thread how you undertake performance review / continuous learning. But if that is not the main purpose of this exercise that's understandable.

Well it wasn't but I welcome it as part of the process!!

I think every process posted here (on ASF) should be open to every sort of scrutiny available.
Otherwise there is no learning!

 

[craft]

SKC I'm sure you have been involved in Montecarlo simulations as part of your analysis.
But for those who haven't --- results of say 5000 possible portfolios will give a range of results ranging from the mean to low to high ranges.

Monte Carlo simulation does not model serial correlations. This means the numbers coming out in each draw are random; there is no way to control what comes out in the next draw based on what was just drawn. But the market is serially correlated – especially if you are playing it based on momentum. What happens today does influence tomorrow – and what happens to one influences others – we know this instinctively as bull and bear markets.

Using Monte Carlo which infers true randomness (ie coin toss randomness) on a serially correlated data set will underestimate the possible range.

Sure we have been through this – that's right, I don’t understand because I know FA.

The questions I would be asking myself - Has the system got an edge or will it only work in a raging bull market when everything 'long momentum' does? Is the position sizing correct to see me through to the next bull market? or am I relying on a Monte Carlo simulation of drawdown that may be understated because MC doesn't take into account serial correlation? Am I using a market trend filter - Is that the primary importance and individual stock selection secondary to my expectancy?

 

[skc]

SKC I'm sure you have been involved in Montecarlo simulations as part of your analysis.
But for those who haven't --- results of say 5000 possible portfolios will give a range of results ranging from the mean to low to high ranges.

So it is highly likely that this portfolio is currently trading at the low end of possible results. Those who have worked with Monte carlo analysis like yourself---will know what I mean.

Actually I've never worked with Monte Carlo simulations although I am aware of what it is. I think a relatively small set of results (i.e. only 30-40 trades) will no doubt experience plenty of randomness and easily swung by a single trade here and there anyway.

Monte Carlo simulation does not model serial correlations. This means the numbers coming out in each draw are random; there is no way to control what comes out in the next draw based on what was just drawn. But the market is serially correlated – especially if you are playing it based on momentum. What happens today does influence tomorrow – and what happens to one influences others – we know this instinctively as bull and bear markets.

Using Monte Carlo which infers true randomness (ie coin toss randomness) on a serially correlated data set will underestimate the possible range.

Very insightful. Totally logical yet not something I've read before.

The questions I would be asking myself - Has the system got an edge or will it only work in a raging bull market when everything 'long momentum' does? Is the position sizing correct to see me through to the next bull market?

I think the trading approach here has plenty of merits especially for a part time trader who's looking to put their capital to work, limit their downside while being exposed to the upside. If I can get my Mum (who's been in the market for 20 years yet doesn't even know whether she's ahead or behind) to trade like this it'd be a huge achievement already. It's perfectly OK (and probably preferrable) to make few trades and make no profit for months when the market isn't treating you well, but with an opportunity to make decent returns when conditions become favourable. I raised the question about performance as the conditions do appear quite favourable at the moment, but I also acknowledge that the results so far may simply be skewed by a bit of randomness.

On the other hand, if you are relying on trading as the primary source of income then you'd want something a little bit robust with more regular and higher levels of activity. Being a full time trader, I often find myself judging approaches with different prupose a bit unfairly at times.

 

[tech/a]

Monte Carlo simulation does not model serial correlations. This means the numbers coming out in each draw are random; there is no way to control what comes out in the next draw based on what was just drawn. But the market is serially correlated – especially if you are playing it based on momentum. What happens today does influence tomorrow – and what happens to one influences others – we know this instinctively as bull and bear markets.

You could be correct I'm only Hypothesising.

Sure we have been through this – that's right, I don’t understand because I know FA.

Are you incapable of refraining from inflaming a discussion?

The questions I would be asking myself - Has the system got an edge or will it only work in a raging bull market when everything 'long momentum' does? Is the position sizing correct to see me through to the next bull market? or am I relying on a Monte Carlo simulation of drawdown that may be understated because MC doesn't take into account serial correlation? Am I using a market trend filter - Is that the primary importance and individual stock selection secondary to my expectancy?

Yes all valid questions and if I were designing a method to trade would ask and test.
This is however AN EXERCISE.

 

[craft]

It's perfectly OK (and probably preferrable) to make few trades and make no profit for months when the market isn't treating you well, but with an opportunity to make decent returns when conditions become favourable. I raised the question about performance as the conditions do appear quite favourable at the moment.

This is why I have asked the questions I have.

If you are going to make your B/E some sort of arbitrary exit for momentum lagging then you are going to have to have a very strong and low volatile market for it to work.

The small cuts in-between time could be very agonising if you enter markets that are not strong enough. MC with its limitations is not necessarily going to paint the true draw down picture.

If you look at the market when things worked it was a fairly rare event with the overall market rising at 70%+ annualised for a few month period with very small ranges. The current up leg is half that speed with larger ATR’s.

If the B/E stop is non negotiable perhaps the market filter need to only switch on in explosive strong markets.

Perhaps consideration could be given to smaller parcel sizes so that pushing quickly to B/E is not psychologically so imperative and retests of breakouts (normal action for a normal up leg) can be allowed to occur without discomfort. Maybe the AW/AL comes down a bit but your W% is likely to more than compensate and your opportunity defiantly goes up because you are not scratching so many trades.

Are you incapable of refraining from inflaming a discussion?.

Probably -Sorry - I'll try harder. [But you gotta admit it takes one to know one]

Yes all valid questions and if I were designing a method to trade would ask and test.
This is however AN EXERCISE.

Ok

I think every process posted here (on ASF) should be open to every sort of scrutiny available.
Otherwise there is no learning!

??

 

[peter2]

...
If the B/E stop is non negotiable perhaps the market filter need to only switch on in explosive strong markets.
...

Or improve your odds (W%) by screening for trading candidates with a strong trend (ADX >30) with low volatility.

Or the top down FA approach, trade momentum in only the strong stocks in the strongest sector.

 

[burglar]

Or improve your odds (W%) by screening for trading candidates with a strong trend (ADX >30) with low volatility.

Or the top down FA approach, trade momentum in only the strong stocks in the strongest sector.

Is there a nice way to say this?
Clearly you are changing the way tech/a trades, to something you are more comfortable with.

 

[fiftyeight]

I hope you continue and I hope some of the less experienced traders who showed their enthusiasm at the start of this thread reappear and contribute more than they have during the thread. This exercise is a valuable real-time learning experience for the forum community and I hope it gathers enough support to rekindle the OP's passion for sharing.

I can only speak for myself as an inexperienced trader but I dont really have a lot to contribute. I learn a lot more when experience traders are discussing an issue as they are at the moment.

Cheers all involved

 

[Gringotts Bank]

Is there a nice way to say this?
Clearly you are changing the way tech/a trades, to something you are more comfortable with.

He is but it makes sense if the problem has been identified as low winning %.

Tech and pavillion may not be interested in my advice either, but here's one possible approach:

Identify a high trending, low volatility stock using ATR, Chaikin, ADX, ROC, long term trendline. My preference would be trendline + Chaikin.
Identify a flush-out bar (much higher volatility red candle with long lower wick).
Enter as soon as the volatility returns to pre-flush out levels.
Ensure the long term trendline is still unbroken.
Backtest the idea for tweaking.

The thing with high momentum low volatility stocks is that when they correct, they correct HARD and often in just one or two bars.

 

[pavilion103]

Happy to hear your thoughts GB when you're not trolling.

 

[tech/a]

Personally I'd like to see a selection of ideas posted here put together in a trading
Method and traded in tandem.

The results would be good to reflect on in a few months.

 

[pavilion103]

Have been watching AGO in recent weeks. Tonight's prospect.

By the way. I'll update the spreadsheet this weekend and then I'll be in India for 2 weeks so it will be a bit of a gap between updates.

 

[pavilion103]

I have posted some analysis in the main thread.

When reading it please remember that

THIS IS THE THREAD TO POST COMMENTS IN

not the other one.

Thanks

 

[Country Lad]

Personally I'd like to see a selection of ideas posted here put together in a trading
Method and traded in tandem.

The results would be good to reflect on in a few months.

Tech, do you mean comments on the ones mentioned here like for example AGO.....

Have been watching AGO in recent weeks. Tonight's prospect.

.......and if so, broke from my setup 7 trading days ago when it went to 110 for a buy for me at 112 that day. I have a stop at 113, the previous high before the break.

Not going on with it strongly, but going on with it. I didn't want to comment like this previously because it may appear to be hijacking this thread.

This the sort of thing you had in mind?

Cheers
Country Lad

 

[tech/a]

Sort of.

There were/are all sorts of ideas to improve the basic method.
I'm suggesting a seperate thread with perhaps a method based upon the ideas mentioned in the thread as improvements---traded in tandem-------trade it however you want.

We should be able to see which improves and how the various aspects affect the method.
Do a discretionary systematic approach.

I'm introducing one of my own conditions I use when selecting a trade into the exercise.
I used it with DNA.
I find it powerful and very useful.

 

[kid hustlr]

Really strong results over the calendar year boys, nice to see the equity curve turning around lately as well.

Can't believe how well you picked the market top back in the middle of the year. Was turning to cash at this stage planned or did you just not receive any set ups at that time?

Should just short the future when the system is turned off!! Reminds me of a thread TH did with regards to hedging an equity portfolio a while back.

 

[Gringotts Bank]

This is the idea I posted above, backtested. Basically you buy the dips of strong stocks. Not optimized yet so plenty of room for improvement.

Strong trend = 280+ days of the last 300 are above MA (low lag MA based on the lows with a high period)
Flush out bar = cross below this MA in the last 10 days >1.
Return to normal volatility = ATR(1)<.05
Buy a $5000 lot every bar that is below the MA when the above conditions are met (price averaging the entry)
Sell when the price crosses above the same MA of the closes, or with a 10% stop loss.

The % return is low, but the CAR/MDD is ok.

 

[tech/a]

This is the idea I posted above, backtested. Basically you buy the dips of strong stocks. Not optimized yet so plenty of room for improvement.

Strong trend = 280+ days of the last 300 are above MA (low lag MA based on the lows with a high period)
Flush out bar = cross below this MA in the last 10 days >1.
Return to normal volatility = ATR(1)<.05
Buy a $5000 lot every bar that is below the MA when the above conditions are met (price averaging the entry)
Sell when the price crosses above the same MA of the closes, or with a 10% stop loss.

The % return is low, but the CAR/MDD is ok.

Excellent
Now trade it live.

 

[pavilion103]

Kid - setups did stop appearing at that time. But this simply confirmed the analysis of the XAO and other indexes at the time.

Tech, you seem to be the only one putting it on the line and trading this live. I respect input from all but am disappointed that none of the others are putting their balls on the line live!

 

[CanOz]

Excellent
Now trade it live.

First of all I would have though you might say "test with more OOS" data, then you might say "walk it forward"...

Why would he test that live?

GB, was this system designed to do the same sort of thing Tech and Pav are trying to do? Is that why you posted it here?