Australian (ASX) Stock Market Forum

@Cam019
Are you looking at using Radge's system? I do recommend it. My super is small but growing quickly. This system is scalable to a large size which is one of the reasons why I like it. (scalable by being able to add more positions which helps the downside risk, and being monthly meaning positions sizes can be large-ish.)
 
Thanks @Warr87.

However, I am slightly concerned that you are quoting average backtest results over such a short backtest period using a monthly system. Not a very large sample size.
I had a crack at trading monthly a while ago. I gave up for the reason you highlight. Even over 25 years I just couldn’t generate enough trades in backtest to give me a high enough level of confidence.
 
I had a crack at trading monthly a while ago. I gave up for the reason you highlight. Even over 25 years I just couldn’t generate enough trades in backtest to give me a high enough level of confidence.
Very interesting @MovingAverage. Two questions if you don't mind answering.

1. What was the sample size for the parameters used on the monthly system?
2. What time frame did you go back to in order to generate enough signals to give you a sample size large enough to give you a high enough level of confidence? Weekly or daily?
 
Very interesting @MovingAverage. Two questions if you don't mind answering.

1. What was the sample size for the parameters used on the monthly system?
2. What time frame did you go back to in order to generate enough signals to give you a sample size large enough to give you a high enough level of confidence? Weekly or daily?

Hi @Cam019,

Not sure what you mean in relation to "sample size for the parameters". If you can elaborate on that I can better answer this question.

For me, when I look at sample size and statistical relevance I basically look at the number of actual trades taken in the context of the total number of available trades. Most systems will generate more entries than you can actually take due to constraints such as limited capital, risk management etc etc. So to me it is not really a question of being able to take X number of trades to decide if it is statistically relevant--the number of trades needs to be put into context. In other words what is the size of my sample (trades taken) in the context of the population size (total available trades). There are many different ways of varying complexity to measure statistical relevance but one crude method I used as a first pass measure is the ratio of the trades taken to the total number of available trades. The lower this ratio the less confidence I have in the sims. I do use other methods, but this is my initial measure.

I think it's been almost 2 years since I went down the monthly system path and I started documenting my paper trading of that system on here somewhere (if you do a search you'll probably find it). I can't recall specifics about sample sizes etc but if you're really interested I can re-run some sims for you--I'm in lockdown so nothing better to do at the moment:) What I do recall, however, is that I ran it over 25 years of historical ASX data and while the sim results looked good I had no statistical confidence in the results.

If memory serves me correctly, my monthly had reasonably long hold times and I've subsequently read several articles/research papers on monthly systems and a number of those have deliberately engineered shorter hold times so they could up the trade frequency and therefore have a higher level of confidence in their sims. I didn't want to go down that path as I wanted a system that wasn't so involved. Not sure if you follow Alvarez Quant Trading (I have no connection to him other than following his blog) I'm pretty certain he looked into this sometime ago.

Anyway, I hope this answers some of your questions.

MA
 
@Cam019 @MovingAverage

Thanks for the discussion. I don't see a lot of people discussing monthly trading as it's not as active or engaging like shorter term systems.

Sample size/trade frequency is definitely going to be an issue for any monthly system. Test it over long enough, and how much relevance does it hold? That is, what if I go back 30yrs. Is the relevance of how well it traded 30yrs ago relevant to how it trades now? It would be a very robust system if it was consistent all the way through, but I'm not sure if my confidence is boosted if I know it did well 30years ago .... as an example. But, as its been pointed out, we need sample size to help validate the system as statistically relevant given the results. Given the system I am running simply takes the top 'x' stocks with the highest momentum, the secret sauce is obviously how you measure momentum. in a lot of ways, this strategy is a lot like a high-beta strategy, though it's much more likely to catch young-up-and-comers than say an ETF running a high-beta strategy. I think the strategy works, and will continue to work, given that while it is a momentum strategy and is also a trend following strategy following the highest trending stocks. The monthly timeframes just means you can be on off with some timing (i.e. buy and then have a negative news event for that stock which immediately affects price).

If the issue is trying to weigh up what is statistically significant or to what extant is trade size sample enough, I think the issue of very long term historical significance may also play a part. E.g. my example of how relevant is the system from now to 30yrs ago?

I have faith in the system because its built on sound principles, simple enough to be robust, and still presents enough trades for me to think it will be close enough to expected results. This isn't an endorsement of Radge's turnkey systems btw, just my thoughts on running my particular iteration of a monthly momentum system. There's certainly other ways to run it (only ASX50 for more conservative, AllOrds for more risk/reward, more/less positions).
 
@Cam019 @MovingAverage

I have faith in the system because its built on sound principles, simple enough to be robust, and still presents enough trades for me to think it will be close enough to expected results.
And that is all that matters, it's your cash and you're the one trading it. After all, if we all traded the same way there would be no trading.
 
Test it over long enough, and how much relevance does it hold? That is, what if I go back 30yrs.
You now my view, what is the value of statistically significant sample if you base tomorrow trade on results acquired before big OS exposure, no super fund, no internet, aud vs usd from single to double, and inflation/interest rates from 16 to negative..
I know i differ with MA but 1000 such samples are as relevant as 12 based on last year. A very simple adjustment of test system yearly return vs inflation would change a lot
 
You now my view, what is the value of statistically significant sample if you base tomorrow trade on results acquired before big OS exposure, no super fund, no internet, aud vs usd from single to double, and inflation/interest rates from 16 to negative..
I know i differ with MA but 1000 such samples are as relevant as 12 based on last year. A very simple adjustment of test system yearly return vs inflation would change a lot
I'd be happy to tune my system on data from 1900 to 1920 for all I care, what will be the true test is how that system performs on forward out of sample data. If it consistently performs ok on different ranges of forward out of sample data then I'm good with that. Granted I may be able to squeeze better performance out of the system if I tuned it on more recent data, but my approach is to get consistent positive returns on a range of different out of sample forward data--not 100% optimized performance.
 
is your system more rotation based?

yup. that's exactly what it is. it's setup that way by Radge, and it's how it's meant to be implemented.

When I have the time, I'd like to approach an ETF strategy with similar rotational dynamics based on momentum. i believe this would work better on the US market, so could be another strategy for the future that is scalable and largely hands-off.
 
Month 12

Beginning of month 12! how exciting. 11 months in and I am still happy. Pretty good month. I decided to rebalance my positions, and there were a few change outs anyway. As you can see below, the momentum is still strong. I only had 1 position in a loss when I re-balanced! Not much more to add, but some stats.

Current return: 21.88%
Yearly return: 23.87% (remember this is the beginning of the 12month so have't done a full year yet)
Sharpe: 1.34


1627980468385.png
1627980474728.png
 
Month 13

End of the 12months! It ended strong. I also saw some numbers for the top super returns, and it's safe to say I did outperform them. Though, I have a small amount and they have a very big amount so not too surprising. My risk is higher too. They certainly wouldn't take the same kind of drawdown I am willing too, not while they handle clients money. But even with that qualification I am happy with my returns!

There are a few buys/sells set for tomorrow. One of my positions was removed (bought out I'm assuming). Metrics are strong as well (see below). The XKO has also done pretty well.

Current return: 32.66%
Yearly return: 32.66%
Sharpe: 1.22


1630403634956.png
1630403644932.png
 
Makes you think if there's any value in trading daily, weekly systems doesn't it?

It's like you read my mind, lol.

In a lot of ways monthly is pyschcologically harder to trade. Buying and then the market has a bad week—it's hard enough with weekly, but with monthly you have to wonder how long you will watch it go down. I recently saw a video from Radge on youtube that talked about trend following and momentum trading. This idea of momentum trading can obviously be used on a macro level. I have to wonder, how well would this trade on the US markets, and/or rotating in/out of strong markets. An idea to follow up later when I have time and capital. Radge listed a book for reading that I think will go over it (will add it to the list).

A lot of my returns in this portfolio have come from a couple of positions that gained 50-150%, while the others only slightly profitable. The bonus of this strategy is that you can hop on early into its upward momentum.
 
Have you noticed that Nick has modified his momentum system. He now allocates 50% monthly mom system, 50% weekly mom system. This was partly to reduce the ugly monthly drawdowns and to reduce the unfortunate timing issue ie starting just before the market dives (you know what this feels like, unfortunately). The weekly positions are closed before the EOM.

BTW It works great in the US. Although the US is in a huge bull trend and has been for ages.

Also, you'll be aware that the XAO has just had 11 up months for the first time in it's history. 12 up months in last 13. I say you got lucky this time.

Wise decision to include the extra 200 as these will show more growth than the top 100.
 
The value in trading shorter time frames comes from their ability to profit from shorter trends.

Say, the market has 8 up months and 4 down months. The monthly system would struggle depending on the sequence of the up/down months. A weekly system will do better in this situation as it captures smaller trends. A weekly system will struggle to beat a monthly system when all the months are up.

The same logic applies to a daily system. The disadvantage of the weekly/daily systems are that they require much more work to manage compared the a monthly system.

If you don't have the time or desire to manage daily/weekly systems, the monthly or quarterly timeframes are the best options.
 
Top