ASX Momentum System Testing

This topic is empty.

Viewing 15 posts - 1 through 15 (of 26 total)

1 2 →

Author

Posts
September 10, 2018 at 10:45 am #101754

RobGiles
Member

As I said in the monthly hookup, I’m in the process of testing a dual momentum system on the ASX100. So far I’ve run the following testing & optimizing:

Initial Testing Period: 1/6/98 to 1/6/07 – this window was chosen off the back of Nick’s suggestion to include a) a longer period than 5 years and b) a period that had a mix of bullish & bearish periods:
Optimized all variables yielded the following metrics:
CAR 22.34%
Max System DD -21.5%
Ttl Return 514%
Ttl Index Return 232%
MAR Ratio 1.04
Profit Factor 4.29

The next step was to use the following Out Of Sample data to see how it held up in a period where the $XTO essentially went nowhere…
OOS test period: 1/6/07 to 1/6/18 using same variable values as IS test yielded the following metrics:
CAR 2.59
Max System DD -27.0%
Ttl Return 32.5%
Ttl Index Return 0.02%
MAR Ratio 0.09
Profit Factor 1.23

So from here, I guess I was interested in people’s thoughts on, a) should I re-optimize the system on the past X (5??) years to ‘tune the system to current market conditions’, and b) are the OOS results given above acceptable, given the index essentially went nowhere and had a 53% drawdown but the system at least gave positive, albeit very small CAR?

FWIW the metrics for the entire period from 1/6/98 to 1/6/2018 using the original IS variable values are:
CAR 12.35%
Max System DD -24.8%
Ttl Return 928%
Ttl Index Return 233%
MAR Ratio 0.50
Profit Factor 1.90

September 10, 2018 at 11:11 pm #109148

JulianCohen
Participant

I tested my ASX momentum system over the same periods as you.

I run it on the whole ASX and exclude resources.

on ASX100 excluding rescources it gave
1998-2007: 24% and -31%
2007-2018: 15% and -23%
Whole period: 19% and -31%

On my universe it gave:
1998-2007: 22% and -38%
2007-2018: 23% and -24%
Whole period: 23% and -38%

And just for good measure on ASX 100 including resources it gave:

1998-2007: 44% and -41%
2007-2018: 14% and -27%
Whole period: 27% and -41%

Hope that helps Rob?

September 11, 2018 at 6:03 am #109152

RobGiles
Member

Thanks Julian….it helped make me very envious of your metrics ….no seriously it gives me a good benchmark to aim for.

Assume that the values for your system variables were left unchanged for each backtest in each period? Also, is the max DD about as much as you can stand psychologically?

September 11, 2018 at 7:37 am #109153

RobGiles
Member

forgot to add, I had no liquidity filters in my ASX100 system (assumed top 100 Co’s would be fine). Assume you’d need them if testing on the whole ASX ($XAO)?

Weird…my results were opposite to yours. When I test the system (with no liquidity filters) on the All Ords I get:
1998 – 2007: 36% and -29% inc. resources
1998 – 2007: 24% and -19% excl. resources

2007 – 2018: 10% and -30% inc. resources

So back to the original question….should one re-optimize on the most recent 5 – 10 years? Anyone got a view on this?

September 11, 2018 at 12:37 pm #109154

JulianCohen
Participant

Well my system was relatively consistent across all three periods, which I think is a plus. I have found that 25-30% for a trend following system is about my tap out point. I wouldn’t have been comfortable when my system was at 38% but it wasn’t there for long and bounced back quickly, so by the time I would have been ready to pull the pin it would have recovered.

You have to think about how you would be feeling when your system is in the middle of the poor performance period. Would you think it was broken after three years and it had made cock all and was in drawdown still? No matter who you are it’s hard not to think that, and chances are high you would drop the system around that point. Don’t forget that for the period you were making 2% CAGR I would have been posting returns of 20 odd percent in the forum. Who wouldn’t think their system was broken at some point in those ten years?

If you reckon you would be able to trade through that period of underperformance then no problem with the system. I know myself and I wouldn’t so I need my systems to show me they are working to keep my confidence level in them high.

September 12, 2018 at 12:04 am #109156

RobGiles
Member

Julian Cohen wrote:

If you reckon you would be able to trade through that period of underperformance then no problem with the system. I know myself and I wouldn’t so I need my systems to show me they are working to keep my confidence level in them high.

You make a good point here, and to be honest I would probably feel the same way. Not sure if you’re willing to share your momentum measuring method, but mine was ROC in this case.

September 12, 2018 at 4:54 am #109157

JulianCohen
Participant

I’m using ROC too.

I am using the whole XAO.au without resources, with a price limit and a volume limit. There’s a couple of other bits and pieces I have picked up over the last two years in there too, but you should be able to get something usable with ROC. Gap filters and Nick’s method of holding the stock if it stays in the top twenty are in there too.

September 30, 2018 at 12:17 am #109155

ScottMcNab
Participant

Rob Giles wrote:

forgot to add, I had no liquidity filters in my ASX100 system (assumed top 100 Co’s would be fine). Assume you’d need them if testing on the whole ASX ($XAO)?

I think it is worth adding some liquidity filters in there Rob..makes a difference even with XTO (for my systems anyway)

October 4, 2018 at 2:44 am #109231

RobGiles
Member

Scott McNab wrote:

Rob Giles wrote:

forgot to add, I had no liquidity filters in my ASX100 system (assumed top 100 Co’s would be fine). Assume you’d need them if testing on the whole ASX ($XAO)?

I think it is worth adding some liquidity filters in there Rob..makes a difference even with XTO (for my systems anyway)

Thanks Scott
I’ve tested Min Share Price $1, max $500 (i.e. don’t care about max) and Ave Volume 500K and lots of variations around these parameter values, and it just kills the system. I don’t how you guys do it but I can’t get mine to work with liquidity filters. Problem is I know I need them as it doesn’t reflect reality in terms of what I’d be able to get away in the market on very illiquid stocks.

October 5, 2018 at 12:07 pm #109240

WENKIT LUI
Participant

Hi Rob

Your topic caught my eye as I have just developed an ASX momentum system on the All Ordinaries universe.

For 1/1/2007 to 31/12/2017, the backtest results are:
CAGR: 21.7%
Max System Drawdown: 22.16%
Profit Factor: 2.48
Payoff Ratio: 2.37
Win %: 51.09 %

As with all systems, the start date affects the metrics. From memory, I think the worst max. system drawdown was 25-26%, which is within my tolerance. In my case, I found that liquidity filters do enhance my system performance. I use Min Share Price $0.5, Max $100, average volume 300k (volume EMA 50 days)

How I measure momentum is based on Nick’s article below.
https://www.thechartist.com.au/Shares-Stocks/52-week-high-rotation-strategy.html

However, I am measuring the market for stocks within 5% of their 52-week highs, instead of 10%. Maximum 5 positions and hold the stock if it stays in the top 15.

It’s taken me a while to test a myriad of different filters, but I was finally happy with the system’s performance metrics and recently started trading the system as of 1 October 2018. Time will tell if it’s a good system and whether I can continue trading it psychologically.

Hope this helps and good luck!

October 7, 2018 at 5:32 am #109242

RobGiles
Member

Hi Kit, thanks a million for sharing all that. I will go away and see if I can get something close. They are really great performance numbers, something that I think would suit my objectives very well. Certainly a lot better than buy & hold.

October 7, 2018 at 11:31 am #109244

WENKIT LUI
Participant

Hi Rob

Unfortunately I made a mistake in that backtest, I ran it with adjusted prices instead of unadjusted prices. I have re-run it with unadjusted prices and here are the results for 1/1/2007 to 31/12/2017.

CAGR: 20.8%
Max System Drawdown: 26.01% (occuring in 2010. and 2011 was a bad year too, the annual performance was -8.2%, but all the other years are positive)
Profit Factor: 2.64
Payoff Ratio: 2.30
Win %: 50.45 %

If I change the start date to 1/1/1997, the CAGR drops to 18.26% and the Max System Drawdown deteriorates further to 28.94%.

Apologies for the misinformation. The maximum system drawdown is probably in the uncomfortable zone for a few people, but I’ve been telling myself that anything under 30% is okay.

Note: To rank the stocks, the system currently uses a combination of ADX and RSI (rather than using percentage distance from the 52-week high, as per Nick’s article – this didn’t work for me).

At the moment I’m testing a ranking criteria which another member has shared under the Trading System Graveyard topic, it’s ROC(C, Duration)/ATR(Duration). For the period 1/1/1997 to 31/12/2017, the backtest shows an enhanced CAGR of 20.54% and a reduction in the max system drawdown to a more tolerable 25.97%.

I would be happy to hear from you (or anyone else!) if there are any other ideas regarding ranking criteria

Cheers
Kit

October 7, 2018 at 12:42 pm #109245

RobGiles
Member

No problem Kit & thanks for letting me know. Performance is still good though.

BTW what is the logic of the ranking method:

ROC(C, Duration)/ATR(Duration) ? Are we trying to buy the stocks with the best momentum & least volatility here?

My S&P500 system used 2 Linear Regression measures (one short and one longer term) and then played around with the weighting of them to give a ‘weighted momentum raking’. also had an Index filter and an individual stock filter.

I’ve tried exactly the same rules on the ASX but the performance is poor, so I’m still looking for something acceptable. I’m testing my systems through a particularly difficult period for trend following / momentum systems…June ’07 to Sep ’18, as protection of my capital is very important to me and I essentially wanted to stress test the systems I’m developing to see how they handled that period of time.

October 7, 2018 at 2:26 pm #109246

LEONARDZIR
Participant

In my experience it is one thing to accept a 28% drawdown in backtesting and experiencing a 28% drawdown in live trading if you have a relatively large portfolio.

October 7, 2018 at 9:56 pm #109247

SaidBitar
Member

Kit Lui wrote:

Hi Rob

Unfortunately I made a mistake in that backtest, I ran it with adjusted prices instead of unadjusted prices. I have re-run it with unadjusted prices and here are the results for 1/1/2007 to 31/12/2017.

CAGR: 20.8%
Max System Drawdown: 26.01% (occuring in 2010. and 2011 was a bad year too, the annual performance was -8.2%, but all the other years are positive)
Profit Factor: 2.64
Payoff Ratio: 2.30
Win %: 50.45 %

If I change the start date to 1/1/1997, the CAGR drops to 18.26% and the Max System Drawdown deteriorates further to 28.94%.

Apologies for the misinformation. The maximum system drawdown is probably in the uncomfortable zone for a few people, but I’ve been telling myself that anything under 30% is okay.

Note: To rank the stocks, the system currently uses a combination of ADX and RSI (rather than using percentage distance from the 52-week high, as per Nick’s article – this didn’t work for me).

At the moment I’m testing a ranking criteria which another member has shared under the Trading System Graveyard topic, it’s ROC(C, Duration)/ATR(Duration). For the period 1/1/1997 to 31/12/2017, the backtest shows an enhanced CAGR of 20.54% and a reduction in the max system drawdown to a more tolerable 25.97%.

I would be happy to hear from you (or anyone else!) if there are any other ideas regarding ranking criteria

Cheers
Kit

I use roc/atr in my systems it does good job in smoothening the equity curve
Logically speaking this ranking is wrong since it will favor lower priced stocks over higher priced ones since logically atr is higher on 300$ stock than 2$ one so it is not fair measure but somehow it works and i couldnt convince myself that it is bad :p
Author

Posts