Australian (ASX) Stock Market Forum

Trading System by Andrew Gibbs - Member of Larry Williams Hall of Fame

Rightio, so let's have a look at stock selection. To do this I use Portfolio123 so we're going to focus on US shares. www.portfolio123.com.au . Btw I am getting ASX and NZX shares added to Portfolio123 so stay tuned.

There are two different ways to come up with a stock basket to trade.

1. The easy way - let other people do it for you
2. The hard way - DIY, we'll look at examples further on.

In these examples I am going to show you the easy way.

Firstly let's look at what a stock index is. Really it is just a portfolio of XX number of stocks with stocks getting added to and removed from the index every so often..it sounds rather like a portfolio doesn't it. In fact most funds are invested into this type of portfolio (i.e. a stock index such as S&P500 in US or ASX200 in Australia etc) and that's where your superannuation goes.

So here's a question for you, which index would you rather trade... the blue one, which is the S&P500 index or the red line which is a basket of 20 stocks that makes up part of the S&P500 index at any one time?

S&P Above Average.png

The red line averages 25% with a 24.5% drawdown whilst the S&P500 ETF Trust (which mirrors the S&P500 index) has averaged 2.5% with a 55% drawdown.

You can find out more info by going to www.portfolio123.com.au and clicking on the Ready 2 Go portfolio tab. This portfolio is put together by a guy called Hemmerling, not by me.

The red line includes survivor-ship bias, brokerage and slippage.

It obviously makes far more sense to focus your attention on those stocks that are more likely to outperform the market and therefore just trading the 20 stocks that make up the red lines performance is going to result in a better outcome than trading all of those stocks in the S&P500 index itself.

Next I'll explain what the red line represents, how the stocks are selected and what you'd need to do each week to make those stocks part of your basket which you can then apply the technical system outlined above to.

ZZ...
 
Portfolio123 runs off the Compustat Point in time database and allows you to backtest based upon fundamental data, ratios, analyst forecasts, estimates, earnings surprises etc etc. This is very powerful software, the best I've seen.
 
screenshot.JPG

I dropped the logic into amibroker and ran against the asx200 - results attached.

Pretty similar in profile to what was described earlier by Andrew for the S&P.

The drawdowns would require some strong medicine though.
 
View attachment 54808

I dropped the logic into amibroker and ran against the asx200 - results attached.

Pretty similar in profile to what was described earlier by Andrew for the S&P.

The drawdowns would require some strong medicine though.

Is the idea to trade the entire ASX200?
 
Nope.. the idea is to have a technical entry and exit rule for trading a basket of about 20 stocks selected based upon their fundamentals.

So the drawdown considering the entire ASX200 universe is a worst case scenario?
 
Nope.. the idea is to have a technical entry and exit rule for trading a basket of about 20 stocks selected based upon their fundamentals.

Yes, sorry should have mentioned that - identifying the preferred sub set is the next step, havn't been through the posts describing how you do that that bit yet, but clearly it should be done in isolation of the backtest results :)
 
Hmmm, 18,000 trades, 4 simple rules across 100 stocks, same rules tested across all of them, no optimisation at all.. tested back 30 years and the system makes money on every single one of them... yeah..crap system, throw it in the bin.

LOL except you once again gloss over the key fact that the 100 stocks chosen are the current 100 largest stocks by market cap in the US, comprising almost 50% of the total US market cap. So these are stocks that have gone up by definition, and any stocks which went down (say, Enron, which used to be a S&P 100 constituent and probably demolished on this system) are not included in the backtest.

You say you tested 30 years across 100 stocks that is a patent lie, as I have already shown. The majority of the stocks you test were not S&P 100 components for a majority of the 30 year "backtest" and many of the consituent stocks have not even been listed companies for 30 years. So considering Google didn't list until 2004 and didn't even exist until sometime in the 90s, how is it possible for you to have tested these particular 100 stocks for 30 years?

If you don't think picking the current S&P 100 as your universe for a 30 year backtest on a mean reversion system is optimisation then you are fooling yourself and I personally don't think you should get the opportunity to fool others on here.

You promised to address this issue and show how your stock selection screen works, but all you've done is instead spruik for another website. So we are up to SCM, Halifax and now add "Portfolio123" to the list.
 
P.S:

Just for the record, I never said the system was crap I said, specifically, that the backtest was crap and your attempt to use the backtest as validity for "robustness of the system" was even more crap.
 
View attachment 54808

I dropped the logic into amibroker and ran against the asx200 - results attached.

Pretty similar in profile to what was described earlier by Andrew for the S&P.

The drawdowns would require some strong medicine though.

Considering that a large majority of the companies currently in the ASX200, weren't before, what meaningful conclusion do you expect to take from this backtest?

Garbage In, Garbage Out
 
Thanks for that Sinner.

Before this gets into an argument, let's digress, you are right with what you are saying and it's an important point you are making, however it is not the main point or even an issue with the entire method.

If this was purely a technical trading system then sure, the backtest was crap. However for the purposes of what I was trying to acheive the backtest is fine. My goal with the system was to come up with a method for entering and exiting a stock that captures virtually every move. Any technical system that captures every signel move will only every have a marginal edge.

In this case 90% of the method/results or edge comes from stock selection using Portfolio123 with the technical system simply being a way to enter and exit the trade and therefore as long as it does this to a satisfactory degree , that is adequate.

The whole point for this series of postings is that the stock selection (for this method) is way more important than the technical system. The point of the technical system is simply to find something reasonable that gets the trader into and out of the stock, which is why I put together an active method that is in the market a lot. As I mentioned it captures almost every move. It does that..job done.

The other thing is I am not worried about the survivorship bias for the backtest, in fact I want to test on a bunch of stocks that has an upward history because the actual basket of stocks we are selecting from Portfolio123 (P123) should have an even bigger upward bias.

P123 includes survivorship bias in its backtesting so in this case it is the engine driving any profitable results, not the trading system.

This system is only suitable for trading with the fundamental overlay, I mentioned that from the start.

Stock selection first + technical system for entries and exits=entire method ex money management

ZZ...
 
The most important consideration is which stocks to trade and how do we select them. Given the system is in the market a lot and very active, say one or two steps down from buy and hold as I just said the key to the entire thing is to find stocks that are most likely to outperform the market, we only need about 20 and that is the real edge and is what I'll cover next.

ZZ..

My experience is that if you can actually identify 20 odd superior stocks then the best thing to do from that point is nothing.

Eagerly awaiting a robust back test on the ACTUAL VARIABLE universe to challenge my beliefs. Does the trading add return after transaction/slippage/taxation? Or are you shooting for risk adjusted return – (assuming that you call volatility risk)

I have used Radges Growth portfolio as a universe in the past on the assumption that the right universe would improve an edge – conclusion. Edges don’t add to each other – they subtract.

I have the same conclusion about adding trading to a FA derived portfolio’s. Conceptually the idea sounds very appealing but reality has always come up different. The best trends to ride in a trading time frame occur on highly leveraged (operational/financial) companies experiencing changing macro conditions or on stocks that are bloody hard to value and the market is performing price discovery in a pretty wide range. The best long term FA stocks are pretty much the reverse.

The portfolio’s you have mentioned with the turnover you have indicated don’t strike me as long term. Since launch, none of them appear to have been running live for more than 1 year in a very strong US market and the since inception numbers are subject to the backtest problems Sinner has mentioned. No real basis for knowing just how robust those portfolios, which are effective trading systms in themselves, will be long term.
 
My experience is that if you can actually identify 20 odd superior stocks then the best thing to do from that point is nothing.

The portfolio’s you have mentioned with the turnover you have indicated don’t strike me as long term. Since launch, none of them appear to have been running live for more than 1 year in a very strong US market and the since inception numbers are subject to the backtest problems Sinner has mentioned. No real basis for knowing just how robust those portfolios, which are effective trading systms in themselves, will be long term.

Hi Craft, lots of great advice there.

Just in relation to your last comment the portfolio/s I mentioned from portfolio 123 are backtested correctly. The database includes delisted stocks, survivorship bias, slippage and commission.

The last sentence is good and without knowing just exactly what rules the systems are looking at are, there is no real way of knowing. However, as traders that is the problem we always face with a new system.. i.e. not knowing what the future will bring and hoping our assumptions of the past will hold up.

The advantage of running custom built Portfolio123 portfolios is that we known the assumptions behind the system, with the Ready2Go models the actual rules is are only outlined in the description but not disclosed. I should also mention that with the Custom built Portfolio123 models we can then also backtest the universe correctly using the technical model and then everyone will be happy ;-).

I'll post some trade examples in the next few days.
 
Considering that a large majority of the companies currently in the ASX200, weren't before, what meaningful conclusion do you expect to take from this backtest?

Garbage In, Garbage Out

Conclusions ............

Once I can find some recent stock selections packs, then you can have some conclusions.
 
Thanks for that Sinner.

Before this gets into an argument, let's digress, you are right with what you are saying and it's an important point you are making, however it is not the main point or even an issue with the entire method.

lol. I am not really "arguing" anything here, let's make it clear, I am making some factual statements about your nonfactual ones.

If this was purely a technical trading system then sure, the backtest was crap. However for the purposes of what I was trying to acheive the backtest is fine. My goal with the system was to come up with a method for entering and exiting a stock that captures virtually every move. Any technical system that captures every signel move will only every have a marginal edge.

This is nonsense. Why do you need a technical system to capture "virtually every single move"? Is it maybe because this way generates a much higher frequency of buy/sell churn of the portfolio which would otherwise have holding periods more like quarter/year/5Y?

In this case 90% of the method/results or edge comes from stock selection using Portfolio123 with the technical system simply being a way to enter and exit the trade and therefore as long as it does this to a satisfactory degree , that is adequate.

Right so this is a system which derives its alpha from well known, well researched and documented factors such as value premium or liquidity, or some combination thereof. Except this is precisely the thing you've not shown, just some technical trading rules optimised for a hindsight adjusted universe which seems to juice returns but really only because it's buying dips (i.e. long term mean reversion) on known survivor stocks and switching to short term mean reversion in high volatility regime (where 6 month momentum is negative).

Oh wait, I forgot you also spruiked Portfolio 123.

The whole point for this series of postings is that the stock selection (for this method) is way more important than the technical system. The point of the technical system is simply to find something reasonable that gets the trader into and out of the stock, which is why I put together an active method that is in the market a lot. As I mentioned it captures almost every move. It does that..job done.

Don't you mean to say here, that the whole point for this series of postings is to add broker friendly overlay to overtrade an otherwise simple and existing quantitative funamental analysis? You have provided exactly ZERO evidence that the technical overlay adds any alpha to well known quantitative fundamental analysis. To make matters worse, the technical system provided is completely arbitrary (i.e. apparently a stock with RSI at 26 is horrible and shouldn't be bought but a stock with RSI at 25 is a great buy).

The other thing is I am not worried about the survivorship bias for the backtest, in fact I want to test on a bunch of stocks that has an upward history because the actual basket of stocks we are selecting from Portfolio123 (P123) should have an even bigger upward bias.

This is complete nonsense. Please explain how Intel at IPO is a valid stock for any form of reasonable testing, simply because it's a current constituent of the S&P 100? Out of the 30 year backtest, for exactly how many weekly periods was CISCO a "value stock" which would have shown up in any of the screens?

You are fooling yourself. Please stop stating nonfactual statements as factual, in case someone who doesn't know any better falls for it.

P123 includes survivorship bias in its backtesting so in this case it is the engine driving any profitable results, not the trading system.

This system is only suitable for trading with the fundamental overlay, I mentioned that from the start.

and yet instead of providing a backtest based on a universe suitable for the system, you generated a charlatan backtest and claimed it tested 100 stocks for 30 years and this proves how robust it is.
 
Thanks again for your contribution Sinner, certainly makes my afternoon a little less dull.

lol. I am not really "arguing" anything here, let's make it clear, I am making some factual statements about your nonfactual ones.

Fact. I tested the system on the current stocks that make up the current S&P100 index. I tested back as far as 1980 on each of these stocks if the data was available.

The purpose of the test was to see quickly and easily if the system worked reasonably well. It does in fact work well on those stocks as it made money on almost every one of them probably because as you say it buys dips on stocks that over the long term have gone up.

I don't have a problem with that for the purposes of this trading idea.

Issue: The universe was not tested on stocks removed from the index nor the constituent data historically which means the results look better than they would have been if bankruptcies and delistings were included... Agree, but it makes money on stocks that have a long term upward bias... did I mention that was the point of the system all along?

Similar results in terms of profit factor and win:loss ratios were achieved on the S&P500 and Russell3000 stocks. Same issue survivorship bias.. you've covered that adequately.

So what conclusion can you draw from this. Sinner thinks you can't draw any conclusions because it wasn't tested correctly. I think the conclusion that can be drawn is that the rules do what they were designed to do in the first place... Show an active method to buy and sell a small number (10-25) of stocks that are likely to go up based upon the fundamental research or stock screen.

This is nonsense. Why do you need a technical system to capture "virtually every single move"? Is it maybe because this way generates a much higher frequency of buy/sell churn of the portfolio which would otherwise have holding periods more like quarter/year/5Y?

The purpose of the system is to give us an entry rule that will get us into the market early on in a new trend and an exit rule that is trend following. The average hold is 26 trading days (5 weeks) so it's not really that high frequency. I think it's safe to say (and correct me if I'm wrong) that it's probably a fairly common time frame for a good many traders. Instead of trading a Ready2Go model traders for stokc selection traders could use Portfolio123 as a stock screener. The point is the fundamentals screening technique can and should also be back-tested as that is how you pick the stocks that "should" have more chance of going up.

Right so this is a system which derives its alpha from well known, well researched and documented factors such as value premium or liquidity, or some combination thereof. Except this is precisely the thing you've not shown, just some technical trading rules optimised for a hindsight adjusted universe which seems to juice returns but really only because it's buying dips (i.e. long term mean reversion) on known survivor stocks and switching to short term mean reversion in high volatility regime (where 6 month momentum is negative).

Yep, the basket of stocks is derived from well researched and well documented factor that should generate alpha, you can use the Ready2Go portfolios as a short cut or DIY. Then we apply a system which is shown to work on stocks like this to get us into and then out of the market, a system that makes money from buying dips or breakouts from consolidations..no rocket science here. Let's cover Alpha below.


Oh wait, I forgot you also spruiked Portfolio 123.

Yep! Portfolio123 is an excellent product, it can give you a real edge in the market and it backtests fundamentals, other similar products do not do backtests on the screening technique or fundamental (quant) criteria so how can you possibly beleive that the screens are likely to give you an edge or generate alpha.. you can't which makes Portfolio123 so good. Don't see a problem with "spruiking it". Thanks for giving me another opportunity.


Don't you mean to say here, that the whole point for this series of postings is to add broker friendly overlay to overtrade an otherwise simple and existing quantitative fundamental analysis? You have provided exactly ZERO evidence that the technical overlay adds any alpha to well known quantitative fundamental analysis. To make matters worse, the technical system provided is completely arbitrary (i.e. apparently a stock with RSI at 26 is horrible and shouldn't be bought but a stock with RSI at 25 is a great buy).

No, my intention is twofold, one we've already covered (the commercial motive), the other to draw people attention to stock selection and it's importance.

I concede there is no evidence to show the technical system adds extra alpha. All it provides is an alternative for getting into and out of the market to simply buying or selling at US open at the beginning of the week.

What I have provided is evidence that the technical system does indeed make profits on stocks with a long term upward bias and hopefully these are the stocks being selected using the P123 screens, R2G models, DIY models etc.

Your other point about RSI. It's an entry rule, something to make you take action, the RSI rule gets you in on a dip. If a breakout occurs before the dip you buy on a stop, which-ever occurs first. More trades were entered into from the RSI rule than the breakout rule.

Fail to grasp your point on your RSI comment when clearly the fundamental overlay/stock selection is what makes the stock a great buy.

This is complete nonsense. Please explain how Intel at IPO is a valid stock for any form of reasonable testing, simply because it's a current constituent of the S&P 100? Out of the 30 year backtest, for exactly how many weekly periods was CISCO a "value stock" which would have shown up in any of the screens?
You are fooling yourself. Please stop stating nonfactual statements as factual, in case someone who doesn't know any better falls for it.

Talk about miss the point Sinner, for up-tenth time....blah blah blah, I am not even going to repeat myself. It is completely valid because the stock being traded with the P123 basket might have recently been an IPO. The system clearly makes money on stocks that go up because it buys dips. It will lose during bear markets and on stock that go down. This can be handled from the Portfolio123 side. The basket of stocks goes to cash during down periods for whatever reason.



and yet instead of providing a backtest based on a universe suitable for the system, you generated a charlatan backtest and claimed it tested 100 stocks for 30 years and this proves how robust it is.

Definately a bad choice of words. To be fair I rushed the back-test because of prior comments however I did mention survivorship bias was an issue and you've made your thoughts pretty clear on that and I've made my point about it, you think it is of utmost significance.


Here's the exact same system rules back-tested on the current stocks listed in the Russell3000 tested back to 1995 where data is available.

Russell 3000.png
 
I'm curious. Can you run the same back test on the shares making up a single sector within the all ords and produce a summary of results similar to the above. For instance the S&P/ASX 200 Property Trusts (XPJ) for the period January 2009 to September 2013, using only those shares listed as at today (don't get bogged down on mergers/acquisitions/restructures etc) trading parcels of $10,000.00??
 
Top