Developing a System to Trade Live

[mit]

Hmm I'm interested in your explaination of "markedly different"

The only things I find different are
(1) length of trade governed by tightness of exit.
(2) Generally more trades---cause your in a trade less time.
(3) With winning systems (I've only seen winning futures short term methods) More winners.
(4) Less Reward to Risk.
(5) Less runs of consecutive losers.

Short term systems tend to look at price action over the last couple of days, for reversal or continuation patterns etc. The reaction to these short term patterns tends to peter out after a few days. They don't all readily expand to a long term system merely by expanding the period.

Also visa-versa. The Alan Hull system does nicely as a weekly system looking back around six months. It stinks as a daily system looking back a couple of months and I would suppose even worse as a short term system. This makes sense as it is a system that is trying to get you into an established trend. I think you would look at that kind of system and it would say longer term system even without optimizing etc.

MIT

 

[tech/a]

I do find this interesting MIT.

I dont think the premiss that the reason for entry peters out in the short term is restricted to short term trades is valid.
I'd go as far as to say that ALL entry criteria peter out and become inconsequential at some point.

Do you mean that a short term system should have an exit at the point where the entry is no longer valid---if so how do you judge this?
Do you also see this point as being the First sign of reversal of the entry condition--IE price goes against the direction of the entry signal.

Would you not loosen the exit criteria as a trade matures and still rises?
Why limit your potential profit simply to satisfy "short term"?

These are points I have often pondered---I offer them not as critisisms but as open thoughts to someone who has veiws which may be helpful to my thinking.

On OPTIMISATION my veiw is that optimising anything other than a singularity is an exercise in futility---to expect NAB and or BHP to perform better to the optimised variables of say WOW just isnt practical.
Sure if your trading 1 stock or 1 future optimise your brains out!.

Im not presenting my veiws as gospel---they are my findings over many years,Like Hawkins Im happy to take on new findings.

Interested in what you have to say.

 

[mit]

Tech,

I don't think we are in fact disagreeing. I am more talking about when you get down to specifics systems. In the general tightening stops etc will make a system shorter term and loosening stops will make it longer term.

A short term system might be for example (not a real system of course)
buy on a gap up opening after two lower lows. Initial stop on the lowest low and trail by 1 atr. You could certainly turn it into a long term system if you wanted to by having a big sloppy 4Atr trailing stop or a cross of the 180MA. I think that loosening the stops would lower the expectancy. My feeling is that the power in this is as a short impulsive move. These kind of events are caused by news and the masses quickly jumping on board. You jump on and hang on until the first sign of weakness.

On the other hand you may have a system that buys when a stock is oversold and the 100MA crosses up over the 200MA. Here I would put a big sloppy trailing stop in as to me it reads like a long term system. Nothing would stop me putting in a 1ATR stop but I would expect to be whipped out due to market noise.

So all I am really saying that as soon as the group decides that we will look at such and such indicators/price action, it will be fairly obvious that it would be a medium/long or a short term system.

Anyway enough from me on this subject as I think I am distracting from the main thrust of this thread.

MIT

 

[RichKid]

I'm very interested in this testing discussion but don't have much to offer due to a lack of expertise but may I suggest a look at resources/energy stocks due to the current cyclical run up for that sector (long term commodities bull market). Maybe ensuring some diversity in the portfolio by picking a certain number of non-correlative markets too. You'll need some rules re reducing exposure to maintain diversity, unless the rule is to just let profits run until the stop is hit (rather than taking profits to re-balance exposure to a particular sector).

You could test Martin Roth's book 'Top Stocks 2005' and previous editions or just his new small co's edition (not enough editions to back test). Since he's vetted the co's already (and has a published track record via his books) you can back test it. Wonder if you've done that already Tech?

 

[tech/a]

MIT.

Indeed you are correct we are both holding similar veiws each expressing the same slightly differently.(I'm glad about that was beginning to question many hrs work!!).
This discussion I think will be important as the method is developed as we will see the points raised in testing---so we will have an idea what to expect.

Rich.
You are talking of a stock selection process and there is no reason why you cant use your suggestion.Im sure we will come up with many universes.

I personally dont see this as a problem as we will be able to apply the method to ANY universe of stocks offered by any member of the forum.

Please if you wish to select stocks please go ahead and submit your universe for testing..

I'm even thinking about testing the resultant method on NYSE,FTSE,HONG KONG and DAX stocks---as I have the data.

 

[Milk Man]

I think Rich's point about out-performing indices is a good idea. Maybe limiting buys to the best couple of indices (diversify but not diworsify)? We could even just buy the index itself. I'm assuming that were all talking shares though. Maybe we should have a vote to decide - or is everyone happy to go with shares?

 

[tech/a]

The Out Performing Index Arguement

Guy's Ive done a lot of work on this well touted theory.

Trade the stock which is the out performer in the out perfoming index.

While a great theory there is one thing that is missing.
Simply by the time you can identify the out performing Index and the stock or stocks within it the move has well and truely gone.--Youve missed it
Particularly if your talking shorter time frame.

If anyone has Metastock or would like the metastock code that determines these out performing Indexes and Stocks within the indexes just let me know and you can have them.Many have taken them in the past and I've asked that if you can find an edge please let me know.

In 5 yrs no takers!

Sorry but I think its untested out dated theory!---used to sell books of little practical use.

Now EMERGING Outperforming Stocks and Indexes--that is an entirly different proposition----this is our challenge when finding a universe and a portfolio.

 

[RodC]

I'm happy to go with shares, I also wouldn't mind seeing a weekly system, although I imagine most here would prefer daily.

Also as tech/a mentioned earlier This shouldn't be anything like techtrader, it's already out there and has been very well discussed in other forums.

What I'd really like to learn is the process of system development.
ie:

Step 1: choose vehicle stocks/futures etc.
Step 2: choose trading universe (test against index)
Step 3: choose daily/weekly etc.
Step 4: choose indicators
Step 5: test
Step 6: refine
Step 7: test.

and so on.

Rod.

 

[cabbie]

G'day

Here's a possible alternative to the outperforming sector idea, why not optimise the trend for each sector, then test for stocks that are breaking out in trending sectors, maybe modify the break out params depending on the sector. Or a similar varriation.

My two bobs worth anyway

Regards

 

[Artamon]

G'day all,

This is a great idea (as was the T/Trader experiment). I am glad i've found this at the start of the developmental process and hope to contribute what my limited knowledge allows and learn along the way. RodC summed it up pretty well for me.

As for the system, my vote is for shares on a daily timeframe.

Many thanks,
Andrew.

 

[RichKid]

Now EMERGING Outperforming Stocks and Indexes--that is an entirly different proposition----this is our challenge when finding a universe and a portfolio.

Well, think of it via the stock analogy, a breakout may look specatacular at first and we may think we've missed it but as T/T has shown if you follow it long you can make big profits.

Now consider something like the commodities bull runs, they look real strong, then correct (as is happening this year) and then they run even stronger, it's all cyclical. So what exactly is an 'emerging' sector? If it's going to run even harder then wouldn't it currently be emerging? Maybe we need to look at longterm commodities charts/indices and then go from there. Oil for example is not going to go down anytime soon so the oil co's will be making money for some time. As I understand it all commodities run together so we need to have some diversification as well so we aren't too heavily in one sector.

As for time frames I reckon weekly is good (less time consuming compared to a daily system) and I'm happy to stick to Aussie shares, makes tax and other things easier as well since it's a familiar universe to start with. Also in terms of commodities we have some real gems on the ASX.

 

[tech/a]

Rich.
There is merit in what you say re indexes---Why dont you persue compilation of a universe based around that.(Or anyone else for that matter).

ON WEEKLY SYSTEMS
Firstly a daily system (even short term if we can find one that is profitable) wont take long at all to trade particularly when we have a full portfolio.

The one huge problem people have with weekly systems is a sell signal on Monday with an exit stratagy on the NEXT open---so you have to wait all week.In which time it can drop further,claimb back OR do little.
Once you understand that the system takes into account these 3 occurences and some will win some back and some lose some more----it isnt a problem.
But most cant handle this aspect--they feel a sell trigger should be acted upon immediately.Sure you can do intraday/or week exits but from a practicality veiw most dont see the days trading results until night so cant react till next day.
Hence the delay buy AND sell 1 day.

A weekly system will be longer term---Shortest would be 2 weeks--2 bars.

ON SHORT TERM SYSTEMS

I've thought for many years that the reason short term methods in stocks particularly portfolios tend to fail(I havent found one that retuns good profit in 12 yrs) is that unlike Futures which have many short term very profitable methods---we are trying to apply short term trading startagies over a GROUP of candidates---with Futures its ONE.

My theory is that if you have as little as 3-5 stocks and design the best "System" INDIVIDUALLY FOR THAT STOCK rather than a blanket universe---youll find a great way to trade.This would be pretty simple and you could optimise your brains out on each individual candidate.
.

What do you think of this idea??

 

[Milk Man]

Will that (short term theory) restrict use of capital? It sounds excellent for options trading where youre heavily geared but will it give enough oppurtunities for shares?

I think the use of a funny-mental approach to selecting a universe of stocks sounds interesting. If we want to be consistent though its a bit hard to back test unless theres a technical entry/exit signal for each idex were using in the universe at the time.

I reckon daily timeframe and don't want to see a too longer term method.
If I have to sit there for twelve months of ebbs and flows in a stock i'll most likely fiddle with the method and stuff it up. Something using eod data unless an automated stop can be used with the broker. I'd be stoked with a 6mth hold period.

what does everyone else think?

 

[mit]

ON SHORT TERM SYSTEMS

I've thought for many years that the reason short term methods in stocks particularly portfolios tend to fail(I havent found one that retuns good profit in 12 yrs) is that unlike Futures which have many short term very profitable methods---we are trying to apply short term trading startagies over a GROUP of candidates---with Futures its ONE.

My theory is that if you have as little as 3-5 stocks and design the best "System" INDIVIDUALLY FOR THAT STOCK rather than a blanket universe---youll find a great way to trade.This would be pretty simple and you could optimise your brains out on each individual candidate.
.

What do you think of this idea??

I agree. There is also the fact that you are in and out so often trying to physically manage more than 4-5 stock universe is difficult. The way I managed it before was to backtest a system using a universe of the asx200 stocks. If it showed some profit and find the best 4-5 stocks. The *BEST* does not neccessarily mean most profit for the period but things like
. Numbers of trades.
. Expectancy still high after the best trades taken out.
. smoothest equity curve.

I would then optimize over all 4-5 stocks. Check that the parameters are still robust. I would then optimize individually. But if the parameters moved too far I would get suspicious.

I hope that nobody blasts me for the word "optimize" I am well aware of the problems of curve fitting and the need to test parameters across years and to test sensitivity and so forth.

MIT

 

[tech/a]

Loakglen.

I'm thinking this method will target traders who ahve about $10,000 in capital.
They will trade unleveraged.
They will have a position size of $3333 for 3 stocks----$5000 for 2
I'm thinking any position size smaller would be less likely to succeed short term.
Universe of stocks will determine frequency of trading.
We could have say 10 or 20 which are continually rotated in and out Think 5-10 best less if sufficient trades.
Only 2 or 3 will be held at any one time.
Profits will be added to capital and position size increased with profit thus compounding wins.
Each stock will have its OWN optimised technical entry and exit (Fundamental can be used for universe but I feel that to find MOVERS a technical approach will suit the short time windows we wish to catch).
Setting of stops will also be governed by BEST performance parameters of each stock.


MIT we crossed---yes I certaintly agree regards comments in particular optimising it is important (and nice to work with people who understand) to understand that optimisation selection doesnt necesserily mean the very best result from optimised parameters.

This is just the basic model---thoughts?

 

[Milk Man]

$10000 unleveraged may leave it a bit thin short-term given all the brokerage. It does help out those low on capital though. It would be interesting to see if we could make it work. Maybe $10000 geared up to $20000 or even $20000 cash would be better- that way positions can be 4 X $5000.

I like the optimisation idea. I'd like to see what everyone comes up with to pick the best potential movers.

 

[RichKid]

Rich.
There is merit in what you say re indexes---Why dont you persue compilation of a universe based around that.(Or anyone else for that matter).
..........

Tech,
I like the ideas. How about using 6-10 stock to ensure diversity across sectors, then if a sector gets hit we still have some runners from another one? I may cross reference with Roth's book too.
I'll look up some commodity/resouce stocks (my favourites) and maybe some cyclical stocks to mimic resources cycles. Getting to know a few stocks is a great idea, especially bluechips with longterm chart trends.

 

[mit]

For the under capitalized could try to use CFDs. The commissions are lower but you would have to account for the spreads, hold costs and the fact you cannot get involved in the opening and closing bids. If I had less than $10k I would use CFDs.

If it is for outright buying of shares. Select shares with the highest volatility (beta) could be part of the final universe, this would give greater bang for your buck.

MIT

 

[RichKid]

$10000 unleveraged may leave it a bit thin short-term given all the brokerage. It does help out those low on capital though. It would be interesting to see if we could make it work. Maybe $10000 geared up to $20000 or even $20000 cash would be better- that way positions can be 4 X $5000.

I like the optimisation idea. I'd like to see what everyone comes up with to pick the best potential movers.

Maybe $20,000 then? That's within reach for most and if they can't then they can just halve it. A small float isn't a problem if we get the numbers right imo, just smaller bets and smaller returns at the start. $40 round turn brokerage is common for discount brokers and is not a big percentage for a $2k or $3k position.

Can you trade CFD's with low margins? ie only borrow the minimum? That'll reduce the downside risk and margin calls and make it closer to share trading.

 

[RodC]

$20,000 with 4 or 5 positions seems reasonable, I don't think smaller positions are worthwhile (there are of course exceptions!).

So it looks like we're heading down the path of short term system with a relatively small universe of stocks with a maximum of 4 to 5 positions at any one time? Sounds interesting!

tech/a, when you mentioned optimising individually for each stock, does that mean you effectively have a slightly different system (within the overall system) for each stock?

Rod.