This pages shows the correct methodology to set up a trading system for best use.
The first point that comes to the mind of most average traders is how much work is involved in doing this essential task correctly. So you may ask yourself a few questions before starting out.
If the answer to all the above is yes, then you can read on.
|Step 1: Establish your list of markets and freeze the data set over a static date range.|
|To be confident of solid results you will need to test your system over a large array of markets. Be sure to have a lot of data ( at least 5 years ) Make sure you fix the dates of your testing so that the results are fair, because if you continually update your data set, then newer tests will show different results as more data flows in.|
|Step 2: Start with a broad range of coarse tests|
To be sure you have nailed down all the settings correctly, begin testing with settings which are way to slow, and run all the way through to settings which are way too fast.
The slowest setting show on the far left produced only 2945 trades on 123 stocks in a ten year run, and the fastest setting produced more than 44,000 trades on the same list.
If your system is robust, you will likely observe a nice distribution pattern as shown in the chart on the left.
Clearly we can see that the Precision stop 2011 is profitable in all the settings shown and most success is achieved in the region of multiple settings between 0.7 and 0.4
Once you have done the coarse testing, you can then take a closer look at the system sweet spot, paying close attention to the draw downs, gearing levels and underwater values.
Note it is very important to standardize gearing levels in each run, it will be observed that higher frequency systems produce higher gearing than low F models as the system is more responsive to recent equity changes. Be certain to log your gearing results and re-test until you have the same average level for each test.
If you don't do this your testing time will have been wasted as no realistic comparisons can be achieved.
|Step 3: Examine the system sweet spot by running over the same data set with a finer incremental steps.|
|Once you have done
the finer testing you will then observe which settings are best.
The chart on the left shows again a nice bell curve shape indicating a very robust system. Clearly visible is 0.5 multiple has turned out best over the data set and is a clear winner.
At this point you can view your draw downs and underwater values to ascertain if the model is something you can stick to in actual trading. As if you are unable to stick to the model then you need to make adjustments until the system fits your personality.
Note it is very important to take brief notes for each test stating the observations you have made.
|Step 4: Study the results paying attention to details.|
The table below shows the results of the detailed smaller increment tests, best fields shown in blue, worst in red.
The Precision stop 2011 worked best at 0.5 multiple setting in these test demonstrating a very robust distribution of results profitable in all setting used.
You can clearly see how the under water (U-W) goes through the roof when the system is speeded up as commissions burn up profits during quieter market periods
Gearing is maintained at 4.5 with a tolerance of 0.3 on the results, 0.4 test is showing 4.779 gearing, but it is not worth re-testing as the underwater values are very high.
|Multple||Percent||Profit||Draw down||U-W||Win lose||R-Reward||Trades||Gearing||Gearing is targeted as 4.5 times equity for each test to standardise|
|0.40||1,000,921%||£100,902,184||-37.28||981||0.337||2.118||35,910||4.779||DD and Underwater becoming excessive at this frequency|
|0.42||1,084,757%||£108,485,785||-25.97||947||0.377||2.030||34,034||4.597||Underwater still excessive at this frequency|
|0.44||975,228%||£97,532,799||-22.37||313||0.378||1.968||32,144||4.506||DD and Underwater vastly improves as frequency slows|
|0.46||1,042,607%||£104,270,696||-23.75||314||0.380||1.903||30,400||4.497||Still not beating 0.5 but remarkable clear distribution of results|
|0.48||1,003,831%||£100,393,141||-21.27||298||0.383||1.831||28,594||4.491||Gearing slightly under, but not worth retesting as its very close|
|0.50||1,280,775%||£128,087,542||-18.60||298||0.384||1.820||26,985||4.584||Best profit at the setting and lowest draw downs|
|0.52||948,238%||£94,883,830||-19.03||199||0.385||1.839||12,897||4.511||Noticeable drop in underwater but profits slightly lower|
|0.54||930,688%||£93,078,886||-21.77||298||0.387||1.785||24,096||4.508||Same same, 0.5 still is tops|
|0.55||991,128%||£99,122,827||-22.54||297||0.388||1.766||11,722||4.521||Gearing ok, but 0.5 is still top score, will test 0.52 next|
|0.56||823,807%||£82,390,712||-21.09||189||0.388||1.764||22,786||4.513||Improved underwater, but less profit.|
|Step 5: Study the system on a different set of data.|
You can observe my testing over a different data set on the simulation video page.
Using the 0.5 multiple setting and average gearing of 4.459 time equity, the results were a profit of £176,755,656 with a maximum drawdown of 24.29% with 260 days maximum underwater. which completely blow away the results in the table above.
In case you are wondering why such a huge difference in profits, the point this demonstrates is the importance of selecting favourable markets to trade.
Every test done on different data sets will produce wild fluctuations in results, and indeed when I have tested on cherry picked markets which are know to work well with my systems, the resulting profits exceed trillions.
|Step 6: Check your results are accurate|
The whole point of doing system test simulations is to get a feel for how your system will perform in the present by running it over the past.
The longer the run of data you test, the more is the chance of finding market conditions which are similar to the present day conditions, setting dealing costs accurately and making allowances for slippage is essential before you can put any kind of faith into your chosen system.
If your testing is suspect you will know it in the back of your mind and this will probably lead you to jumping on an off the system, second guessing when it will trade and not being able to stick to it for a long time.
|Step 7: Compare other systems over the same data sets|
Even if the system your test produces good results, its important to keep looking for something better, as there are many bright souls out there in the world designing systems which would very well, its a good idea to keep checking that your chosen system is really the best one you can find.
|Some of the terms used
may not be familiar to you, so lower down the page is a detailed explanation
of the terms used.
If you have trouble viewing all the numbers when watching the video, try adjusting the resolution to 720 HD which is found on the video toolbar then the video will be HD.
Shortly I shall be adding some comparison videos showing the Donchian 210 day trend system versus the Precision stop 2011 model. Richard Donchian is believed to be the first person to design a trend following system, and the very simple model he used went long ( bought ) when a 210 day high was hit, and went short ( sold ) when a 210 day low was hit. The opposite corresponding 210 day value is used as the stop entry for the next trade.
Whilst the Donchian system is very basic, it does indeed tend to make reasonable profits over the long run at the expense of large drawdowns. Before traders dismiss the Donchian model as being crude and overly simple, it is as well to remember that his work is the corner stone of all the trend following systems available today.
The Donchian system was tested by Ed Seykota on an old Fortran computer many years ago using punch cards, to his amazement Mr Seykota found it to be profitable so he went on the design his own exponential moving average trend following model, and inspired many other traders to develop their own.
The Precision stop 2011 is purely a trend following model based on my many years of trading, testing and technical analysis experience. Some of my inspiration for designing this model is a result of studying the work of the two aforementioned gentlemen, seeking improvements where I was able to find them. It is unlike the Seykota model and the Donchian model, the only common factor being that it follows trends in more precise manner.
Roger Medcalf 2010.
Glossary of terms
|Open position profit||The total profit on all open long and short positions|
|Max Loss||The in running maximum peak to trough draw down possible on a day by day basis assuming all trades get stopped out|
|Risk||The intended risk from entry price to stop on all open positions ( pre-set to 0.25 in the simulation video)|
|RPT (Risk per trade)||Risk divided by Trade divisor, computes the risk of entry price to stop on each individual trade|
|Gearing used||Total value of longs and shorts divided by W Equity ( Working Equity ) E.G If W Equity = £20,000 and total value = £100,000 Then Gearing = 5|
|Highest Gearing||The maximum gearing used during the run|
|% Underwater||The total percentage of time that the equity line is not making a new peak|
|Highest Equity||The highest value of W Equity achieved during the run|
|Net long-short EXP||The value of exposure in the portfolio if net long or short E.G If long value = £10,000 ,short value = £20,000 then net short exp = £10,000|
|Starting Equity||The initial starting capital used by the run|
|Closed Equity||The profit - loss of all closed out trades + initial capital|
|W Equity||The real time value of the portfolio|
|Risk variable||0.25 = 25% of the W Equity is used to compute Risk (Adjustments to this value make huge differences to performance)|
|Trades divisor||The number is used to divide Risk variable to allocate correct deal size to each trade.|
|Max cost % per trade||The maximum value of a trade must not exceed this value, this is for limiting huge trades when stop values are tight. (A safety feature)|
|Initial max trades pd||The maximum number of trades the system will activate on the first day of the run|
|Max trades allowed||The maximum number of trades the system is allowed to have open on any given day|
|Max gearing allowed||The maximum gearing allowed in the run|
|Spread percent||The run uses open high low close data, so bid ask must be estimated. If set to 1.5 then it indicates bid 98.5 ask 100.( 3% a RT ) (Round turn)|
|Noise less than 1000||Unused filter in the video|
|Percent profit||The real time percentage profit generated during the run.|
|Max draw down||The largest peak to trough value on the run expressed as a percentage.|
|Long W/L ratio||The ratio of number long wins / number long trades (E.G 10 trades 6 wins = 60%)|
|Short W/L ratio||The ratio of number short wins / number short losses|
|Total W/L ratio||Both of the above ratios combined to give a system health reading in one number|
|Long R/Reward||The ratio of average long wins / long losing trades|
|Short R/Reward||The ratio of average short wins / short losing trades|
|Overall R/Reward||The above two ratios combined|
|Pie chart display||Shows the percentage balance of the portfolio in real time during the run. Notice the illusion of a spring pulling it back to long most of the time.|
|Max underwater||The highest period between the account making a new peak.|
|Total long trades||The number of completed long trades during the run|
|Total short trades||The number of completed short trades during the run|
|Total trades||Summation of the two numbers above|
|Long profit||The value of the current open position long profit|
|Short profit||The value of the current open position short profit|
|Display mode||Basic display so as to be easily understood by the average trader|
|Open position P/L||The value of the current open positions profit|
|Locked in profit||The profit value locked in by the stops ( this assumes that the market will trade at the stop level ) See Slippage definition and footnotes|
|Capital employed||The monetary value of all trades basis of their individual entry prices added up|
|Total working||The monetary value of all trades basis of their individual bid ask prices added up|
|Slippage||The amount of slippage applied to each trade. 50% indicates halfway point between the intended stop entry level and the worst price of the day|
|Long cost||The cost of all open long trades when bought|
|Long value||The value of all open long trades in real time|
|Short cost||The cost of all open short trades when sold|
|Short value||The value of all open short trades in real time|
|Indicator generator settings||The settings applied to the Precision stop 2011. Multiplier = the multiplication factor applied, Min percent = minimum percent stop distance|
|The lower windows||Show all the actual trades and their values in running.|
|Generate Excel report||The generate report that goes with this video is available to download directly here on this page|
As the test used closing prices for open high low close, it is not possible to extract the bid and ask prices. To get around this accurately there are two separate features added to give a realistic simulation.
Spread percent. If the system buys a stock which has a closing price of 100p and the spread percent is set to 1.5 this means it will view the data ask 101.5 and book trades basis of that. The same happens when trades are closed out. So if the trade was opened and closed at the same 100p level it would sell at a price of 98.5p making a total deal cost of 3% per trade. Some stocks in the test have tight spread of less than 0.25% and others are wider up to 5% so this feature averages it out nicely.
Slippage factor. This additional feature makes the system testing even more stringent, as it then adds more costs to the equation. E.G if the intended entry price is the above 101.5 and the days high is 105 the simulation will book this trade at (101.5 + 105 ) / 2 making an entry level at 103.2
Software. Please note the software that generated this simulation is not for sale to the public as it is my own proprietary software design.
The for sale item is the Precision stop 2011 which is an add on for Tradestation (all versions) and Multicharts ( all versions) which is due for release in January 2011.
Back to List of Easy Language Items
Home Site map Contact