Diversified Trend Trading – Excel Projects

Cheers guys

AJS

I thought I would do a run on a 45 instrument portfolio from woe to go using your settings supplied in Post 70 (using the data history supplied in Post 74) to test the speed of the process from portfolio creation to final outputs of the test.

I started at 8:39pm my time and it is currently running. I will let you know how long it takes and supply you with the results. The aim of this exercise is to demonstrate that this is a pretty quick process once you get the models up and running, and also provide an example that you may wish to attempt to replicate using a far larger investment universe.

I am hoping that the instructions given in the last few posts have worked for you. If not, then just let me know and we can troubleshoot together.

The good thing about working on projects such as these is that we can exactly replicate results across the group provided we utilise the same source data and broker settings. Once we get a handle on the excel spreadshsheet modelling, we will then be in a position to splinter into project groups to collaborate and run comprehensive backtests and robustness tests with our models which we can then apply to our live trading activities.

It is 9:00pm my time now and already we have a portfolio of 26 instruments created. Once the 45 instruments have been created it is going to take about 3 minutes each to run the PM file across each instrument so the complete run should be finished at about 10:30 to 11.00 my time. Not a bad effort considering (a 1 to 2 hour exercise). Obviously the processing time is dependent on your CPU but many will have a gruntier PC than mine....but try to achieve this speed of portfolio testing in the MT4 environment across a diversified portfolio. This level of processing used to take me a full day to 2 days to achieve within the MT4 environment...and that is without the bells and whistles of the outputs delivered by these models.

While it is not as speedy as Strategy Quant, TradingBlox, TradeStation, Amibroker or Adaptrade.....it is fast enough to pull out some heavy duty testing.....plus you will feel like a master of the universe as it will truly be your creation.....and you can then take it and further develop tests beyond what is available with these off the shelf solutions.

I will post the output results once completed and let you know the actual time for processing.

Cheers

C

PS Down the track a bit we will add a further module that allows us to compound results (reinvest profits) and also blend systems to produce some pretty impressive equity curves utilising system diversification principles.

PPS If you have run a test on a strategy and wish to change the variables and run it again, you do not need to recreate the portfolio. You simply adjust the settings on the PM file and run it again (after you have copied your files of your previous run to another directory location for future refence and portfolio system blending).

Eeeeek......It took an hour to process (much faster that anticipated) but I had assigned a 2% risk per trade....... The resulting equity curve and associated drawdown gave me heart burn.

I am running again on 0.25% trade risk or $500 on $200,000 trade capital. Fortunately we can bypass the portfolio creation phase and simply run the PM 'update' function again. It will take about 1 min to process each instrument.

While I am waiting, here is a handy resource for Excel developers.....'TradeInformed.com'. It contains numerous simple back-tests models and You Tube demonstrations that trial different systems and provide comprehensive details for each strategy. I frequently use this resource for new ideas to test in the Excel environmemt.

Below is a useful You Tube presentation on how to use Google sheets to obtain live data for a range of markets.

Here we go AJS

The results of your strategy applied to a universe of 45 markets across asset classes. Of the 45 markets, 26 were profitable contributors. CAGR is 4.61% with a max draw of 20%. The Pwin of 19.3% is very low when compared to other Donchian breakouts. I would suggest we need to increase the ATR multiplier to smooth returns as it appears we are being nailed by a tight initial stop. By increasing the ATR multiplier we will give a bit more room to breath.

The good thing about testing across a very wide universe is that it avoids any tendency to curve fit. With these simple stategies, the aim is to get a robust general performer. We can then improve our performance metrics by system diversification to increase our ability to perform in a range of market conditions.



I have included a run below that has altered the ATR multiplier to 3 just to observe the impact of giving the trades a bit more room to breathe and (as trade size is determined by volatility adjusted position sizing) the increased ATR multiplier also reduces the average trade position size.

I have taken a 'snap' of the prior run to use for comparison purposes against this amended variation.

Below are the results of the amended run. Despite the improved Pwin% from increasing the ATR multiplier, the overall performance results are worse (eg. CAGR/Drawdown ratio, profit factor ). The impact of a tight ATR multiplier therefore improved performance returns, as a few of the successful trades significantly benefit from the enhanced position size that results from the tight initial stop.

However......this is the important bit...... despite the better performance of the tighter ATR multiplier, the higher Pwin% of the amended variant (below) is what we are after in terms of it's contribution to overall robustness of a composite system portfolio. Why?.....because the improved performance from the tight ATR multiplier is attributed to only a handful of trades..with the vast majority of trades being losers. In terms of it's potential impact in a composite portfolio, this attribute has the potential to contibute to higher drawdowns for the overall portfolio.



The thing to remember with your testing is that you need to preserve the results of each run as the unique characteristics of each equity curve provide information that we use when we blend systems into robust solutions that are applicable to a range of market conditions. As a general principle, provided we have an edge in our individual system contribution, we are less concerned with individual system contribution to the overall portfolio and more concerned with the correlation between this system and the balance of the portfolio including it's unique characteristics that will contrbute to overall portfolio robustness. As we build our arsenal of equity curves, we use them later in our composite system solutions. Remember to use that 'snapshot' function to retain the results of your run and refer to them later.

Anyway, let me know if you have been successful in your efforts to broaden your universe. Cheers C

All I can say is "WOW" I was just trying to add another 5 forex pairs.

Stick with it ajs. After we have got into the the spreadsheet contents and you get a feel for it, you will realise that in this free creative environment, you have total freedom in where you take your models. It is quite a liberating experience weaning yourself off the broker sponsored or 'pricey' off-the-shelf solutions.

I will hopefully start getting into more of the detail this weekend.

C

Thanks I will Copernicus. One question, do you use your Excel Project to actually place trades in MT4?

Not yet ajs but am planning to link this backtester to a live trader through Excel like the demonstration in the You tube presentation on post 23. It will be a future project that we might be able to develop on this thread if we get sufficient interest from members with skills in VB and API interfaces.

One of the projects I hopefully will be introducing at a later date in this thread is a few projects in the MS Access or SQL environment (database solutions as a back-end to the testing models) that can host the data and results of this testing in the Excel environment in a seamless manner through further automation. That way we can really get some good stuff from the testing environment and be able to retreive and re-utilise our research that will assist us in live trading.

Copernicus, Did you use the "RLF InstrumentDump v3.ex4' for the original Instrument details? I have used it and updated the worksheet and ran a test using different parameters and I keep having failures. See screenshot. I have also uploaded a comparison of brokers cost / swaps. Do you think this is the reason for the failures?

Yep

The problem is with the SWAP calculation. Look at the huge cost of SWAP in your results of ($2.8M). The rest seems to be working well. So close mate. Well done. You are well ahead in the learning curve already.

Try finding which instruments are producing the distorted SWAP results. To find them look at the instrument spreadsheets created by the TM file and find those where the average loss is significantly different from $500 as a max loss (recognising that exhange rates) may vary this amount +/- $500. If you see some that have exceptional max losses....you will know they are not calculating properly. Otherwise you should be able to eyeball the culprits if you are able to see the individual instrument contributions on the right hands side of the portfolio equity curve. If all your instruments are not being displayed, if you understand formulas, you might be able to extend the formulas across the additional instruments if you follow the first few examples already provided.

I can't see the entire screen shot and identify those that have distorted contributions to the portfolio but would suggest it may be the instruments such as J225 or WS30 or worse still, every forex pair as the SWAP setting for these might have to be 0.

If you want to include those 'funky instruments' and try again, set the 'for SWAP' collumn to 0 for the forex pairs such as AUDCAD, AUDCHF etc. (look at the comparison you have made and use the SWAP settings if you can match them).

For example for those instruments with whacky results set the SWAP to 0 and see if that works. If not, then your broker (Darwinex) may be using a different method for calculating it. The 1 or 0 in the SWAP collumn provides two different methods of calculating SWAP. The 0 option is the usual method applied to forex pairs. The 1 option is the usual methods for CFD's such as indexes, commodities etc.

Below is a screen shot of the collumn and possible settings to use for Darwinex.



Let me know how it goes.

C

PS You won't need to create the portfolio again. Just adjust on the PM file instrument details worksheet and then go straight to the update function of the portfolio. The changes will be applied as it updates the instrument files.

PPS If you have trouble and you want me to fix the formulas for the rest of your portfolio on the equity curve instrument contribution list so you can see the entire list of contributors, then send me your PM file (as an attachment to your post or via messages) and I could fix it quickly. The model still works without them but it is useful to see the contribution from each. I will demonstrate how this is done for everyone a bit later.

Yes it's working now. Changed the 1 to a 0 on the worksheet. Thanks Copernicus.

Guys

I am sorry to say goodbye but inevitably something has got to give. I thought that I might have some spare time to get back into the forum....but unfortunately there are more pressing matters to attend to.

If the situation changes down the track, I might be back but in a far less involving manner.

Thanks for all the friendships gained along the way.

All the best with your trading endeavours.

Cheers

C

PS Guys, please don't be offended with the unsubcriptions. I just need to free myself from email notifications.