Simple Shelley Backtester

[garyfritz]

Yeah. Unfortunately adding the backtester functionality to the EA was NOT a simple process, and I hate to repeat the process on regular Shelley. I'll be working on designing a library interface to make it easy(er) to convert any EA to a backtester, but for now it's all by hand.

I can port the backtester mods into regular Shelley if I have to. I can diff Simple Shelley 1t against my backtest indicator, and see exactly what changed. I really tried to minimize and isolate the changes so that works well. It would still be a bit of a job and I'm really short on time this week.

[Colt45]

Just loaded it up and this is what I got.

I believe one of the previous versions of simple shelley calculated the 4hr woh along with the daily slope buy/sells which is what I believe is what you have posted. The new version calculates the daily woh with the daily slope.....correct me if i'm wrong

[jboy]

Gary,

Does this indicator back tester account for the spread?

[garyfritz]

Currently no. It just uses the close value. Since it is limited to the close value of the *candle* -- and currently I'm testing with M15 -- the spread is a small issue. I will see if it's possible to do better once I get a chance to work on this more and design Steve a backtester interface.

[Verto]

I am very interested in the time of day test results.

Even though the time of day seems like a blunt instrument it works rather well for one. I have been opening a full position with 20 pip tp and 30 sl, just for sake of simplicity with test results. I close half at 20 tp and move the other half to BE trailing the pair until it falls from the list. This seems like it would work well to catch a good long trend.

You only need to catch a good long trend on a few pairs to really make up for the RR on closing out the half. So it seems. I have been forward testing this for the past week and half. Right as the trend changes, but still am in positive territory! Grr, my luck.

Edit: To add as to why I like the idea of Simple Shelley it is because it is a "dumb" EA. It looks for the trend and enters. There are not many working parts and I like that. I understand it and I believe that there is less curve fitting (besides time of day).

Naturally we all want to 'improve' our entries using many different indicators or methods - such as retrace, TMA band. That is fine! I, personally, just don't have confidence in my ability to 'time' the markets in this way.

[garyfritz]

I am very interested in the time of day test results.

Remember I said these are VERY PRELIMINARY results. I found some issues in the tester implementation that change the time-of-day results. I've hit some issues (trading and non-trading) that have prevented me from posting updates, but I hope to post something this week.

[Verto]

I am very interested in the time of day test results.

Remember I said these are VERY PRELIMINARY results. I found some issues in the tester implementation that change the time-of-day results. I've hit some issues (trading and non-trading) that have prevented me from posting updates, but I hope to post something this week.

Understood. I wasn't taking the results as set in stone - that is for sure. I never do when it comes to backtesting. Take care of life stuff before trade stuff Gary. That always comes first! We do one to support the other, not the other way around.

[f451]

I am very interested in the time of day test results.

Remember I said these are VERY PRELIMINARY results. I found some issues in the tester implementation that change the time-of-day results. I've hit some issues (trading and non-trading) that have prevented me from posting updates, but I hope to post something this week.

Gary,

I just want to say thanks for the work on this. Empty4's inability to test multi-pairs is a serious drawback, and it looks like you've put together a framework to provide useful backesting info for these kinds of scenarios.

I'm a chronic lurker but as Steve can vouch, I am battle scarred veteran - some of my recent Forex trading work has been in manually (demo) trading Steve Mauro's Market Maker Method which has some things to say about market sessions and market behaviour at these times. While overall MMM is probably long on marketing hype and short on solidity, I have added a couple of useful insights around sessions and market timing to my own trading because of it and I am have (non-rigorously) tested these which some success. I'll list the ones I feel are relevant to NB 10.2 / Shelley's basket trading approach:

1/ Sessions and time do have an influence on market direction and behavior - but we all knew that right?

2/ NY equities open (9:30 NY time) is a known time for reversal of the current trade bias - this one I have annecdotally tested - it is by far and away the most reliable trade setup of the MMM approach - known as the NY reversal - it usually good for about 40-50 pips

3/ semi regularly there is a "head fake" shortly before the end or or shortly after the end of the asian session. Some traders call this the Frankie fakeout - it often coincides with the open of the Frankfurt market one hour before London open. This is a move - usually around 40-50 pips in the opposite direction to the subsequent move in London.

4/ rarely, although reliably when it occurs, the high or low of the trading day is made just after the end of the asian session - this is where there is a push to a new high or low (frequently to a weekly or monthly pivot BTW), and then price reverses for a large move. (see 3 above). When this happens, it is usually the begins of a major move

5/ trend direction often reverses mid-week - monday, tuesday one direction, wednesday or thursday reversal to the other direction, friday back to the range is the most common pattern

6/ frequently price moves back towards the 50, 100 or 200 moving average to end the NY day's trading, often ending up having traveled quite far across the day, but not so far away from the longer term moving averages when examined on a day to day basis

7/ unfortunately the session based behavior, while it does seem to have distinct patterns - these patterns seem to shift around - for example while I was trading MMM, major USD moves started mid asian session three weeks in row, then for the next three weeks shifted to mid london. I think your results show this quite clearly. The common belief is that the major trading banks are now global outfits and they co-ordinate their moves - handing off the buying or selling to a particular country desk for initiation. Then they shift the pattern around to make sure no-one can anticipate their plans.

How does all this assist in finding optimum times for closing Shelley baskets? Well, I think it gives us a few directions we can explore in testing.

Here are some ideas:

Firstly NY open. Open positive trades are likely to move against you at this point, and so it may worth testing to see if a good place to close out profitable baskets is just prior to NY (equities) open, e.g. just before NY session open, or just before 9:30am NY time (equities open)

If NY as a time to close out the trade is unprofitable, then perhaps this is a time to initiate SRF or reversal scalping instead (flip trade direction on the negative results to see if this gives you winners)

analysis of the time at which baskets are in maximum profit across the sessions might lead to interesting and tradable results

segment the pairs by major currency - USD, JPY, GBP, CHF, CAD etc - and do the time analysis on this basis - they are typically traded by differently by different traders in different sessions, and have different rhythms

some pairs have definite seasonal patterns, particularly the commodity pairs (AUD, CAD etc) and so we might find a that rather than closing baskets at the same time every month of the year, shifting to a pattern of time for closure depending on the month is more profitable...

it may be worth segmenting strongly trending pairs from everything else to prevent premature closing of runaway winners.

cheers
Simon

[Jemook]

Cheers Simon, interesting insights!

[garyfritz]

Yes, that's the kind of thing I'd like to be able to backtest. Unfortunately my situation right now means I can barely keep up with the forum posts, let alone do any coding or design. I don't know when I'm going to be able to get around to creating this indicator-backtester framework.