| stevehopwoodforex.com https://www.stevehopwoodforex.com/phpBB3/ Print view |
|
| The purpose of this forum. Please read this sticky. https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=519 |
Page 2 of 2 |
| Author: | Jonathan2FI [ Fri Nov 30, 2012 2:38 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
Gary, Steve, I think I understand .....your aversion to Empty4 backtesting with people diving in on poor data, mismatched errors between TFs, kidding the system with unrelaistic spreads, curve fitting on limited trades and declaring a massive victory for their 5 min backtesting. ... with clean data , running scripts to update and eliminate miss matched errors (datacleaner data updater), looking for >50 trades over a significant period 3 - 12 months, running MTi4 spread controller and selecting realistic spread based on looking at spread hourly history charts, not over opimising to the nth degree and using VPS with limited latency then it does and can reflect fwd results. I take up the challenge The difficulty is with Empty4 you cannot see in the backtest result on a forum that someone has consistently applied this discipline and therefore you open the flood gates to postings of BT results pointing to results in 100 directions with no consistency. Have I missed something ? Jonathan2FI |
|
| Author: | garyfritz [ Fri Nov 30, 2012 4:16 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
I'd say that's a good start. I think 50 trades is an absolute bare minimum for any kind of statistical confidence. Considering that an EA often has multiple entry modes, multiple exit modes, etc, you really ought to have at least 30-50 trades for EACH of those cases to have good confidence. If someone posts backtest results on a forum, it's often hard to know how reliable their backtesting is, given Empty4's funky backtester. But you can't control what people do if they trade the EA and don't follow the rules. |
|
| Author: | Jonathan2FI [ Mon Dec 03, 2012 2:15 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
Gary, I extensively test and don't trade if I can not get the return with a low RDD etc as you indicate there is a skill to backtesting and many pitfalls including "the past does not equal the future". A couple of Things I have been playing about with that I would appreciate your response on 1- Trade 20 bounce where TMA >0.8 or <-0.8 no brainer 2. Trade variation of GDay based on different levels of Stoch OB OS and L/ S triggers depending on TMA level e.g strong TMA>0.8 requires weaker stoch retracement say 60 , weaker TMA <0.4 requires stoch >80 etc. 3.Using MA of equity curve on slave/mirror account to switch EA on and off on live account. Clearly the equity curve demonstrates if the market has the correct sentiment that the EA was designed for ....If its falling switch it off before it trashes your account ...if its rising happy days - switch on off appropriately automatically! I have simplified the above for brevity I am aware that strategy is much much more than entry! anyway 1. where would I find closest EAs from Steve for above all based on single pair . I have looked at Gday;trendy;bigmac 2. I assume you can not test any of the basket systems on Empty4 or ? Jonathan2FI |
|
| Author: | garyfritz [ Mon Dec 03, 2012 6:09 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
However the future generally resembles the past. If it didn't, trading would be a random activity. What's a 20 bounce? Bounce off a 20-bar MA or something? Remember that TMA (presumably "TMA True") levels do not guarantee market direction. The market can go up while TMA < 0, and go down when TMA > 0.8. Trading is very seldom a no brainer. Might help. I don't have a good feel for it. Sometimes it works. Generally it doesn't. It's functionally identical to trading an MA strategy, i.e. buy when market closes > MA, sell when market closes < MA. That works great when the market trends like crazy, but it kills you when the market goes sideways. Which it does, much of the time. See http://www.stevehopwoodforex.com/phpBB3 ... 780#p28780 for an explanation. I'm not aware of any EAs that are "closer" to your ideas than Gday &etc. No, the Empty4 backtester can't test basket systems. Single symbols only. |
|
| Author: | Jonathan2FI [ Mon Dec 03, 2012 7:34 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
Gary, Crystal clear I'll bring something fwd with more meat ref. TMA true levels & retracements with Stoch levels ....... visual testing is leathal to the biased eye looking for confirmation ....long live stats Jonathan |
|
| Author: | maahome [ Thu Oct 31, 2013 11:28 pm ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
Hi Guys, Very interesting thread! I have done some back testing of a particular EA I was interested in (commercial offering). The method I used was to setup an offline instance of Empty4 and used imported 1 minute data obtained from a colleague who works for a major hedge fund here in Australia (sorry can't disclose name or he would probably get shot!). We tested the EA for 8 years including the meltdowns in 2008 and the Japan Nuclear disaster. the EA actually survived but the drawdown in 2008 was pretty scary. Point of interest here, if we ran it on an Empty4 data feed it just blew up (regularly) because of all the data inaccuracies. Since then the developer has removed that version from use and replaced it with a newer version which I have not tested as yet. Anyway, the reason for the post here was to reiterate what is being said, unless you have a very controlled environment in Empty4 the results are far from accurate. you can run a back test on different time frames and get completely different results because Empty4 "guesses" the ticks in the data. Having said that, I believe a back test can give some indication of whether the EA has a chance or not. I would however be very careful of testing on straight feed data from any broker. If you do test this way - I suggest you look carefully for any possible data anomaly that may appear (and they will!) My friend tells me that if you were to offer a back test from Empty4 to a professional firm they would just roll around the floor laughing at you. Unless you have results that are compiled and obtained for example from say a Bloomberg terminal / feed then they just tell you to go away. Also any black box method /system has to be audited and again only results from a professional level platform are acceptable. Regards John |
|
| Author: | garyfritz [ Fri Nov 01, 2013 12:08 am ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
That's why I don't pay attention to any Empty4 backtest results. The Empty4 backtester is utter crap. Even with good data (hard to find) its results are generally total fantasy. Nobody should risk any money based on the Empty4 backtester. I use Tradestation for my testing. That's a backtester I can trust. I know it works right, and I know if TS says the system did X in the backtest, it would have done X trading it in realtime. (With a few well-understood exceptions like bid/ask differences, etc.) |
|
| Author: | macman [ Fri Nov 01, 2013 8:40 am ] |
| Post subject: | Re: The purpose of this forum. Please read this sticky. |
Thanks for posting this, it should become essential reading for everyone joining the Forum as a huge percentage of assumptions made by contributors seem to be based on terribly inaccuate Empty4 backtests. Best regards, |
|
| All times are UTC | Page 2 of 2 |
|
Powered by phpBB® Forum Software © phpBB Limited |
|