This is my interpretation: From the paper: "The basic concept of model MACDR2 is the same as in model MACDR1. However, the buy or sell signal is given, if the difference between the moving averages is bigger or equal than a certain percentage of the stock price at the end of the third day after a crossing." There is no profit target. This is a trend following system. This is not MACD. It is a moving average crossover system with a three day delay parameter. "If (EMA1 - EMA2) > 2 % of the closing stock price three days later then buy" might be a rule. The authors might obtain similar results by using longer term moving averages instead of the three day delay. The system trades long or short as the signals appear.
Hi Nuke, I started daytrading a couple of months ago and consider myself a newbie. I'd love MACD and use it extentively along with candlesticks, but have not yet been doing well . I happen to find this thread and the article of the "refinded MACD Indicator...." which are of great interest. I went through all 17 pages, but don't completely understand both the MACDR1 & MACDR2. Since you have extentive research into this refined method and may answers for my questions as follows: 1. MACDR1: do we use the same 26-12-9 as MACD1 with a 3% or 5% target after the crossover? Often time whipsaws are created after the cross, leading unneccessary losing trades. How do you address the pitfall? 2. MACDR2: on page 13 of the article it saids the "best result is achieved using the moving average combination 73, 34 & 25". Does it mean replacing with the 26, 12 & 9? 3. MACDR2: 1.5% crossing level for MACDR2. How to calculate it? As with ironfit, I am very interested in your stock version of the m-macdr and would like to know if you can also send it to me at neoncity@gmail.com. Many thanks in advance. HAPPY HOLIDAYS! Collin,
what have you been using it on because this matters.. in equities forex.. etc. MACDR1 is a revision of the MACD1 or the original MACD the macd doesnt have the crossing level built in its macd1 the original cross wait 3 bars if trend is still intact enter withdraw at TP this was a large increase i think up to the 50's% success. macdR2 is a revision of the MACDR1 and basically is macd line - signal line / price @ close. *100 this gives you a % 1% 2% 3% 4% 5% so you would enter on a predefined % level 1% being less accurate but having more trades available 2% being more accurate but less trades than 1% and so forth with accuracy increasing but number of trades decreasing. and you exit at a statistical TP point like always this was macdr2 and since then ive modified it for forex and its a beauty and also built a macdr3 which has almost the same setup as macdr2 but adds confluence and thus isnt ton more accurate but does provide on average bigger returns macdr1 is simply on cross wait 3 bars enter in direction of trend if still in place exit on recross or tp point its 34,73,25 in that order yes you replace 12,26,9 with 34,73,25 well quickly calculating you take the simple formula macd-signal line / price *100 so lets say macd reads 3 and signal reads 1.5 and price of the equity is 100 then ((3-1.5)/100)*100 = 1.5% i have to dig through mystuff i know i wrote the mmacdr2 for equities somewhere but if it doesnt involve FX market i just file it away so let me look. its pretty simple to code if i cant find it ill just recode it. you to
Hi Nuke, thank you so much for sharing your work in this thread. It's been quite helpful and early tests show some promise but now I'm a bit stuck. Could you shed some light on your correlation coefficient, please? I'm not really a maths guy. What is a correlation coefficient?
Anyone still working on this? I just found this thread and yesterday I spent several hours on it. I have coded it up and tested with equities and the S&P 500. In both cases I find that it's barely breakeven. I can't find any edge. With the S&P500 futures using daily data, it's 56% win rate with 0.80 avg win/loss. Equity curve oscillates like an MACD. If anyone has any ideas I'd love to give this some more time. It's an interesting idea, I'm just a bit frustrated that I can't make it work.
trading the market is like competing in a triathlon... the edge is not in a single discipline... you don't have to be the best of everything, but you have to have all angles covered.
yes this works when setup properly but its not a holy grail remember everything always changes.... i know i shouldnt give this away but if your looking for a way to trade .... price action is the way to go... if your trying to automate .... there is better than this while it tends to give you good entrance and exits and can be profitable .. the triple histogram will help you read the market better.
I am thankful for this thread, since I was trying to implement MACD for myself since there is a lot of documentation on it. The paper addresses two problems I was experiencing: waiting all the way through the profit window until convergence kills everything, and noise on the entry condition that would cause basically false triggers. I happened to already express the divergence as a percentage of the total value so that I could compare signals from multiple equities, so I had that much. I tried to sloppily clobber together the additional rules this morning in about 20 minutes, and wound up with an indicator that never triggered. I am sure it is a bug in my code and will fix that tonight when I can give it the proper attention. NukeTheWhales--if you're still following along after so much time--what do you mean by confluence? I don't know what people mean with some of these terms, and this one I couldn't infer from looking it up real quick.
real money trial or simulated? ok, probably, simulated, ok, then will give you a simulated answer... good job! good results! just flip coin and do exactly opposite and reverse your results!
I think that was part of the point of the original paper picking on MACD. Given how poor it performs if you execute it based just on the crossings, then it might as well be random. They dressed up the paper by showing a strategy that goes beyond the noise. IMO they didn't do too good of a job at it, since they didn't really show, say, normal distributions of their trades against random ones, or do anything like monte carlo or bootstrapping. They added some rules that makes the indicator smarter. It's the kind of thing one generally should do, but I hadn't seen any real examples how people do this before, so it gave me an idea of what people try to do to refine an indicator into a viable strategy that factors in the personality of the indicator. That being said, I personally haven't been able to find success with the rules yet. I haven't finished though; next I will try to use a 5% limit order to achieve the 5% rule in MACD-R2. And I am also considering stops. The limit order is probably my best bet. However, there is still something about it that grinds me gears... Using the closing price in the transactions really grinds my gears. A lot of simulation is this infatuation with using information from the current close in order to trade on the current close. In my case I'm making my decisions in the evening after the market has closed. The next day, my first real trading opportunity then is the market open, and it's a whole different game. So I feel these strategies should trade on the open of the next bar. Either that or they assert they're putting in a limit order based on the close of the last bar, but somehow they need to address how they're going to get complete data for that close bar, crunch on it, turn around, and trade on it. The bar has closed; that opportunity is gone. I should probably start a thread and rant about it, since maybe there's something I don't understand; maybe you really can market on the close after having all that information. But it seems like a temporal contradiction.