With the additions of BAMM & GDP from today's Hotlist, our Final Universe List now stands at 40 companies. - Spydertrader
ALRIGHT SVRZ! Let's discuss and see if we can get some NAILING down going here. It's been a huge & superior effort that this thread is making as a collective and I applaud and commend the contributors of this thread. Let's see if we can squeeze some more refinement for the task at hand. Underneath all of this lies a primitive transition function (state machine) which will provide even furthur comfort to answer definitively where we are now, and anticipation options for what's next. Before pulling apart nuts and bolts, quick question about the MACD you have charted. What kind of bias exists between EMA(MACD(5,13),6) vs SMA(MACD(5,13),6). I understand the exponential averaging aspect, but mathematically, is it more advantageous to exponentially as opposed to simply amplify the difference between EMA(Price, 5) and EMA(Price, 13)? I'm finding the SMA to be smoother than the EMA with less false signals on the derivative of this indicator curve as a proxy for price. Using the indicator as a proxy for the price curve enables the hist to be the first derivative (price velocity). Fundamentally, the XO of the hist to positive should in application dictate positive price velocity (our long bias... for now). Although for my race mindset, I am hypersensitive to the actual as opposed to approximated magnitude of the price velocity, I will settle for this nugget for now. Some visuals for the math jargon despisers. Check out the SINA chart. Again, I just want to discuss. This is why I always end my postings with "Newb", since I am not in any way or form an authority on any of this. Just wanted point out any item of interest that pops out when the visuals and math line up... http://www.elitetrader.com/vb/attachment.php?s=&postid=626496 On your SINA chart, I noticed the solid red line on your EMA(MACD(5,13),6) and then noted it's relationship/movement with the actual price bars. I am hoping that it too looks interesting to you... Then consider the difference between the solid red and blue lines of the EMA(MACD(5,13), 6) in terms of a derivative. All being equal then, the regression on the indicator should provide a briefly reliable indication of an upcoming transition into either price congestion and/or trend reversal I would presume... The slope of this regresion line should mimic this differnce stated above since the regression reduces to the tangential to the price curve... Any thoughts? G33M4K the Newb
Hi G33M4K Thanks for the excellent post. Yes, this was where I was going. I think it was you who posted a link to a newsgroup posting which discussed MACD in depth. This discussion made some interesting points in regards to the slope of the MACD and signal lines. I thought that if we could obtain the slope of the signal line, we could determine congestion or reversal by simply checking if this slope is positive or negative. Unfortunately it seems that nothing in financial time series analysis is that simple. WL has a function called LinearRegressionSlope. The arguments to this function are series (which is simply a time series) and period. I plugged in our signal line and a periods of 5-13. There was nothing meaningful I could decipher from the resulting graph. Take care
What you have is right there Svrz. It's evident that you understand the relationship we sought for. I don't want to lead anyone astray so I'll get specific. It's there because I see it and hopefully you will see it too. The two graphs you have below the chart plot is a suitable space for the time being, so I'll recommend some adjusting so they tell you what we have been striving to see... Briefly cast aside the regression stuff for P[t]. It will be needed it for money flow for reasons I'm still working on. The slope we desire is already before you. Keep in mind derivative flags f'(x)=0 w.r.t. f(x). Reduce your macd plot to only the MA of MACD plot and if possible use the SMA of MACD rather than EMA and note extremely well it's parallel movement to price (NO LAG... LOL). Below it, on your second graph, plot the difference between your MACD and MA of MACD. Flag, the zero line XO (f'(x)=0) of this bottom plot and note the relative min/max of the MA of MACD. Voila! Derivative (simultaneously, the function of regression slopes)... Do you see it? G33M4K the Newb
Our system failed to generate any signals prior to 11:00 AM. However, the system DID generate two signals shortly after the 11:00 am time frame. Both BCSI & NAVR had actual volume exceed their calculated dry up volumes before 11:15 AM. MACD and The stochastic indicator for both these stocks fell within normal parameters. We see price improvement for both stocks, and price for each, when their respective volumes exceeded dry up, were: BCSI ($19.90) & NAVR ($16.26). I did not take either of these two trades as they did not appear prior to 11:00 AM. However, I have read in Jack's writings where he has used 11:30 as a deadline for dry up in the past. Since we seek FRV by end of day as our confirmation of remaining in a trade, I do not think extending the 11:00 AM time frame to 11:30 would result in additional risk exposure. As a result, taking on trades that signal relatively close to the 11:00 AM time frame is a matter of personal preference. Your mileage may vary of course. I hope you find the above information useful. - Spydertrader
G33M4K Very interesting, indeed. I followed your instructions. Take a look a this picture that I've attached. Read my next post as I will attach the performance of a buy and sell system based on this very simple indicator.
Here is the performance of the following system: var Bar : integer; var ma1 : integer = EMASeries(#close, 5); var ma2 : integer = EMASeries(#close, 13); var mcd : integer = SubtractSeries(ma1, ma2); var signal : integer = EMASeries(mcd, 6); var pane1 : integer = CreatePane(60, false, true); PlotSeriesLabel(signal, pane1, #red, #thick , 'MACD Signal (5,3,16'); var diff : integer = SubtractSeries(mcd, signal); var pane2 : integer = CreatePane(60, false, true); PlotSeriesLabel(diff, pane2, #blue, #thick, 'MACD - Signal'); DrawHorzLine(0, pane2, #black, #thick); for Bar := 20 to BarCount - 1 do begin if not LastPositionActive then begin if (@diff[Bar] > 0 ) then BuyAtMarket(Bar + 1, 'Long') else if (@diff[Bar] < 0 ) then ShortAtMarket(Bar + 1, 'Short'); end else begin if ((@diff[Bar] > 0) and LastShortPositionActive) then begin CoverAtMarket(Bar + 1, LastPosition, 'Cover'); BuyAtMarket(Bar + 1, 'Buy'); end else if ((@diff[Bar] < 0) and LastLongPositionActive) then begin SellAtMarket(Bar + 1, LastPosition, 'Sell'); ShortAtMarket(Bar + 1, 'Short'); end end; end; Honestly, it does not mean much. Until we do a sample back test, out of sample back test, Monte Carlo Simulation and then forward test this system, any acquired result is meaningless. However, with that in mind, this looks somewhat interesting. It may have potential. Although our system did not lose money, it peformed very badly compared to buy and hold. This is simply due to the period tested being in a fairly strong bull market. This is also verified by the fact that short side of this system performed very badly. If we were to take the long signals only, we would have done fairly well. The problem that we have is the age old problem of whipshaws. This is perhaps the number one achilles hill of any mechanical trading system. Discuss.
Two things to note, I see that your graph is intraday, this is where the limitations of my Excel sheet fail. I've only been refining on EOD bars and use the currently developing EOD of day bars. Basically, at any point of the day, my current bar is a single incomplete EOD bar. I assume there are some transient effects when changing to faster fractals. ALSO, NOTE VERY WELL, the relative extremas of velocity that are close as opposed to far from zero have different aticipatory options as opposed to those which are far from zero. Personally, I have EOD temperment. Also, price is only 1/2 of the cruder PV relationship. I don't want to in anyway or form suggest this is a trading system, just a tool to reduce some of the anticipation subjectivity. Positive price velocity means positive price displacement. JH's equity approach is long biased, hence positive price velocity. Compounding is biased to positive acceleration. Conditionally "AND" these two and contract the duration and this will have a significant affect on returns. I haven't worked out acceleration yet but the math should be similar. Volume will be a whole other ORDEAL to identify congestion!!! Y? Volume BO's during congestion precede a new trend. It cannot furthur congestion. This has to be. Don't even know where to start with that.... A/D... yeah . YIKES! G33M4K the Newb PS. One other question, is AA in our Universe (ie. does it meet our culling constraints)?