Judging by the thundering silents it seems that almost none of you realized the significance of being able to assess time series fractal statistics. Once again it falls on the lowly market mole to elucidate and reveal you the error of your collective ways. As you all know – if necessary I will drag you crusty butts kicking and screaming into the 21st century. If you still think that drawing lines on a chart gives you a competitive advantage then 1991 called -they want their MOTORAZR back. Sometimes I wonder why I bother…
Anyway, before we get to all that a quick visit to the equities side – not looking so hot over there and it seems the spoos are going to take a b-line to the 100-day SMA. Today’s failure at the 25-hour SMA pretty much sealed the deal here. I don’t see a good entry opportunity except maybe if we see a bit of a bounce near the EOD:
The NQ has been pretty stubborn today and that’s why I would take a long on a breach of today’s highs tomorrow. It still enjoys support from below via the 25-day SMA and a daily NLSL at 4019 and a quarter. Let’s see if it can close the session above and then we talk. Until that happens I’m on stand-by mode on all things equities.
Alright, now let’s talk fractals, folks. So, why should you care about them? Well for one they can be pretty – see Mandelbrot et al. And they can be profitable - if harvested properly they offer us statistical edge – very simple. And that in turn leads to profits – ka-ching!
In order to prove it to you guys and myself of course I coded up some basic reporting based on the Fractal Monger code I put together last weekend. It was rather basic – all I needed is to parse through my list of 50,000 candles and sort all fractals into a sorted list. Each fractal entry contains is higher, lower, equal, and of course total closes. From that I can do some sorting and in the process extract the respective ratios – i.e. higher vs. lower closes and the inverse. Let’s take a look at what I found starting with the 60 minute ES futures:
Shown here is are the top scoring higher closing fractals – my cut off threshold is 1.5, meaning I want to see 3 higher closes for every 2 lower closes. That btw will be my filtering threshold for the Fractal Monger notification module starting next week. The fractal on the very top is a pretty sweet one and in the past eight years (50,000 hourly candles leads us back to roughly to September 2006) it has only stuck 46 times (total closes). It closed higher 33 times and lower 13 times, giving us a ratio of 2.54 – I take those odds every time.
As you scan down the list the ratio slowly drops toward 1.5. In total we would have had 1665 entries in the past eight years – that’s roughly 420 weeks and that’s almost four per week! That actually struck me by surprise as this is only one direction on one symbol. Thus we may be able to scale it back to a ratio of 1.65 or higher.
And here’s the lower equivalent. This time we are rating short fractals – meaning we would want to go short when they trigger. Top scoring has stuck 67 times – not bad and the total on the short side is 1622 – almost the same as the long side. So that puts as at 8 hourly fractal patterns per week for just one lousy symbol. Are you paying attention yet?
Now let’s look at the EUR/USD – here we have a top scoring higher close fractal with a ratio of 3 – it’s a rare beast but when it hits I will be there to either go long or grab myself some binary options for a bet higher. The total number of high scoring fractals comes out to 1739 – almost 4.2 per week.
The short side is a bit weaker and we’re counting 1209 – about three per week. So if you count those two symbols together we are at about 15 entries per week on average. Obviously there is no reason to not run Fractal Monger on at least 20 symbols out there. This would probably allow us to crank up our ratio filter to around 1.7 or higher.
Now let’s take a look at one that was triggered on the ES futures during a few hours of testing the messaging system today. This is actually how the email looked like as I managed to get that HTML format working.
The current fractal has occurred 62 times in the past.
It ranks as the 164th most frequent fractal during the past 50003 bars.
The next candle has closed higher 37 times, lower 23 times, and equal 2 times.
Frequency of occurrence: 0.12%
Higher/lower ratio: 1.6
Lower/higher ratio: 0.6
Pretty sweet, isn’t it? The higher ratio was 1.6 and that would have been above my desired 1.5 ratio threshold. And it DID close higher. Just one example – let’s see how we flow. For now I want to see how many of those we get. Then next week I’ll turn on the ratio filter of 1.5 and add more pairs and other symbols. This should be fun!
By the way, I also switched to 5-digit fractals after some extended testing yesterday. I may also slightly modify the normalization algo to be a bit more precise..
Alright before I run – two more setups are waiting below for my intrepid subs: