AET Short R Long Chart 3 .......... experimental trade Layer 2 added <img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=3836047> Vertical lines at right ends of plots are spread size of 1st layer.
Chart 3 RAI Short PM Long Illustration of trading layers. Chart posted while trading in session today. <img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=3840560>
Do you trade LO RAI? The correlation has been more intact lately. I entered it on 6-20 and out 6-24. The pair converged ~2.5% from it's 6-21 high. The attached chart illustrates the 7 level trade. PG
<img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=3840848> I like the pair. Thank you for sharing. I suspect you are oriented short term in trades ? I swing trade over days and weeks to take additional advantage of chart history where there is existing, obviously good probabilities of success. I would have stayed in this trade just a little longer, considering the history of RAI to be the top plot. Also, as illustrated in your link, I am wondering why so many layers were installed so quickly. It appears there is trigger happiness. Of course it worked out, but would you not have considered while in the trade there being possibility of the spread getting wider than it fortunately turned out to be as you are installing layers and running out of powder prematurely ?
Iris is designed to trade off standard deviations from the historical mean of two stocks, the pair. I had RAI LO customized to enter it's first level at approximately 2.30 STD...which puts it in the top 20%+ of it's normal trading range (from a standard deviation perspective). The pair was entered based on probability and statistics. There wasn't any "trigger happiness", all of the entries were automated and executed at defined STD increments from the initial entry point. I entered 7 of 20 available levels, so I still had plenty of "powder" available. If the pair were to reach it's maximum 20 levels, that would be it's extreme standard deviation over the compiled data's time frame. At that point if it continued to diverge I would not allocate any more capital, as something was awry with the relationship between the stocks. The pair's auto exit settings would have already been initiated and it would stop out if triggered, due to either max standard deviation exceeded or simply max % loss exceeded. My average hold time varies, but typically it's under two weeks. This is primarily because I'm entering pairs only when they reach their standard deviation extremes--and a convergence in the very near term is expected. I exited the pair because the STD reverted back to zero, which is again how Iris is designed to trade. Put in another way, when RAI LO "crossed" on the chart you provided, that by definition gives an STD value of zero and is a full mean reversion. Another reason for the brief hold time is cost of capital. I don't like 'dead money'. If I'm entering a pair b/c I've determined that it's very near to a time of mean reversion, than that should happen soon (within a couple weeks), or my capital should be allocated elsewhere to more potentially profitable opportunities. At the time RAI LO was 1 of 30-40 pairs that I had open and I was monitoring an additional ~400 so there is always a place to allocate the capital. PG
<img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=3841165> There are two basic elements in your explanation IrisPairsGuy. (Reference your post immediately before this one.) First is long and revered by pairs traders which is the use of standard deviation. I have two questions about that as per YOUR identification of SD in the case of pair RAI LO. Second is style. Your average time is two weeks to hold. For myself I happily am willing to apply more time which is where I find my greater booked gains are derived associated with intelligently applied timing of layering. The gains are so meaningful they easily compensate for what short term traders label arguably "dead money." But these different holding times between the two of us are because of our different approaches which are huge. The hackneyed phrase, "Whatever works for you" is appropriate here not just because results are satisfactory for the respective trader, but because of what ever subjective assessment a trader judges to be en enjoyable level of success vs the risk. Regarding standard deviation ( SD ) in this case of pair RAI and LO........ You stated your approach is based on the historical mean of the two stocks determining SD. You also identified the crossing of the two plots as being where SD is zero. (1) Am I correct from your view that looking on a chart, should two stock plots for a given time period be perfectly superimposed on each other the entire time, the entire term of that chart would have this pair as zero SD ? (2) If the answer is "yes" to the above question, looking at this chart (a historical chart) that has RAI traditionally above LO, would not the SD being zero be somewhere between the two plots, not their crossing ? It is the last question that drives MY trading such that I would have believed the rubber band is still stretching within the context of the traditional spread at the crossing of the plots such that there is more gain to be had by holding past your exit point at that plot crossing. I easily would have held on to the trade past your exit until it approached approximately where the spread occurred a few days past your exit (zero spread) and where the spread was continuing to re-develop in my favor the last week of June. That exit (my exit) would have been where the spread would have appeared more like that of periods Apr 15-22 and Jun 10. My exit would not have been swinging for the fences aspiring to a spread as it was in mid-May, but instead aspiring to the reasonable expectation of a spread similar to the dates I cited.
(1) Am I correct from your view that looking on a chart, should two stock plots for a given time period be perfectly superimposed on each other the entire time, the entire term of that chart would have this pair as zero SD ? (2) If the answer is "yes" to the above question, looking at this chart (a historical chart) that has RAI traditionally above LO, would not the SD being zero be somewhere between the two plots, not their crossing ? Very good observation deucy...and yes your are correct in #1. My lookback period (set of data points) is slightly different than yours. With that said for this discussion I'll adjust my lookback period to 80 days, or 4 months (to match your chart example), so we are comparing apples to apples. With an 80 day lookback RAI LO hit zero STD around 1:30 EST on June 25th. So you are absolutely correct in that zero was not reached at the crossing point on your chart...rather the crossing point of the historical price ratio mean (which is just another way of saying the average price ratio over your lookback period). FYI...I actually exited RAI LO at .44 STD so prior to full mean reversion. That is my style. My auto exit settings will always close the pair prior to zero STD b/c I've found that there is a lot of 'noise' around zero STD. Again, nice catch on seeing that at the crossing point of the two stocks on your chart example was not zero STD...it was a bit later the pair reached full reversion. Not to sound pitchy, but you should check out our software...I can clearly see how it would benefit your trading. Just looking at your few examples I can easily model an automated system around it. Although you are much longer term and less active, Iris is able to monitor and manage as many pairs as you can input...far exceeding human capabilities. I traded pairs manually for 8 years before I co-founded a fund out east, at which point I turned to full automation. After seeing much success from the automated pairs trading approach, a few years ago I decided to create a software company to develop a retail version of our pairs black box...and Iris was born. Iris can be traded in full automation, semi-auto (auto entry only or auto exit only), or manually...obviously I prefer full automation. PG
Deucy, How do you calculate SL for these type of trades that continue to diverge. I assume you don't just keep adding layers.
A favorite question of mine I receive from those who appear I am taking on much risk. The answer is at the heart of why I am so passionate about MY STYLE of pairs trading. I expect questions when I post on this thread. I prefer to answer at another thread, quoting your question. The link below will take you to my answer at Pair Trading Strategy Journal. But first, let me point to the coincidental issue of adding layers from another trader's style. He added 7 layers in the example he cited ! Just scroll up to IRIS PAIRS GUY post of July 13. Other than his trade reaching pre-set standard deviations of the pair's spead which triggers for him an added layer, I know nothing more about his capital management or why he chooses those SD's or how much capital he adds with each layer or when he would stop out. He does explain why he would stop out, as general as it is. My answer to your posted question.... http://www.elitetrader.com/vb/showthread.php?s=&postid=3842038#post3842038
Chart 4 RAI Short PM Long Illustration of earlier spreads similar or larger in size to current one. <img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=3843714>