House cleaning
the first thing i need to get out of the way is i nixed the '60s psychedelic theme in the beta 2.0 spreadsheet release. i thought the color coded column names matching the color of it's corresponding chart line would be a nice usability feature. the next day i opened the file and said NOT. so i came up with a more aesthetically pleasing and professional color theme matching the color of the new company logo i created. that too will probably get updated but it will probably stay roughly the same color. i hope you like it.
Overview of the market
ok, this week's stock market performance was interesting to say the least. as i have posted previously, we have been in a trading range from Apr-9 when the S&P 500 closed at 856.56 to friday's eod close at 866.23. the intraday high in-between this period was 875 on April-17 and the intraday low was 823 on this past tuesday Apr-21. this suggests to me that buyers got cautious after the index has run up so high and so fast from the 666 low on Mar-9. and the bears have gotten their shorts squeezed so they had to cover early on but weren't willing to begin selling later due to the high volume of buying. the index could have gone either way this week and euphoria won out over pessimism.
on last night's show, jim cramer said it's the mutual funds getting back into the game, not the hedge funds, causing the buying frenzy. this may or may not be true but even if it is, i have to ask "what's the thesis for all the giddiness into the 7th week of market recovery?" one possibility could be the market really thinks 6 - 9 months from now the economy will start showing signs it's working it's way out of the recession. another thesis could the market got so sold off that equity valuations were just too enticing to let go by, even up to friday's close. another thesis is regardless of the state of the economy, the fed and the treasury are on the mark with saving the over-leveraged banks and liquidity is unfreezing the credit iceberg.
there are other plausible theses, but i think it's sort of a combination of these 3 above. the markets did get too over-sold. a snap-back was over-due. earnings and guidance from some big banks, tech and industrial companies beat estimates and projected somewhat rosy guidance for the year. quantitative easing by the fed has given confidence and leadership around the globe. but all this good news may be a little over-rated after the 5th week of the run-up. are these signs of an emerging and imminent bull market? i don't know. i do know that regardless of the direction the stock market takes, we can still profit.
What the model is telling us
the model says, as i started off from above, we are in a sort of "no man's land" particularly for aggressive users of the
S&P 500 Index Arb Model. no man's land in this instance is a region where PID vacillates close to zero. this movement is accompanied by several crossings through zero which is our indicator to re-balance the ETF pair to neutral weights or for aggressive users to change their weightings bias from one extreme to another. non of this type of vacillating movement generates profits for aggressive users. for delta neutral users, the motion is a bit of a nuisance but no big deal one way or another. see the green line on the chart below.
the model generates profits when PID is on either side of the x-axis. the coefficients i have chosen for the model provides the above characterization. remember, the x-axis is like the fulcrum point of a see-saw. the longer PID stays above or below zero, the more time we have to accumulate gains. the same is true of the magnitude of PID. the greater the absolute value of PID, the rate of gains increases. so time and value are important to making gains using this model. if time is short and the magnitude of PID is small between transitions, the less potential to make gains. the behavior describes a see-saw or a swing that is oscillating ever so slighlty about equilibrium. not much potential for fun for anybody.
the mechanics of the model is based on classical mathematics used to derive equations of motion for a system of two point masses or two solid bodies in coupled oscillation with respect to one another. in this case i have used calculus of variations and my own variant of the hamiltonian equation to derive a long/short ETF dynamical relationship. finally, the key to completely describe the system is assuming a set of boundary conditions (such as sliding versus rolling motion for example) and a method of integration to solve the equation of motion. once an engineer has the equation of motion for a system, theoretically the engineer has enough information to control or stabilize the systems motion. thus, there is yet another level of control we can pursue which is a pre-set rate of return using our paired ETFs. this may come as a special update or in a new version release of the model.
bits and pieces of this discussion and other details will be background for a users guide i am writing for these ETF models. i understand the concepts behind the model may still be a black-box but stay tuned for future alerts and updates with more illustrations to fill in the blanks.
what action should be taken now?
aggressive users should maintain a significant over-weight to SSO and significant under-weight to SDS. i let you decide that percentage based on your tolerance for risk. be alert though because changes can quickly ocurr as we've seen recently.
users wanting to stay in the ETFs but remain neutral, make sure you re-weight your pair on monday using the neutral weights specified in the spreadsheet that were calculated for yesterday, friday APR-24.
comments and questions are encouraged.
Posts
7 comments:
Hi Mike,
I have read a few of your posts on seeking alpha as well a few of your posts here on your blog. I have a pretty clear understanding of how these ETFs function and how they react to different market sitiatuions, but I am still having trouble understanding what your trading system tries to profit from? I am wondering whether you are trying to profit from directional bets(in your blog you seem to state your directional bias for these trades)? pricing errors that occur as a result of leveraging daily returns? or is it something else? I understand that you may not want to go into to much depth about this system as it is something you have worked hard to develop, but if you could provide a basic concept on what you are actually trying to profit from, that would help alot.
Thanks.
hi J.
to put it simply, gains are actually generated from long-term slippage and daily tracking errors in the long & short ETF. another words, incremental gains are accrued from exploiting an arbitrage between 2 reasonably behaved inversely leveraged ETFs.
the model tells me how to maintain a delta neutrally balanced portfolio by slowly adjusting the weights of each ETF. neutral is very rarely a 50-50 weighting for the pair, hence the arbitrage.
when bias changes in the model move from 1 ETF to other, i re-weight to neutral or re-bias the weightings to exploit these imbalances for a gain. of course, a neutral model implies minimal to no gains or minimal to no losses over a given period of time.
let me know if you have other questions about what i said above.
Hi Mike,
Thanks for the reply. I have a few more questions for that I hope you can answer…In this period of trading where volatility is high, by holding both sides long and not rebalancing every single day you may be fighting some possible decay related to volatility(although it would be small in these 2x broad-based index ETFs as compared to something like FAZ and FAS), is your model strong enough to beat this volatility decay, or am I misunderstanding how your model would work? In these more volatile markets would it make sense to change your model so you shorted both pairs of ETFs to also take advantage of this volatility decay instead of going long? Also, if we were in a steady and directional market like that of 2006-2007 would your model outperform further just as the pairs of ETFs did during this period due to the low volatility?
Also, I still am a little confused as to how your system works…it sounds like most of the time your model wouldn’t consider a 50/50 balance to be an equal dollar amount balance…so how is it deciding which side is overweight and which side is underweight? And how is it not doing this so the trades don’t take a directional trade form? (Not sure if this is too much information to ask for.
Also, if you are trying to profit mainly from the long-term slippage and daily tracking errors in long and short ETFs, why not use this model on some of the more volatile and more decay heavy ETF's like SKF and UYG, SRS and URE, and FAS and FAZ?
Do you have any information on the compounded or uncompounded percentage gains this model has had since it has been implemented in real time and also how it has performed when it is back tested?
Thanks again, Mike.
hi J.,
thank you for asking some very, very good questions. i've spent a lot of time doing research on most of the topics you raised and watching how the model behaves in various market conditions. give me a little time to thoughtfully follow up with answers to your questions as best i can. my model has only been commercially available for about 6 weeks. a lot of people will appreciate the questions and answers.
cheers!
hi J.,
here's a reasonably rigorous attempt to answer your questions. leaving comments to posts on Blogger has much to be desired, but that's another matter.
J. writes........
Hi Mike,
Thanks for the reply. I have a few more questions for that I hope you can answer…In this period of trading where volatility is high, by holding both sides long and not rebalancing every single day you may be fighting some possible decay related to volatility(although it would be small in these 2x broad-based index ETFs as compared to something like FAZ and FAS),
Q-1) is your model strong enough to beat this volatility decay, or am I misunderstanding how your model would work?the short answer is YES but the limitations are not dependent on the model but the ETFs in the model. i screen the ETFs i would like to consider as candidates for a long/short pairs trade. the screening involves looking at long-term relative rates of changes of each ETF. doing a little bit of analysis and charting, i can determine whether a neutrally weighted pairs trade ever breaks down. that's the first criteria that must be met for consideration.look this chart of DIG & DUG and then look at this chart of SDS & SSO. the green line in both is the models computation of neutral. in the DIG-DUG chart, notice how the green line drops negative. right when it dropped negative the pair was no longer capable of being balanced to neutral. with SDS-SSO, everything has been smooth sailing.
my work-flow goes like this. i compute neutral EOD weightings for each ETF in a prospective pair going back as long as possible. this is done by using a model i developed based on the techniques of calculus of variations. the result yields a homogeneous partial differential equation very similar to Hamiltonian-Lagrangian-Euler equations used in classical mechanics of machines analysis. new neutral weights are obtained by solving this equation as a line integral along a path of one day. it's important to note the solution to each line integral stipulates the least amount to change BOTH ETF weightings to maintain neutrality (equilibrium) corresponding to long-term rates of changes of each ETF. the most optimum solution involves changing both weights, not just one. this follows the low of least action built into the mechanics of the model. since these ETFs compound or slip at different rates as a function of market changes, precise weightings are required to maintain a neutral pair. finally, i evaluate and plot the sum of each ETFs weight times it's relative long-term rate of change. the sum as defined above, for a long/short ETF pair that is capable of being balanced, will always equate to zero (0). if the sum is ever < or > 0, the pair is not in equilibrium and thus the pair is not a good candidate to be employ while this condition is true. the sum is computed going back as far as the ETFs have been on the market, ideally more than a year. otherwise there's not enough information as far as i'm concerned to determine fitness for my model. there has been a lot of extreme market activity in the last 12 - 18 months.Q-2) In these more volatile markets would it make sense to change your model so you shorted both pairs of ETFs to also take advantage of this volatility decay instead of going long?no because as long as i'm working with a long/short ETF that is capable of being balanced, the long ETF and short ETF take care of each other.Q-3) Also, if we were in a steady and directional market like that of 2006-2007 would your model outperform further just as the pairs of ETFs did during this period due to the low volatility?i'm not quite sure i completely understand the question. but here's something i think will address what sounds like the issue. remember, the pairs would at all times have a long & short component. steady directional changes in the market, either way, make it even easier to exploit whichever ETF is offering the gains. now, the absolute rate of growth of the net pair value will NOT be as much as potentially in a volatile market but the model should beat the tracking index for sure. by how much, there's no way i or anyone else could know this ahead of time. but compounding is not as much a benefit in steady markets as it is otherwise.now, during a period of range-bound trading like we've been in since APR-9, it's rather frustrating for broad-market long/short hedging models. i employ moving averages in the analysis, no longer than 12 trading days. when net market variations frequently vasillate about a given point, the net growth in pair values doesn't change much either. so know that my model is not an end-all-be-all model for every type of market condition. there needs to be net changes in the tracking index over a 12 day trading period for the model to generate gains.Q-4) Also, I still am a little confused as to how your system works…it sounds like most of the time your model wouldn’t consider a 50/50 balance to be an equal dollar amount balance…so how is it deciding which side is overweight and which side is underweight?ok, if one ETF is growing at a faster rate than the other ETF in the pair, that's the ETF to overweight until the next re-balance signal or if polarity changes again before a re-balance. the model keeps track of polarity (bias) and most of the time bias conditions exists long enough to accumulate gains. moving averages are employed so some days there would be counter trends during volatile market conditions. but not significantly so on average to change the bias from one ETF to another. this is statistically shown in the analysis and charting of the data sent to subscribers each night.Q-5) And how is it not doing this so the trades don’t take a directional trade form? (Not sure if this is too much information to ask for.a distinction should be made between using my model in a neutral scenario (preservation of capital) and an aggressive high beta scenario. the model always keeps track of stability and indicates when to re-balance. when a re-balance indication exists, the indication is for both scenarios. there aren't separate re-balance indicators for different scenarios. there are different re-balance indicators for different pairs, though.but in between re-balances, polarity can change and this is a secondary re-balance indicator for everything but the neutral scenario. a conservative low beta scenario that is not neutral and is not aggressive can be achieved as well. say you start off neutral and the net pair value begins to drift upward (not an unusual circumstance.) say the net pair value has approaches a level you decide is good enough to maintain before the model issues a signal to re-balance. you can re-balance ahead of time to keep that pad of gains under your pair. your are more than free to re-balance at that point and let the model drift from there until a re-balance signal is issued or at a point you want to reset again yourself. there's nothing stopping a subscriber from doing so. the model is a tool for subscribers to have a better picture of overall market direction as opposed to just looking at index valuations. the model is both a technical analytical tool and a hedging mechanism wrapped up in one product.Q-6) Also, if you are trying to profit mainly from the long-term slippage and daily tracking errors in long and short ETFs, why not use this model on some of the more volatile and more decay heavy ETF's like SKF and UYG, SRS and URE, and FAS and FAZ?i have studied pairs like SKF & UYG and DIG & DUG. DIG & DUG break-down last october and hasn't yet returned to balance-able levels. for various reasons not germane to using the model, DUG slipped down for a while just a fast as DIG and that ruins the viability of a pairs trade. SKF & UYG are intact but SKF still suffered a considerable amount of slippage and bid/ask values varied wildly during extreme volatile trading sessions. this doesn't make for a preferred pairs trade either. i haven't studies SRS & URE. i'm not sure how long URE has been on the market. but new levered ETFs with less than $1B in assets are taken off the table. this goes for FAS & FAZ. very unreliable and tons of slippage making it off limits for a consistent pairs trade. i've limited the retail models down to what i feel are 3 best choices for conservative reasons. i don't like unknowns without at least 1yr of trading performance, i don't like low volume and low asset values and i don't like wild bid/ask spreads.Q-7) Do you have any information on the compounded or uncompounded percentage gains this model has had since it has been implemented in real time and also how it has performed when it is back tested? have lots of back-test data i can send you if you wish. just let me know. the models stand up well against more than 18 months of back-testing. beyond that, the data isn't as reliable nor have some of these proshares ETFs been in existence much longer.all back-testing is based on actual market performance. in general, compounding is a straight forward analysis using reliable historical market data. currently my analysis does not list out the amount of compounding on a daily or long-term basis but whatever the slippage is, it's built into the neutral weights calculations.
Hi Mike,
You can send any backtesting material you have to user83248324@gmail.com . I would be most interested in seeing the percentage gains that your system has built up and what your systems max drawdowns have been(largest one day percentage losses).
Thanks again.
ok J.,
please give me till this time tomorrow at the latest to generate a snapshot and send you a couple of back-tests results of my SDS-SSO model. i think you'll like what you see.
best regards,
mike
Post a Comment