Hedging

Manage
1

As much as I really like the Portfolio123 approach to stock investing, it is really tough to make a profit, when we are faced with the bearish kind of a market we have had for the past 6 months. Just a quick glance at the model ports will bear out this fact. The models are down from 8% to 24% on 26 week performance, with drawdowns ranging from 24% to 33%. The GE fiasco last Friday really doused hopes of a near-term market breakout, and what with the financial giants’ poor earnings reports portending more doom and gloom in the coming days, the financial horizon currently looks a bit gray for the markets.

My sense is that the prevailing philosophy of Port123ers is to stick it out through thick and thin, keeping the faith that, although good ports will have big drawdowns at times, sooner or later, they will rebound in a sharp upward curve, justly rewarding those with the nerves and stamina to see it through the dark days. My observation has been that good ports will return 50% to 100% or more(on an APR basis) when the market is having a good run. That is why we love them, and that is why we are here!

So my thought is that, since you can make the big returns when the market improves, if you could devise a hedging strategy to offset losses, when the market is bad, then the bite that the DDs take out of you wouldn’t be so painful. It would make it a lot easier to have the stamina to stick with your ports, when the market is sinking. When the market turns up, you may lose some money on your hedges, but the rate of loss would pale in comparison to the rate of gain on your ports.

Specifically, I am looking at the inverse ETFs, like QID, DXD, SDS, etc. So my question is, “Is anybody using these ETFs to employ the type of hedging strategy that I am considering?” If so, can you offer any tips? Specifically, a formula for how much of an inverse ETF to buy per $1,000 invested in your live ports, to offset losses, would be extremely helpful to me. W ould you advise going with the short ETFs or the ultrashort ETFs? I am not looking to make profits per se, but rather to self-insure my ports against loss to some degree, enough to take some of the sting out of a serious drawdown.

I am not as good with numbers and technical formulas as many of our resident geniuses here. That is why I am asking for help. Also, I figure that some of you may already be actively pursuing this kind of an approach, and have already figured out the basics. Are you employing or considering a similar strategy? Do you have a different, better strategy for hedging, or making profits in this market? Anyone developed a successful port for shorting the market? All feedback will be appreciated by me, and, perhaps, by others here, since we are all swimming in the same turbulent waters at this time, IMHO.

2

Unfortunately I don’t think there is any “free lunch” when it comes to hedging. The way I see it, there are 4 things you can do to reduce drawdowns in bear markets/correcting markets:

  1. Go completely to cash. Zero drawdown risk
  2. Hedge using index futures / etfs
  3. Hedge using index puts
  4. Hedge using a “short” strategy

Lets analyse the problems with each one of the above:

Option 1. Going to cash - this is ok in theory, but can have significant opportunity cost. If you wait for the market to turn before you get in, you miss out on the inital moves, and they can be extremely powerful. Take a look at this, my old “Advanced momentum w/value” port fared quite poorly during the 2002-03 bear market:

http://www.portfolio123.com/port_summary.jsp?portid=338910

But then when the market turned, it returned its best returns of the entire timeframe:

http://www.portfolio123.com/port_summary.jsp?portid=338955

Note the returns there 178% in 12 months - that is nearly treble your money. Go and look at these ports again. One follows directly from the other. How do you think you would have felt at the end of the first one? Pretty bad. The port looks “broken”, doesn’t offer any protection during the drawdown for nearly a full year. Yet you were richly rewarded if you just stuck with it.

Option 2: If you hedge by selling the index, you may be in for a nasty surprise. The problem is that most indexes are capitalisation weighted, and therefore heavily concerntrated in megacaps/large caps, which tend to be lower beta. When the market falls, they often fell less than small/microcaps, and as a result, you may find that you don’t feel hedged at all. Also, if you get a strong market rally, as you often do in the end of a bear market, you will have significant opportunity cost in terms of missed upside.

Option 3: In a way, buying index puts seems like a no-brainer. You get to have the protection against downside but no limitations on upside participation. However, as is often the case, there is no free lunch. My own back-of-the-envolope calculations indicate conclusively that buying puts is a losing strategy in the long run. The reason: all options have a “time value” element which decays as time goes on, just like the way your car depreciates. Continually buying and selling depreciating assets is a way to lose money. In a sense this is obvious, if it wasn’t a money spinner, no one would be prepared to write puts.

Option 4: Hedge using a shorting strategy. This is my favoured route, the short strategy can make money as a “standalone” strategy, so there is little or no opportunity cost. However, I tend to find that shares I want to short already have a high degree of short interest (indeed is a factor in some of my systems). This can lead to powerful short covering rallies, where short strategies significantly outperform the market, at least in the short term. As a result, such a strategy would be far from drawdown free, and indeed, it is not certain in my mind that the drawdown characteristics would be better than a long only portfolio. I am looking at spreashees of this sort of thing, it would be nice if market-neutral portfolios were a feature in p123.

So there you are, there is no perfect answer to this. When things are going well, it almost seems like a license to print money. But there is a piper to be paid; investors must endure significant periods of pain and drawdown: that is the trade you make!

3

Here is a link to an article outlining historical testing of protecting portfolio with various put strategies (the first one I came across):

http://www.fpanet.org/journal/articles/2008_Issues/jfp0208-art8.cfm

They find value in straightforward put buying strategy.

4

Jaybee and Olikea,

thanks for the posts.

I also opted for a SHORT portfolio for my hedging needs.
Olikea: You have described the problem with this: Bear markets can have nasty counter rallies which are difficult to withstand. With SHORTS we have limited potential gains but unlimited potential losses, so it is not for the fainthearted.

Also the most difficult puzzle to solve is: When do you expose yourself to such a SHORT port? One may opt for a “permanent” SHORT port (which will look bad over the long run) as prices advance.

This means we need a TIMING vehicle in P123 to go along with our ports. Personally, I think this is the most important missing feature of P123. But I think Marco is working on this (Marco are you reading this)?

Wern

5

I believe I might have an answer to that. I have been paper trading a few shorts, the “worst” performing ones are thus:

http://www.portfolio123.com/port_summary.jsp?portid=334373
http://www.portfolio123.com/port_summary.jsp?portid=356531

That’s no hedge, its more like an anti-hedge!

However, I was determined to find out why this was the case. The short portfolios already have a high degree of short interest, indeed, if they seem to pass on your other factors but no the SI, it means the smart money knows something you don’t. So I don’t advocate throwing away SI as a factor. However, a while ago i noticed this:

http://www.portfolio123.com/mvnforum/viewthread?thread=3190

In other words, if people are bearish on the market, quite a lot of them don’t just short the index, they short the weakest stocks. Makes sense, I would do the same. Unfortunately, then when they want to get out of the trade, if they head for the exits all at once they are going to get crushed. I believe that is what we saw.

Therefore I believe it would be very useful to know the aggregate level of short interest in the market, and indeed I created a feature request (please vote if you haven’t already done so):

http://www.portfolio123.com/mvnforum/viewthread?thread=3196

That way, perhaps you could avoid following shorting strategies when aggregate levels of short interest are very high.

I guess this is analogous to avoiding the dot-com bubble at its peak, when everyone is on one side of the trade, its about to crack severely!

6

Olikea,

there are several indicators that I would like to see for MARKET TIMING. The data is all readily available, so I don’t know why it takes so long to implement this:

  1. Advance / Decline Ratios
  2. Put Call Ratios
  3. Number of New Lows
  4. Number of New Highs

I am not sure how much lag is involved in this number. Will the lag still make this valuable?

Wern

7

As you are aware by now, P123 foundation is relative ranking. This means that it is geared towards keeping you in the best stocks, ALL THE TIME. There is nothing wrong with that since most studies show that active traders / market timers (as a whole) usually end up losing out to buy & hold.However 100% invested can be painful, and the conclusions from the many studies are based on averages, but who wants to be average?

I think Oliver spelled out the two main ways to achieve more “straight line” results:

  • Market Timing (MT)
  • Hedging (multiple ways)

if you believe that there’s always a bull market somewhere, then sector rotation might be another. There might be ways to do this now in P123, so I’ll leave it out of this discussion.

We’re taking steps towards MT. The FED Model and all its components are available via the Buy/Sell rules. I’ve had some success using them this weekend (see attached charts). We might release some of these tests soon.

There can be many inputs into an MT signal generator and we’re collecting them. Soon we’ll release another page where you can chart several macro indicators, and plan to add many more. It will only be data though. Putting it all together and making them available to sims/ports will take serious brainstorming, but some ideas seem promising.

As far as hedging, I hope you can all appreciate how big this field is. There are many ways to do it. We will need to focus on the simpler kinds for now, which means options are out. I think the two that are within our reach are: hedging with index/etf , and shorting. I know we’ve been promising this for a while, but there is just a lot going on. One way you can help to speed this up is to spread the word around about P123. A modest increase in memberships could do wonders.

In the mean time, check out these simple tests of MT using RiskPremium. They only run for 5 years since the RP was too low before that, and nothing was being bought. Personally I like the DOW one. It’s only a total return of 66% vs 30% for the DOW, but it does it with less volatility (average of 3% dividend is not included). In addition you make some return during the cash periods as well.

NOTE: I only did this to start testing ideas about simple MT. Many say it’s not possible, but the results look encouraging (everybody needs positive reinforcement once in a while)

markettiming.png

8

Marco,

thanks a lot for this post.

Glad to hear that P123 is on its way to MARKET TIMING (MT). I really think it will make a BIG difference in the attractiveness of P123 for many if there is a decent MT approach we can use. As far as I know there is no competitor of P123 that has a Timer included, so there is business for you to be had.

We all know that we have fantastic Sims and Ports here but we also know that many have gut wrenching drawdowns which will be very difficult to withstand in real life. That’s why we need hedging and specifically MT.

I outlined some specific ratios I would like to see in my post above. The Fed model is also a start. There are many indicators used for Timing. We need to be able to test them and maybe even have a “Ranking” for various Timing indicators (just an idea).

With this new MT strategy of yours, I think P123 is heading in the right direction.

Wern

9

Jaybee:

Toughing it out - that is my current approach.

What helps me hold on with my portfolios is the equity graph comparing my ports with the benchmark in blue. Each week my confidence remains strong because none of my ports is doing worse than the most relevant index (the RUT in my case since most of my ports buy smaller cap stocks). I am “content” if my ports do no worse than the general market during bear periods. My belief is that they will greatly outperform during the next bull – whenever that might come.

Oh, I use larger ports than many P123 users since I have found that 20 stocks per portfolio gives a less volatile ride during bear markets and still have great upside potential in bull periods. True, I may be giving up 5%-10%/year when the market turns to a bull, but I am very happy that the larger ports are “well behaved” during this bear market. I would find it a lot harder to continue to hold smaller, more volatile portfolios. In fact, I retired two of my smaller and volatile portfolios early last fall and replace them with a 20 stock port built on similar principles. Since then I have watch the smaller ports (which I put on auto rebalance) bounce around a lot. I am glad I moved to the smoother ride of a larger port. I currently have money in 3 ports of about 20 stocks each (total of about 55-60 stocks). Two ports are doing no worse than the RUT benchmark and the third is ahead by a few percent. So I am “content” to hold on waiting for the bull to show up.

A year ago I was using PUTs to hedge. That worked fine up till the middle of 2007 but since then the PUTs have become a lot more expensive. A year ago I could get PUTs one month out that were 10%-12% out of the money for about 0.2%-0.3% of the value of my portfolio. That worked out to about 2-3% cost for a full year. But now PUTs are very expensive and the same protection would cost me 10%-15%/year!!! So I am currently without any PUT hedge in place. I was willing to pay 3% for a smoother ride, but not 15%. That 15% works out to double the money in 5 years.

As an alternative, I have been “toying” with putting more money into some type of sector rotation strategy to serve as a hedge. Last fall I put a bit of money into the DecisionMoose system when it rotated out of Latin American ETF and into a Gold ETF. I made about 10% on the Gold ETF from November until its exit signal in March. I only had a few thousand in Gold (just a small test amount, one step up from paper trading it) so it did little to hedge my overall account.

The reason I am hesitating putting more into the DecisionMoose model is that it “only” averages 25-30%/year and I am expecting my P123 methods to do twice that once the bear market is over.

Once I retire I may well move a third of my money into DecisionMoose or something like it just to smooth the ride out during bear periods. But at present my “greed” makes me reluctant to use DecisionMoose and thereby give up 10%-20% per year.

So far my search for a hedge confirms that there is no free lunch. I am willing to give up a little bit (larger ports vs smaller ports) to get a smoother ride, but PUTs and sector rotation just seem “too expensive”.

Brian

10

Wern:

I agree that P123’s business model could be helped by adding market timing. I expect it will be good for P123’s bottom line since it will attract more users and prompt existing users to upgrade to “gold” status.

But I do not think market timing will add much value to accounts of P123 user’s. I’ll come back to this in a minute.

I expect market timing will generate a LOT of posts on the discussion boards. By now the P123 community has largely finished the search for which fundamental factors “work”. Most work now seems to be in the area of buy/sell filters and rules. When the extended database comes out there will be a brief burst of activity as users test their strategies on “out of sample” data. This will not discover any new factors but it may show that a couple of factors are not as good as we thought. Most importantly, the extended database will give us a much better handle on draw down risks and how long “good” factors may temporarily stop working.

But once market timing can be done in P123 the discussion boards will be filled with timing ideas. From past experience on other boards, the ideas and variations for TA timing are virtually infinite.

I for one am not looking forward to the addition of market timing. The reason is that TA based market timing makes it very, very easy to mistake “noise” for valid signals. As one author wrote, it is easy to be “fooled by randomness”. The multitude of TA factors, most of which can be twigged by adjusting parameters via step based optimization, virtually guarantees a user can construct a “curve fit” method that will give the promise of outstanding gains in back tests which will almost always fall apart when one tries to use real money. The rule of thumb is: the better the timing back test, the less likely it will work in the future. Sad but true in my experience.

What many fail to realize is that one needs decades of data to be able to validate most market timing systems. The more factors one includes, the more years of data one needs to avoid unintended curve fitting. Look at it this way. If I construct a set of timing rules that lets me avoid major market dips and crashes, P123’s current data only gives me 4 “data points” to test my idea since there are 3 major dips in the 2000-03 period and one more so far in 2007-08. With the extended P123 data, we may be able to go back to 1995 or 1993 which will give us a couple more dips (1995 and 1998). So in total we will have 6 to 8 down turns on which to test our timing ideas. However, any statistician will tell you that you should have 30 or more data points before you can get a rough idea of how something behaves! So will will need 7 times more years of data than we currently have, and 3 times the data we will likely get with the P123 extended database. Sure, I can develop and “test” ideas on the data that is available, but the test results will be virtual meaningless. It will be like trying to build a house on the sand of a seashore. Shortly after one gets the building “finished” it will usually collapse.

So take care. Here are two tips.

  • 1 - If one wants to use market timing, only use it with strategies that one knows will “work” without market timing. Treat market timing as a supplemental add-on. Hopefully, the underlying fundamental strategy will continue to “work” even if the market timer stops working.

  • 2 - Never use margin with market timing. Even if you get lucky and find a market timer that really works, back tests results will almost always greatly underestimate the timer’s draw down risk. The last thing you want is to have a margin call liquidate your account.

Oh, one final point. I believe market timing does work if one is trading futures, especially commodities, since they tend to trend more. My caution is about market timing of stocks.

Regards,
Brian

11

Brian,

thanks for your thoughtful post. You are making good arguments against Market Timing. I agree that we have to be very careful not to curve-fit such an approach (as always) and that we can easily fool ourselves with things that worked in the past but will not in the future.

But still, I am not as pessimistic when it comes to MT. Your argument about the “number of data points” is true but it all depends on what you are considering a data point or “turning point”. Depending on your time frame there could be easily more than 30 data points to choose from during the last 10 years.

There are a number of factors and ratios that DO work imho because (as an example) fear and greed will never go away in the markets as it is part of human nature. That is why sentiment numbers give you a good indication what to expect in markets (excessive optimism/pessimism).

The # of New Lows is declining rapidly at Bottoms (as can be seen right now) etc. This indicator is quite reliable.

None of this stuff works perfectly because nothing does in the markets. There will always be this element of uncertainty otherwise we would not have markets and nobody taking the other side of a trade. No TA can protect us from “outside shocks” etc. So we have to live with all these limitations but MT would give us another useful tool in our toolbox, nothing more and nothing less.

Wern

12

Wern:

I see your point. It all depends on how many data points a particular timing system generates over the data history. For example, the “turn of the month” calendar timing system generates 12 entry signals per year which gives over 80 data points in the P123 data we currently have. As I am sure you know, the Turn of th Month system is based on the idea that there will be increased “buying pressure” from pension and mutual fund managers who often receive new cash inflows at the start of the month. Since some trader/speculators have caught on to this, the buying pressure seems to have shifted a couple of days earlier. So now the rule is often “buy at the close 3 days before the close of the month” and sell after holding for 5 (or 7) trading days". That method is supposed to give one most of gain for the entire month while being out of the market for over half the time. Because volatility is much lower (at least in back tests on indexes) some think it is “safe” to buy on margin or use the leverage of index futures to generate enlarged returns. Since this method gives 12 buy signals a year, I would call it a “fast” system. We have ample data to test fast systems (as long as the fast system issues its signals only on weekends since P123 back tests are limited to weekly data).

An example of a medium speed system would be the “Sell in May and Go Away” calendar system. This buys back in either the middle of October or the first of November depending of the variation one is considering. This method gives 1 buy signal a year. So for it we have at 7 buys signals (and will soon have 7 sells). To test this timing method we need 30 years of data. Once we get P123’s 15 year extended database we will have half of the minimum needed. Sometime around 2020 we should have sufficient data to test the “Sell in May” timing method.

And what about the famous 4 year Presidential Timing method? That’s the one that predicts the 3rd and 4th years of a president’s term will be the best for the stock market while his 1st and 2nd years will be the worst with the 2nd year being worse that the 1st. It also notes that the 3rd year will be better than the 4th. Assuming we want 30 data points, we would need 120 years of market data to test the presidential timing method. Is the presidential timing method built on a solidly recurring underlying characteristic of human nature or is it built on market “noise”? Fortunately we do have more than 120 years of data for this. Researchers report its presence from 1868 onward. For a summary report see

How does the presidential timing method work when matched up with good stock picking method? One study looked at the results of combing stock picking by O’Shaughnessy’s method and by ValueLines:
http://www.mechanical-investing.com/stock-market-timing-3.html
What I take from this study is that a good stock picking method will be able to make money even in the “poor” years of the Presidential Cycle (years 1 and 2). So even though the Presidential cycle is “real”, anyone with a good stock picking method would would loose money by using that timing cycle.

The Presidential Cycle illustrates what I consider to be a major challenge in evaluating timing methods. We have various stock indexes with long enough histories to test virtually all timing ideas. But we do not have enough history of stock fundamentals to determine if adding market timing to a good stock picking method will increase or reduce returns.

A couple years ago, I did a test to see if the “Buy in May and Go Away” method increased or reduced the gains of my favorite P123 strategies. In general using that timing method reduced returns. Of course, this is not conclusive since I would need 30 years of P123 data to get enough data points to know this with reasonable confidence. However, initial results from this preliminary study suggest that the “power” of good P123 stock picking methods can more than make up for the general market weakness during a typical summer.

As for “Turn of the Month” timing, there are enough years of data (6.8 years x 12 > 80 data points), but P123 back tests use weekly data which makes it hard to implement the “buy 3 trading days before the end of the month and hold for 7 days” rule. Nevertheless I did a quick test of this by putting “in cash” dates into P123’s “exposure” feature. The results were not encouraging, but this of course is not definitive since P123’s weekly data forced me to test a crippled variant of the Turn of the Month timing method. Although I can not be sure, my gut feeling is that P123 stock picking methods would generate profit during the “in cash” periods of this timing approach. Thus, I expect Turn of the Month timing to reduce gains when using P123 methods.

There is more that I want to talk about in your thoughtful post. But I will do that in a separate post.

Regards,
Brian

13

Wern:

I agree “fear and greed will never go away”. What makes timing so challenging is that the market players, whether they be professional or retail investors, have memory. Thus, when their greed or fear lead them to stay too long or get in too early, or whatever, those players will plan to “do better next time”. So they will do things a bit differently the next time. Perhaps they will exit a bit earlier or hang on a bit longer or jump in sooner or later. As someone has said, “history does not repeat exactly but it does rhyme”. Because investors have memories, the future will be different than the past, perhaps only slightly different or perhaps more so. Thus even if there were no curve fitting risks, human memory makes TA timing challenging at the fundamental level. But I am starting to do theoretical rambling.

Let me suggest something more practical.

Can you point me to a study on the timing calls based on the number of new lows? We can use the signal dates from such a study to produce “in cash” periods that can be placed in P123’s “Exposure” feature. We do not need to wait until P123 adds the ability to generate signals internally. We can import those signals via P123’s Exposure feature.

As you know, I have concerns about curve fitting risks with TA methods due to the number of variables that can be studied. However, we might be able to get a quick answer by trying out the “best” signals generated by an external timing program. If the “best” timing signals do not improve P123 returns then we do not need to get into long and convoluted discussions of how much curve fitting may or may not be involved in various TA signals.

As you may guess, my hypothesis is that even if a TA signal is “real” the power of P123 stock picking methods might be so strong that one is better off ignoring the signal and being “all in, all the time” using P123 selection methods. That is the pattern I see in the proven Presidential timing method when combined with good stock picking methods like Valueline or O’Shaughnessy. My previous post gives the URL for that study which showed that the most profitable approach was to ignore the timing signal and be “all in, all the time” using either Valueline or O’Shaughnessy.

With my Regards,
Brian

14

“A couple years ago, I did a test to see if the “Buy in May and Go Away” method increased or reduced the gains of my favorite P123 strategies. In general using that timing method reduced returns.”

There are a couple of issues here.

(1) you shouldn’t expect a strategy that has been fully optimized to get better results when you apply additional rules or constraints. Therefore if you take a favorite strategy and apply a TM exposure list to that strategy don’t expect better results. What you really need to do is design and optimize systems with the TM exposure list built in from the start. Then compare those systems against your favorite P123 strategies on an out of sample basis.

(2) Market timing may reduce risk without necessarily increasing profits. For example the “go away in May” timing strategy results in similar profits to buy and hold but has much less market exposure (risk). This has been demonstrated over the last 50 years or so on the major market indices.

Steve

15

“When do you expose yourself to such a SHORT port?”

My philosophy is not to try to improve results by hedging but to reduce risk. I generally hold back 25% of my available funds in something liquid so I can quickly hedge using an ultrashort ETF. This can also be done by using 100% of capital for investments and using margin for hedging. This technique won’t cover all of the downside but will take the bite out of drawdowns and allow one to still make a reasonable profit if the stock market goes up. I used this method quite successfully in the last year. During the major drawdowns I managed to get out will little damage by introducing hedges when things started to get rough. The hedge bought me time - I realized that things were serious and I managed to exit all my holdings.

I usually implement the hedge when I feel uncomfortable (i.e. not sleeping). This usually occurs when markets are very volatile. I also find that it is good to hedge after unusually good portfolio performance. If my results appear too good to be true then I am usually about to get slammed. This happens more often than not. I also like to hedge mid-summer until some time in October/November.

Steve

16

Steve:

I always look forward to reading your posts because I know I am going to learn something. This is especially so when you are responding to something I wrote.

Good point. I am not sure it applies to my example because my favorite strategies are not “fully optimized” at least not so with respect to profits.

When I test multiple variations for a strategy I look to see consistency and to find any “cliffs” to stay away from. So the factors I end up using are not necessarily the one’s with the highest profit in the back test. For example, back then I was testing a pair of value strategies with price momentum: one had price momentum of 75% with 25% value factors like p2s and p2fcf and the other had the same factors but with the weighting switched. Both tested well as did all the variations in between and more extreme weightings as well. In the end I did not pick the strategy variation with the highest profit in testing. That might have been 69% price momentum or 82% or whatever. I don’t remember. But I do know I did not pick the “best” (most profitable) one. Testing can only give one a very general picture of the “lay of the land”. Once I have that picture I figure arbitrarily picking a value is equally good compared using the value that had the highest profit on the tests. In that case, I picked two weighting values (75% and 25%), each from opposite sides of the “plateau” of good results. My reasoning, which might be wrong, was that if the market pattern shifted a bit, at least one of the two would be in the “good”.

Since I am not “fully optimizing” my strategies, at least not in the traditional sense of optimizing, I am not sure your point directly invalidates my the way I did a preliminary test of the “Sell in May” timing method. Of course, giving the fact I would need at least 30 years of data to get a statistically valid test of the “Sell in May” method’s effects on P123 strategies, I can not be dogmatic about any of my initial conclusions. Market Timing may well help P123 strategies over the long term, but we (me at least) will not be confident in definitively concluding one way or the other until we have many, many more years of data. Until then we each take our best “educated guesses”.

It may surprise you but I think this is creates more problems than it solves. In particular, it increases the amount of “educated guessing” we have to do after getting results. If we had unlimited data for testing, I would agree that one should vary P123 fundamental factors along with TA factors when developing a timing system. But there is not enough data, no where near enough data. So adding more factors in the design stage increase the amount of guessing we will have to do once we can see the results of the tests.

Even without including fundamentals, there is a major problem when evaluating 1,000s of TA timing variations. It has been shown that one can be guaranteed to get outstanding test results for several variations EVEN IF all the variations tested are completely devoid of predictive power. That is the nature of randomness when testing a multitude of variations. When I get time I will find the title of the book on my self that demonstrated this problem with an extensive study of index timing methods. From my perspective including fundamental factors as variables along with TA factors when doing strategy design will increase exponentially the problem of being fooled by random patterns when looking at back tests results.

Thus, your point about the need for “out of sample” comparison of methods with and without timing is well placed.

This leads to the question of how long a time period one needs for the “out of sample” test. The general rule of thumb is 2/3 of data for in sample and 1/3 for out of sample. That makes sense for verifying a single strategy or timing method, but it gets more complicated when comparing two. It seems to me that one should subdivide the “in sample” results to see how frequently and for how long the weaker method out performed the other stronger method during the in sample period. The “out of sample” period would have to be large enough to ensure that one did not mistake a short term out performance as being conclusive.

Good point about the reduction of risk.

Regards,
Brian

17

Here’s my $.02. I am 50% hedged at all times. I use inverse ETF’s in my IRA and a short strategy in my taxable account. If my strategies truly produce alpha, then I am willing to accept lower returns in up markets for a smoother ride.

It seems to me that the most violent swings happen without warning. Yes, timing models will get you out for a good chunk of a big move, but they can chew you up in a sideways market. If I were to attempt timing in any fashion it would be in a separate account and this account would just be part of my portfolio of accounts. I think timing has it’s place, but I think of it as almost another asset class to add to diversity.

18

Brian - it is true that mixing MT with the fundamental strategies introduces a new level of complication. But you would think that simple strategies such as “go away in May” and the presidential cycle have some potential given their long history of performance. In theory they should generally reduce risk in the long hall.

A possibility is to keep track of when your favorite strategies were developed and released. Then apply these MT strategies to the strategies post-release system and compare on an out-of-sample basis. Unfortunately it is difficult to make any conclusions without waiting a fair amount of time. For “go away in May” you would have to wait several years.

19

Some good discussion on market timing here.

I would just add a few of my own comments:

-I find that mechanical marktet timing difficult to implement. It doesn’t really obey simple rules such as “buy when above 50 day moving average” etc. and something really complicated is likely to be a curve fit.

-Discretionary market timing seems easier. Last year I knew at the time when we were in trouble when the S&P 500 made a double top. I have been recording my own calls on where I think the market is going, and I have been fairly accurate (though not yet traded on this).

Has anyone read the “TrimTabs” book? (Charles Biderman and David Santschi) I found it pretty interesting, talking about money flows as a way of market timing, this intuitively makes a lot of sense to me, and because probably few people are looking at this, it may be more predictive. E.g. the number of IPOs tends to correlate with the level of the market, and the number of share buybacks tends to inversely correlate with the level of the market. Also, fund flows etc. If you are at all interested in market timing then I think it is a must read.

20

I think we are missing one of the cheapest and better hedging techniques available to us: combining ports that have low correlation to each other.

I ran some port combinations on a spreadsheet, taking a look at each port’s weekly gains/losses. I combined two Hi Dividends, Growth and Value Small Cap, and a large cap. None of these were the eye-popping ports that we occasionally see. Good solid ports, but with drawdowns that still hurt.

Indepedently, most had draw downs in excess of 20%, one past 30%. However, when I combined them, the max drawdown was 13%. I then played around with a hedging factor, using S&P and R2000 ETFs. After testing I went with the R2000 at a 30% hedge. This dropped the max drawdown to 8%, comfortably within my target of 10% drawdown.

The results since 4/2/2001 to 2/25/2008

5 poorly correlated ports
No hedging
Total Return 694%
Max DD 13%

Same 5 ports with 30% R2000 ETF short (IWM)
Total Return 628%
Max DD 8%

I know I could do much, much better if Port 123 let me select sims and generated a correlation matrix for me. That would lead to much smarter combos than what I hypothesized would work together.

Even without hedging, the 13% drawdown was pretty amazing. What could we do combining sector specific ports? Style ports? other? The possibilities are exciting. We need a sim correlation tool: with that, the value of this site becomes, well, “uncorrelated.” :wink:

I found a feature request that addresses this topic. Please vote for it here
http://www.portfolio123.com/feature_request.jsp?poll=176

Carl