Monday, July 13, 2009

Popular Moving Average Models: Another Look

Over the past couple of weeks, several popular moving average timing models have given buy signals on the S&P500. Prices on the S&P500 finished the past month above the simple 10 month moving average, and this is a buy signal according to the model put forth by Mebane Faber. Then there is the "golden cross" or the 50 day moving average crossing above the simple 200 day moving average. This model has gained the backing of CNBC news anchor Dennis Kneale who has adopted this strategy as one of his talking points that the recession is now "officially" over.

{Editor's note: This post was written Sunday night, and following a 2.3% "pop" in the S&P500 on Monday, readers may not feel that it is relevant. Nonetheless, I thought it was worthwhile to publish as it gives you insight into how one can use the characteristics of a trading strategy. My apologies for the untimeliness of the post -not really- but I had a more pressing engagement as I spent the day with my daughter at an amusement park in Santa Claus, Indiana of all places. Now that is cool!}

While many pundits ridiculed the simplicity of such models to predict the behavior of a complex beast such as the market, I think they missed the point. By themselves, the models aren't useful as timing models. They are loss avoidance models. With these models, you avoid the big hit. You avoid the bear markets. If prices close below the simple 10 month moving average, then you are out of the market. It is that simple.

On the other hand, within the context of the appropriate market conditions, these models can provide a useful way to navigate a potentially hostile environment. For example as showed in the article, "Inflation Pressures Are A Legitimate Concern", I can improve the efficiency -i.e., make more money with less time in the market - of the Faber strategy by over 50% by avoiding those times or signals when inflation pressures are high. I have added context to the signal and made it better. In essence, without the context of some other factor such as sentiment or inflation or an economic variable, these simple models are just two squiggly lines crossing on chart.

Another way to use the signals is to use the characteristics of the signals themselves. By this I mean that one would look at the how the trading strategy performs and then choose the buying and selling points. So let's look at the Faber strategy a little more closely, and to do this, we will look at the maximum adverse excursion (MAE) graph. See figure 1. MAE assesses each trade from the strategy and determines how much a trade had to lose before being closed out for a winner or loser. For example, look at the caret in the figure with the blue box around it. This one trade lost 2.5% percent (x-axis) before being closed out for a 15% winner (y-axis). We know this was a winning trade because it is a green caret.

Figure 1. MAE Graph/ Faber Strategy

The current buy signal with the Faber strategy was on the monthly close with the S&P500 at 919.32. At this past Friday's close, the S&P500 was at 879.13. Since the buy signal, the S&P500 has lost 4.4%. So if you acted upon this buy signal from the Faber strategy, your current position would be underwater by 4.4%. Because the low was at SP500 969.32, your MAE for this trade (initiated at the close on June 30th) would be 5.4%.

Now let's return to figure 1. The blue vertical line is set to an MAE of 4.5%. What we notice is that 12 of the 33 trades had MAE's greater than 4.5%, and this is to the right of the blue vertical line. Of these 12 trades with MAE's greater than 4.5% only 2 ended up being winners. As a matter of fact, if a trade had an MAE (or lost) more than 6.5% it would end up being a loser; this is to the right of the red vertical line.

So how can we use this information? The current signal already has an MAE of 5.4%, so it has a high likelihood of ending up being a loser. On the other hand, this would seem like a good low risk time (remember this was written Sunday night) to actually act upon the signal. Why? By acting now with the strategy already 4.4% underwater, we are reducing our risk. If the S&P500 drops another 2.1% making the MAE 6.5%, we are pretty much assured that the signal from the Faber Strategy will be a failed signal. Furthermore, another 2.1% drop would take the SP500 below our important support levels of 876 and the down sloping 40 week moving average, and as stated previously, this would be a good time to "get out of Dodge!" if this were to happen.

This is a good low risk opportunity for a trade. Why a trade? There is no real compelling reason to be long this market; the context for which the signal has occurred hasn't really changed over the past two weeks. "Green shoots" are looking more brown; investor sentiment is not favorable; inflation pressures, as measured by the trends in crude oil, gold and 10 year Treasury yields, has moderated substantially (but two weeks doesn't make a trend). Also I would consider this a trade because the characteristics of the strategy suggest that this particular signal has a high likelihood of being a loser. But the most compelling reason to get long for a trade is the well defined risk. We know from the MAE graph that if the S&P500 losses another 2.1% then the signal from the Faber strategy is unlikely to recover; in addition, this would also leave prices below our "key" support levels.

{Editor's note: this section was written Tuesday morning.}

The market's down draft of the past two weeks probably left some analysts bewildered and worried about their positions. "That darn "golden cross"! They told me it signaled the end of the recession." I am sure many threw in the towel too early and are now sitting on the sidelines. They were unfamiliar with the characteristics of the strategy.

On the other hand, there were some analysts out there who stated, "Stupid technical analysis; there's nothing bullish about two lines crossing on a graph." I am sure these folks were feeling all pumped up the past two weeks only to be deflated on Monday.

As we all know, the market will do its best to confuse the most.

Having a strategy, understanding the characteristics of that strategy, and executing the strategy according to plan are difficult but worthwhile tasks.

8 comments:

jbr said...

But, what about the head and shoulders in the indices? :)

Guy M. Lerner said...

jbr:

I would reply (but not trying to be smug or offensive): what is the significance of the head and shoulders pattern? Look at past data and determine what that pattern has meant for prices going forward. Does the pattern work across multiple assets? Are there other factors that add context to such a pattern or signal?

jbr said...

Ran across this interesting post:

Testing the Head-and-Shoulders Pattern

It would be interesting to see if there were any differences between the period tested (90s) and this decade.

dacian said...

Guy,

Thanks again for this great post. I agree with you it's never enough to emphasize the "risk management" aspect of things.

There seems to be more added value to these simple (yet efficient) strategies using some filters (which are a bit more involved). Is there a way, in order to offer protection for these "trades", for managing stops?

If we consider Faber's model, a close below 10 month MA at the end of the month is a sell signal. A more aggressive trader might want to go short for say 10% of it's portfolio instead of being on the sidelines. How we could setup a protective stop for that trade, if the model prove to be a false sell signal (the same is true for going long for a trade).

thanks

Guy M. Lerner said...

JBR

Thanks for the link; that was interesting and their findings can be summarized by the following:

"In summary, head-and-shoulders technical analysis is likely unprofitable as a standalone trading strategy in rising markets, but it may work well in designing hedged portfolios."

I think this is saying something akin to all overbought signals in a bear market are a sell signal. Plus head and shoulders tops are typically found at market tops but they don't mean a market top; similar to the work I did on negative divergences.

Guy M. Lerner said...

Dacian:

Thanks for the feed back.

You wrote: "Is there a way, in order to offer protection for these "trades", for managing stops?"

I think I have an idea of what you are asking. The MAE graph is helpful in this regard; looking at all trades from the strategy we can set a stop that will always capture the winning trades or we can set a stop outside the noise of the market. So looking at the current MAE graph from the article, we know that an MAE >6.5% showed no profitable strategies and in fact they did not even recover. So this would be a good place to set a stop -outside the noise of the other trades.

I think your second concern is best summarized as to what happens after a failed signal in the Faber strategy? I have thought about this; what happens after the losing trades? I will check it out and let you know.

dgau said...

Holiday World rocks! Glad you enjoyed it. Best amusement park in the world. The rides aren't the biggest or fastest but they're fun. The park is rarely crowded. They have free parking, drinks and sunscreen and you often see the owners walking around the park checking on things and talking to customers.

Guy M. Lerner said...

dgau:

You are absolutely right!!! There are advantages to living in the midwest and out of the fray.