Over the years, I have become friendly with Mebane Faber, co-founder and the Chief Investment Officer of Cambria Investment Management. He manages an ETF called the Cambria Global Tactical ETF (GTAA).

Back in 2007, Meb authored an excellent paper titled “A Quantitative Approach to Tactical Asset Allocation.” It was published in the Journal of Wealth Management, Spring 2007.

That analysis tested and reviewed simple timing models, using a 10-month moving average on various asset classes as a signal to enter and exit asset holdings. Compared with traditional “Buy & Hold” investing, the performance improvements across all asset classes were quite significant.  The methodology has the advantage of being objective, unemotional and mechanical (fee free to insert “first wife” joke here).

The timing strategy works as an effective risk management tool that enabled investors to miss most of the fall in major bear markets. It also captured a significant portion of the subsequent rally. Signals were infrequent, not prone to false positives, and avoided “whiplash” – the many false buy and sell signals that typically plague timing systems.

As an example, look at the 2007-09 bear market. It peaked in October 2007 at 14,100, and bottomed in March 2009 at 6,500. Using the 10 Month Moving Average, traders would have exited the Dow Jones Industrials ~12,650 (1/31/08), and avoided the next 6,000 points down. The re-entry was July 31, 2009, at 9100.

Improving the methodology

I describe this quantitative method as symmetrical: The identical signal that triggers exits (markets breaking below their 10-month MA) also triggers entries (markets breaking above their 10-month MA).

However, in my experience, market tops and bottoms, are asymmetrical. They have very different characteristics in terms of timing, investor psychology, trading activity, market breadth, economic factors, and internal metrics. Tops are much more of a process, as buyers slowly lose their ardor for equities. Bottoms are more of an event, as sellers capitulate and dump holdings. Hence, tops develop over longer periods of time, while bottoms tend towards a faster more climactic event.

Thus, a ripe place to explore for possibly improving the original quantitative methodology might be on the entry side. Is it possible to identify a better entry than the 10 month MA?

I have been discussing this with Meb, and we plan on exploring a variety of factors related to (in alphabetical order): Earnings, economic data, Fed funds, insider buying, market price, reversions, sentiment, trend, valuation, volatility, and yield curve. If anything comes of this, we will publish our findings at SSRN or some other suitable journal.

Question: What other metrics are worthy of market timing exploration? If Traders use a 10  month MA as their exit, what might enhance their entry price?

The goal is an entry determined by a systematic, objective, data driven, metric, with modest drawdowns and limited false signals.

Feel free to add any ideas, suggestions, research, metrics . . .

Category: Investing, Quantitative, Trading

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

43 Responses to “Revisiting “Quant Approach to Tactical Asset Allocation””

  1. CANDollar says:

    Quoted from a Globe and Mail article from August 5 2011:
    “S&P’s Sam Stovall recently examined the S&P 500 over the past four decades to see how well investors would have done if they had employed moving averages as market-timing devices, versus simply buying the entire index and holding it.

    He considered single moving averages (50-, 65- or 200-day, taking both the moving average and the exponential moving average, which places more weight on the more recent days), using the moving average to mark the buying and selling point: The investor is fully invested whenever the index is above the moving average, and fully on the sidelines whenever it is below the moving average.

    He also looked at pairs of these moving averages – in which the point where the two averages cross each other marks the buying and selling point. (The investor would be fully invested whenever the 50-day or 65-day moving average was above the 200-day average, and fully out of the market when the shorter average was below the 200-day.)

    While the results were mixed for strategies that simply tracked the shorter-term moving averages, Mr. Stovall found that investors using the 200-day average and the average-crossover strategies would have generated appreciably higher returns over the past four decades.

    Longer term better, cheaper

    Mr. Stovall found that while none of his moving-average strategies had a particularly high frequency of beating the index on an annual basis (all were in the 30- to 40-per-cent range), they were all quite effective at protecting investors from the deepest market downturns while still capturing most of the market rallies. What really makes following a longer-term moving average system desirable, in my opinion, is that it forces you to stay in the market for as long as possible, but then gets out of the way before any real damage is done to your portfinvesting strategyolio, he wrote.

    The longer-term and crossover strategies have an added benefit: lower transaction costs to managing your strategy. Because the 200-day average and the crossovers are hit much less frequently than the shorter-term averages, the investor isn’t buying and selling nearly as often; in the case of the crossover strategies, buy/sell events occurred, on average, less than once a year.”

    Theres a chart that goes with the article but you have to be a member to read the entire article. The original S&P research may be available on their site. They studied the market timing approach from Dec 1969 to July 2011.

  2. constantnormal says:

    For starters, I think that the entry point determination is going to have to be a function of more than price and/volume. Markets can hug the bottom for a long time until the right combination of circumstances arrives.

    Possibly the re-entry algorithm might have to be a heuristic that attempts to distinguish between such things as a (lack of) liquidity condition being resolved, an inventory overhang being resolved (with situations like the present being an excess of debt inventory), …

    Something that looks for the right conditions to ignite a sustainable rally. That will likely *not* be something that looks only at price/volume movements.

    Or it might. Looking at historical PE’s might prove useful — although they did not get down into the single digits in 2009 (which might have been the influence of FASB 157 meddling, given the large role the financial industry plays in our economy today). A sudden plunge in index PE’s might indicate a selling climax, especially if prices are plunging faster than earnings, vs an index-wide earnings collapse. An earnings collapse is probably going to be associated with the beginning of a crash rather than the end of one.

  3. constantnormal says:

    Has anyone done anything looking at valuation metrics? Stuff like price/book, or price/sales ratios? Presumably, sales would pick up before stock prices, although that is more of a question than anything else …


    BR: Consider these:

    Mkt Cap/GDP ratio
    Dividend Yield
    Tobins Q: Equity Value/Book Value
    Shiller CAPE
    Value Line VLMAP

    I am sure there are more but these are worth exploring

  4. inthewoods says:

    I like using the percentage of stocks below a 50 and 200 day moving average for the S&P 500. I’ve found it works very well for entries.

  5. znmeb says:

    Oh, man, where to start! ;-) I could recommend a few books on the subject, but my guess is you’ve read them all. I’ll be blunt – technical trading rules for most markets are at best only slightly better than wishful thinking, and at worst a complicated morass of brittle code waiting to bite bodily parts off of users. Developing technical trading systems require hours of hard work and the resulting systems offer minuscule edges.

    There was a time when this was not the case. That time was well-captured in books like LeBeau and Lucas’ “Computer Analysis of the Futures Markets” and Katz and McCormick’s “Encyclopedia of Trading Strategies”. There’s plenty in these two books on how to go about this if you still want to go down that path, but I think you’re wasting your time, and more recent research indicates I’m not alone in that thought. Good luck with it in any case.

  6. Veneziano says:

    I have seen similar systems with percentage “deadbands”; i.e., if the deadband is 2%, the price must go 2% over your signal before entry, and drop 2% below before exit. the deadband could be asymmetric, (e.g., 2% entry, 3% exit, or something like that) if a data analysis suggests it.

    The deadband mainly prevent whipsaws.

    I have played around with SP500 data with a couple different methods, and they all seem to work fairly well over the past 10 years, vs buy and hold. You could optimize the system (MA and deadbands) to optimize results, but that is just look backwards. IIRC, a 10 to 12-MA with a 3-5% deadband is pretty solid, looking ~10 years back.

    A bigger point, I think, is that these timing methods work much better in secular bear markets (e.g., now) vs secular bull market (e.g., 1982-2000). In secular bulls, they seem to lag. This seems true of a couple different timing systems I have seen (Faber’s 10-mo or 12-mo MA, or a 200-day MA, or a 50-day MA/ 200-day MA crossover); Buy and Hold rules (or ruled, I should say) in past secular bulls.

    If I could identify the next secular bull / secular bear switch point, now there is big money….

  7. Well, Buy & Hold has revealed its shortcomings — disastrous during secular bear markets!

    Active trading does not work or most investors.

    The alternative is using tactical strategies to get out of the way of the collapses.

  8. nizer says:

    Here is a link to a Google Spreadsheet summary that compares some SP500 strategies from 1950 – 12.23.2011: http://goo.gl/8zZtT

  9. TraderMark says:

    I *think* I am replicating this in freestockcharts.com now

    Barry can you confirm this 10 month MA would have created a lot of noise summer 2010? Looks like it would have triggered an exit on 5/31/10. Then crossed back over 7/31/10. Then back below in August, then back above (for good in September)… until 8/31/11 when it broke back down. And has only crossed back over 12/31/11.

    Are you seeing the same thing or am I charting with wrong setup – this is my first time on their site.


    BR: Look at the monthly CLOSE — not just the prices during the month. (Candlestick charts might help)

  10. TraderMark says:

    Also it crossed below the 10 month MA on the monthly Dec 31, 2007 rather than Jan 31, 2008 as you noted above but got the same re-entry on July 31, 2009. So wondering if I am slightly off since I show a break down 1 month earlier than you noted

  11. znmeb says:

    “Well, Buy & Hold has revealed its shortcomings — disastrous during secular bear markets!

    “Active trading does not work for most investors.

    “The alternative is using tactical strategies to get out of the way of the collapses.”

    Modern portfolio theory, options pricing theory and their cousin arbitrage pricing theory were supposed to be the answer. However, as Bob Haugen put it, “modern portfolio theory only works if everyone uses it.” ;-) Meanwhile, Professor Haugen has moved on: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1306523

  12. TraderMark says:

    So if I am viewing this correctly, only the DJIA has signaled a ‘return to market’ as of Dec 31, 2011. The NAZ, SP500, and R2K are all still below the 10 mo MA as of that period. Probably why Barry is in the cautious 50% involved with equities stance, or at least partly.

  13. refhounds says:


    You have got it backwards…

    “Markets form their tops in violence; markets form their lows in quiet conditions.”-One of Richard Rhode’s trading rules and certainly true for most markets.

    Not too sure why you want to improve the 10 month moving average rule .Adding to it will only increase whiplashes (I have the bruises to show for it)

    Another way of using this rule aggressively is to apply it to sectors.Not only it gives more opportunities but also some easy pickings especially on the downside in these market conditions

  14. The Window Washer says:

    Seems Fed funds is your only macro.

    How about CapEx and Z.1 flow of funds.

    Two of my favorites but now that I typed it I doubt they will fit in the 10 month time frame.

    So do you mean improve the method or improve on the 10 month method?

  15. CapArb says:

    We’ve found that signals from the credit market can be used effectively in TAA. For example, the empirical relationship between high yield credit and the Russell 2000 can be used to drive a long/flat strategy across most North American equity indices.

    Details here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1872163

    What’s especially nice about this approach is it tends to be uncorrelated to other popular security selection and TAA strategies and works well as an overlay.

  16. kenny powers says:

    The only way to find out is to TEST. This takes time and effort, which is why so few bother…

    I recommend Trading Blox (www.tradingblox.com) as a good value. It is not cheap, but cheaper than the comparable software out there, and top notch. It also has a forum to which full admission is worth the price of the Blox Builder alone.

    For data; use CSI Unfair Advantage or a cheaper option would be Norgate data, which is accurate and very reasonable.

    One of the very nice things about testing is that you debunk a lot of the things you belive to be true about the market which are in fact total bullshit. I am a professional discretionary trader, but I benefit greatly from backtesting because of this. And my Blox Builder has paid for itself many, many times over.

    Btw, I have no affiliation with Trading Blox other than as a customer.

  17. Gloobus says:

    I think that keeping this one simple is the main objective. So a possibility to time entries would be to go with easing into these positions with parts of the funds after some time interval. It would look like a pension scheme. The same for selling the position.

  18. kenny powers says:

    I also recommend Anthony Garner’s book “A Practical Guide to ETF Trading Systems” for a good read on this subject.

  19. ValuValu says:

    Adding complexity to a simple but efficient trading strategy improves theoretical results, but at the cost of robustness. Especially if you don’t factor in additional information that may mitigate the over-fitting.

    In plain terms: a strategy based solely on historical prices is either a momentum strategy, that will work well to follow trends but will miss reversals, or something that draws nice shapes with the wishful thinking those shapes will re-appear. in this case one may as well try tea leaves and astrology!

    If we stick to momentum, the question becomes: what OTHER information can we factor in to anticipate reversals? I think there are 2 paths:

    1 – Valuation (Yield, P/B, whatever), because the stock market has to go back to reality… somedays.
    2 – The market’s mood.

    We decided to use the mood, but both have their merits.

    PS: Maybe there’s something with volume as well, because it could show the significance of the moves

  20. JasRas says:

    Short answer: yes. I’ve got something interesting. Backtested with supporting data. In use now.

    If you want to discuss it more, I will be glad to meet with you and discuss it. After you sign a non-disclosure form.

    You have my e-mail.

  21. BigDaddy says:

    An analysis of Value Line data going back to 1973 shows that their forward projections work reasonably well near bottoms, not so much at/near tops. There have been 148 weeks of the 1969 since ’73 where they projected forward returns above 125%. Of those 148 instances, there were 22 that showed negative returns in the following year averaging a decline of 16%. In the other 126 instances where the following year was positive, average returns were over 28%. Last instance of over 125% – the weeks between 2/20/09 and 4/10/09 – investments during those weeks had subsequent year returns over 40%. Unfortunately the very low forward estimates by Value Line are not as profitable. Put me down as one following Value Line for market bottom signals.

  22. Conan says:

    There are basically big waves or major market turns, medium sized waves which are notable turns in a big wave and small waves that are short term fluctuations in the market. What you are really talking about in this approach is focusing on the big waves. You want to be in the major turns to BULL Market and out or short in the major turns to a BEAR.

    So to see this you either have to use Monthly charts or long duration number on daily charts. So look at a MACD of say 200,400,100 or 100,200,50 or some such combination and you will see this effect more clearly. As stated before in one of the posts above you can also use 1 to 2% filter to keep from getting whip sawed. But the point is in order to see these big market changes you have to up the elevation of the helicopter.

    Good luck, nothing is 100%, it is all just probabilities and risk management!!

  23. orvil tootenbacher says:

    As an example, look at the 2007-09 bear market.

    Hmm. Seems a little too convenient to identify a short market window where a quant strategy worked. why not a broader more rigorous backtest in several markets? Without rigorous sampling, dismissing a strategy as random is legitimate and boxing it into the all technical analysis is crap box is where it ends up.

  24. the_imperatore says:

    I also use technicals, mainly Movering Averages and MACD. Sometimes I look at the RSI, but that’s it.
    I came to realize that with TA, just like in life, having too much is like not having anything at all.

    The timing is really the most important thing in trading.



  25. dougc says:

    Margin debt, the least capable tend to think they can borrow at 10% and beat the market. They tend to buy small marketcap stocks and iwm will be outperforming spy.

  26. Gator81 says:

    Your point regarding the asymmetry in market tops vs. bottoms has been the focus of my practical research for some time. After digging into many of the potential signal carriers you mentioned, as well as others, I have found one to be very useful for both its content and its simplicity: the VIX.
    A quick view of a VIX chart against SPX from 2-Jan-90 reveals that the VIX contains a similar and opposite asymmetry. To make it useful, the VIX must be smoothed, and of course there are unlimited ways to do that. I found an exponential moving average to be helpful, then set upper and lower trigger points. Within my model, I can vary the period of the EMA and the trigger points to find the combination that best avoids the waterfall SPX declines while capturing most of the steep recoveries that follow.
    For example, looking at the 10-year period 21-Dec-01 through 20-Dec-11, a 35-day EMA on the VIX, with a sell trigger occurring when that EMA falls through 36.0, and a buy trigger occurring when that EMA rises through 10.7, will generate a net gain on the SPX of 94.9% with 69.8% exposure. This compares to a Buy & Hold (i.e., 100% exposure) net gain of 32.3% for the period. The key outcome for this parameter set is that there are just 2 buys and 1 sell during the entire 10-year period: 2-Oct-02 (Buy), 20-Feb-07 (Sell), and 20-May-09 (Buy). Declines of 27% (20-Dec-01 thru 1-Oct-02) and 35% (21-Feb-07 thru 19-May-09) are skipped.
    What I like about this application is that it captures the essence of the asymmetry, which is too complex for my mind but I believe is mostly about psychology: fear and panic are not the opposite of greed and euphoria, and it shows in the VIX.
    What I don’t like about it is basically two things: 1) the VIX only goes back to Jan-90, and I much prefer more robust datasets; and 2) this approach is rife with the pitfalls of data mining, requiring the user to constantly “back out” and look at the bigger picture to see if it still makes sense.
    Another concern is that there hasn’t been a closing VIX print below 14.6 since Jun-07; how much of that is because SPX isn’t near its next top, and how much is because HFT and the algos have changed the basic nature of trading volatility? Nevertheless, the concept seems sound.
    I’d be happy to share the data file.

  27. inthewoods says:

    @Gator81: I don’t understand a few points about your comment:

    1. Sell trigger occurring when that EMA falls through 36.0 – Given the inverse correlation of the VIX to the SPX, why would you sell when the EMA falls below 36? Shouldn’t you buy? I’m not sure I understand what your technical definition is of “falls through” – can you clarify?

    2. Same question on your buy – buy when that EMA rises through 10.7 – so buy when volatility is rising? Don’t get it.

    Overall, I think you are right in terms of your own critique of your algo – 35d moving average seems data mined (could mean over-optimization). Also anchoring on specific VIX numbers seems, to me, to be asking for trouble – I’d use a relative measure. Thanks for sharing!

  28. CANDollar says:

    Here is one simple formula that gives excellent risk adjusted returns:

    Go long using the 50 day EMA – but not if trending downward.
    Go to cash using the 10 month moving average or the 200 day E/SMA but do not go to cash if the 50EMA is still above the 200SMA.

    In the Canadian large cap equity index (which is more volatile than SP500) this would have resulted in 5 long trades and 6 trades going to cash in the last 10 years. Presently you would be in cash instead of equities for that portion of the portfolio.

    Picking individual stocks doesn’t work for most investors so use broad indexes represented by the lowest cost ETFs.

    Nobody knows which asset class will be outperforming at any given time so own:
    Major world equity indexes SP500, EAFE, Emerging
    Domestic and international REIT indexes
    Commodity index
    Bond universe index

    Consider using value averaging approach to go long various asset classes and the 10 month MA to go to cash instead of the above rules with 50 day MAs.
    Potentially goose return using small cap index, maybe quantitative fundamental
    Rebalance annually (any more doesn’t make a difference)
    Reinvest dividends.
    Optimize taxable income using registered accounts.

    Since your fees will be extremely small (probably on the order of 0.15% of assets under management or less) you will keep approximately 40% more of your money over a 30 plus year period than if you had a traditional money manager who will not be able to exceed the return of this portfolio structure.

    For retirement account:
    Consider a very simple strategy of holding mostly the inflation indexed bond indexes from around the world and using small amounts of out of the money longer dated call options for long exposure to equity indexes. Use options that make money to buy more TIPs. This is Zvi Bodie’s “Escalating Life Annuity”. Real return bonds are best held in registered accounts.

  29. Gator81 says:

    Sorry, yes, I stated it backwards. Not enough coffee for the wetware!
    Generally, a falling VIX signals a rising SPX; so BUY when the EMA of the VIX falls through the trigger point, and vice versa… thanks for the correction…
    As for the data mining problem, it is very real. What I’d like to do with a longer data set is look at what VIX levels correspond to capitulation in each of the long cycles. With the VIX data available, there are only two or three “panic”-type time periods. There must be some other data source that tends to have that asymmetrical behavior that mirrors SPX but with a longer data history. Bond prices?

  30. Hey Barry, first time commenting here but long time reader, thanks for the excellent blog.

    I have a number of ideas, I’ll start with 1 on the longest timeframe.

    I call it the “Two consecutive quarters chart”.

    I first observed this effect here in Aug 11. Although not very scientific, over the past 2 decades, any instance of 2 qtrs lower has been a reliable top sign. Bull markets do not have more than 1 quarter lower in their runs. Similarly, coming off of a bear market, 2 quarters up signals a buy.

    ————> the thinking behind this is the mega money, the pensions et al, move on such a timeframe (and slowly) that after 6 months, they are either making money on their respective plays buying on dips, or losing it. That fuels the next part of the move.

    It also fits the concept that tops take longer to form.

    Here is my first observation in chart form: http://goo.gl/uRBn8
    Here is a followup chart a friend made following our discussion: http://goo.gl/85ygH

    I have not personally checked the numbers in chart 2 (not my chart) but assume they are correct, and for visual explanation of the concept, do a good job.

  31. CANDollar says:

    Great charts! Useful and begs further inquiry.

  32. LevyBrain says:

    Playing devil’s advocate here.

    GTAA is performing pretty poorly versus the SP500 since it’s inception.



    BR: Its global — the benchmark is not SPX but MSCI or some other global benchmark

  33. ty CANDollar!

    unscientific as I said, but definitely was/am keen on the observation. And, for what little this is worth, I felt creative bc I have not seen similar analysis (re the 2 consecutive bars theory) elsewhere.

  34. Eric Kennedy says:

    Barry, although bearish sentiment can be hard to accurately quantify, at extremes it does a good job of signaling a market that has gone as far as it will. AAII suffers from a small sample size and varying group of people submitting surveys, so I have to agree with Helene Meisler that Investors Intelligence is the best one to look at. Investors’ Intelligence mentions the importance of finding a positive divergence in their sentiment data, where the market continues to decline but newsletter sentiment isn’t quite as bearish. That happened in 2003 and 2009.

    Trying to incorporate too many factors can make it hard to have confidence in the signals. I was following way too many indicators in 2008 trying to determine when it made sense to get back in the market, and it was really stressful. It’s hard to argue that when sentiment is a negative extreme, a buying opportunity is still far off. It might just be a bear market rally, but I bet you’d still get a better price than if you waited for the market to close over the 10 month sma.

    Excessive bullish sentiment doesn’t work as well to determine when to get out of the market, so trend following works great to stay with a trend.

    Barry, sentiment is half-way down your list and should be higher if you look at _extremes_ of sentiment in a quantative measure like Investors Intelligence

  35. J. Francis says:

    This ties in with sentiment a bit but why not use a Google search on the ticker or even just the company news/product searches and see what the relationship would be with a high-hit count entry vs. a low-hit count entry.

    What I’m trying to suggest is maybe it’s a more realtime and accurate coincident indicator that can give you an idea of how popular or crowded a name is getting….

  36. [...] I’ve also chatted with my buddy Barry Ritholtz, and we’ve been batting around some ideas with various other factors that would apply to timing stock indexes. [...]

  37. adid says:

    Use the S&P500 percent of stocks above their 50DMAs. This indicator does not usually go above certain levels during bear rallies and below certain levels during bull corrections. As soon as this rule is violated you have an entry signal.

  38. kbt says:


    The ECRI WLI is available in Excel dating back to 1967:

    Another short little book that looks at moving-average crossover systems using both technical and fundamental variables is _Time_The_Markets_ by Kirkpatrick (2012).

  39. Gator81 says:

    BR, here’s another shot, more directly at your challenge: “an entry determined by a systematic, objective, data driven metric, with modest drawdowns and limited false signals”.

    The metric is “moving average stacking”. The concept is that an entry is signaled when the asset price rises above its short MA, which is above its mid-term MA, which is above its long term MA.

    Here is an example, using SPX as the asset, with a 30-year daily-close dataset (7,570 data points, 24-Dec-81 thru 23-Dec-11). Setting 3 exponential MAs at periods of 21, 63, and 126 days (i.e., 1, 3, and 6 months) generates 320 entry signals, which is 10 or 11 per year. Let’s assume you can execute the trade at the closing price on the same day the signal is generated.

    After 5 days, 246 of those entries are winners, 74 are losers, a ratio of 3.3:1. The average gain for all 320 trades after 5 days is 1.2%, with a best of 6.8% and a worst of -8.3%.

    After 21 days, 231 of those entries are still winners, 89 are losers, a ratio of 2.6:1. The average gain for all 320 trades after 21 days is 1.5%, with a best of 11.6% and a worst of -21.98% (ouch!). Presumably, stops could mitigate those loss numbers.

    There’s nothing magic about which 3 periods to use for the EMAs. Shorter periods generate more trades with smaller average gains and smaller maximum losses; longer periods generate fewer trades with larger average gains (up to a point) and larger maximum losses.

    If you exchange the [21,63,126] period set in the example above for, say, [50,100,200], the number of entries goes down from 320 to 214; the 5-day average gain rises from 1.2% to 1.3%; and the 21-day average gain rises from 1.5% to 2.0%. The winner:loser ratio after 5 days improves from 3.3:1 to 3.5:1; at 21 days, it improves from 2.6:1 to 3.0:1. Meanwhile, the maximum 21-day loss worsens to -24.2%.

    So, I’m curious: do those winner:loser ratios meet your definition of “limited false signals”? As for the “modest drawdowns” portion of your challenge, that would be determined in large part by the exit portion of the strategy, which is not addressed here. True?

  40. bcca47 says:

    I have to agree with LevyBrain (above): GTAA has performed poorly since inception and, contrary to what you say, it’s not because the fund is global. If the fund is following the strategy outlined in the prospectus (and in the white paper that you cite), it only invests 20% of assets in international assets (and only when the timing model is favorable.) The other 80% are in bonds, real estate, domestic equities, and commodities (again, when the timing model is favorable) in roughly 20% blocks. The fund’s prospectus and white paper explicitly compare its performance to benchmarks like the S&P. The fund aims to control drawdown in down-trending markets, but so far it has been trounced. The strategy does not seem to be working: even after GTAA pulled assets out of all categories except bonds, and at a time when bonds were still going up, the fund lost ground. Since the fund is expected to trail the relevant indices in an up-market, I don’t see how it is going to outperform overall if it underperforms on the down-side. The strategy sounds good and looks pretty on paper, but something is not working, and so far Cambria has offered no explanation.

  41. Scott Teresi says:

    I think GTAA is underperforming because the market is rising while bouncing around the 200-day SMA (similar to 10-month SMA). GTAA went risk-averse just after SPY plunged 15-20% in August. Then as SPY recovered, GTAA was still in risk-averse assets and has fallen behind. If the market plunges again soon, GTAA will look very smart. Otherwise, its market call based on the moving averages had a cost in lost opportunity.

    With not much data on GTAA over very many market conditions, it’s hard to see if this will be typical of the algorithm or just the occasional price of using a risk-managed asset allocation fund.

    Maybe so many advisors are now using trading strategies similar to GTAA that the fund will become a contrarian indicator. (Hope not, since I have some money on it, believing in the strength of Mebane Faber’s long-term tests in his 2007 paper!)

  42. BobDobalina says:

    I missed this earlier, but a friend sent me the link -

    You could make it really simple and get back in when the price crosses above its 5-mo moving average, instead of waiting for the 10-mo MA to get back in. A quick scan over the last 10 years shows it should get you in 1-3 months earlier and there appears to be no “false” signals. Once you dipped below the 10-mo MA, when you cross back above the 5-mo MA, we’re off to the races for a little while anyway.