Core Wealth Management


Since launching MarketSci in 2006, I’ve always been successful building active strategies for generating outsized volatile returns (see what I’m trading today).

But I often struggled with more conservative strategies for the core of my wealth, the types of strategies that produce more than equity-like returns with less than bond-like volatility, and minimize exposure to market shocks and other risks.

For the last few years, my core wealth management solution has been Tactical Asset Allocation, trading from a diversified basket of asset classes once per month.

My model, like most similar models (1), has been subpar in recent history; such is the ebb and flow of such long-term approaches. I’m confident that over long time horizons, TAA will provide value because these models are based on concepts that have endured for as long as modern markets have existed (trend-following, momentum, diversification of risk, etc.)

But I’ve grown less happy with TAA as the solution for managing my core wealth. Admittedly, part of that unhappiness is the lackluster performance of the last 18 months. But if that were the only problem I would stay the course; an advantage of taking a quantitative approach to the markets is justifiable conviction during lean times.

The bigger issue for me is the one-two-three-punch of (a) middling returns, (b) long hold times (i.e. holding trades for months at a time), and (c) the limited upside remaining in Treasuries.

Bad days trading are inevitable. But having a bad day without being able to get right back into the fight with a real chance of making up losses is hard to stomach.

Yes, you could design your TAA model to be more active, jumping in and out of asset classes more quickly, but I think you lose a lot of the benefits of those long-term concepts that make TAA so powerful (and increase trade frictions as well).

Add to this the limited upside remaining in Treasuries, one of only a few major asset classes remaining without a strong positive correlation to equities. With Treasuries, all of these asset allocation models are going to suffer in terms of return. Without Treasuries, they will all suffer in terms of volatility. It’s just math (read more and more).

The question then is how best to build a strategy that generates consistent solid returns and minimizes beta risk, but is also much more active, and doesn’t rely on Treasuries as a major component.

I’ve been trading a big portion of my wealth with my own solution to that problem since late last year.

Readers know I don’t talk about a program unless I can show real-time, verifiable trading results (because backtests are worthless (2)), so starting this month I set up a small tracking account with IB to track performance just like all of our strategies.

More thoughts to follow. This is meant to be a conceptual post, my mea culpa if you will, explaining why I’ve begun to fall out of love with TAA (and yes I appreciate the irony of falling out of love with TAA as the rest of the blogosphere is falling head over heels).

And just to be clear, even though I never sold the TAA model, it’s something I’ve traded and talked about extensively in the past, and something I’ll continue to trade with a portion of my portfolio in the future. Readers know I abhor cherry-picking returns, so even if one day I stop trading the TAA model altogether, I’ll continue to carry results of the TAA model here at the blog and at indefinitely.

Happy Trading,

(1) Of course, I’m referring to real-time, out-of-sample TAA models (i.e. the only kind that really matter), not the backtested variety burning up the blogosphere.

(2) To clarify, as a research tool, as a way to share and discuss ideas, backtests are super duper. But as a way to judge the efficacy of a black box, they are worthless.

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.

21 Responses to “Core Wealth Management”

  1. Are you going to give us a hint at what your alternative is?

    • 2 MarketSci

      Hello Brian – sure, in a future post when I have some real-time results to share. The point of this post is just to lay bare why I’ve had a change of heart on this whole TAA concept.


  2. 3 ATrain

    A lot of TAA models have been under-performing. Decision Moose comes to mind. There are several reasons for this…

    1) Lower transactions costs, electronic trading, and algos have reduced the efficacy of momentum investing. Trends have become shorter and turning points are much quicker

    2) You mention this: Treasury yields are near the zero-bound. TAA backtests owe a lot of their high sharpe performance to a 30 year bull run in US treasuries. That bull market is near its end

    3) this is a key point: cross asset correlations have become more volatile, making diversification difficult. During a crisis, correlations converge and diversification fails.This is s a symptom of high wealth concentration

    4) most TAA models ignore currency risk. In the new world of countries competing to debase their currency, a naive approach that ignores this will under-perform. Just look at the Nikkei (in yen). It should be near the top in a TAA model but since most models only consider the Nikkei in dollars, its not

    5) probably the big one: government intervention.

    There are better approaches out there, like Adaptive Asset Allocation, that incorporate advance risk management techniques.

    • 4 MarketSci

      Hello ATrain – I agree with most of what you wrote. I take issue with the last point.

      Not to deride the work of GestaltU or anyone else, but trying to make an apples-to-apples comparison between the sea of purely hypothetical backtested models to the small handful of models (like Ivy or GTAA or DM or ours) that have been trading out-of-sample in real-time for any length of time, is silly. Everyone looks good in a backtest. But funny how rarely anyone has the guts to share verifiable real-time results, sans cherry-picking, for any significant length of time.


      • 5 PLM

        The main reason people don’t track TAA strats real-time — besides laziness — is that it won’t prove anything. How would 1-2 years worth of a rebalance-once-a-month trading strategy returns (12-24 data points) prove anything about the efficacy of the strategy? Unless the results were extreme, it wouldn’t show anything. By the time you had a large enough sample, much of your sample would be too old to be terribly predictive (unless you assume market behavior is static and permanent). No amount of backtesting or live testing is going to validate (predict future performance of) a very low-frequency strategy. That’s not to say all such strats are invalid; it’s just to say that testing is not very useful with tiny samples.

      • 6 MarketSci

        PLM – I agree that those real-time stats would be less informative than a strategy trading every hour, day, or week. But to say they would have NO value over backtested numbers? No, I don’t agree.

        I think anyone who’s been around this topic long enough and seen the vast, vast majority of programs fail to live up to backtested expectations understands the importance of taking advantage of whatever real-time data is available (insufficient as that may be).

        The alternative? What we have now: investors perpetually led astray by the constant stream of eye-popping, but (likely) totally unrealistic backtests from the blogosphere, books, and Wall Street.


  3. 7 Yaba

    “Mea culpa” means “my mistake”. Hence, “My mea culpa” means “MY MY mistake”. That’s funny. ;)

    Congrats for your great blog: you’re a genius. Best wishes for 2013.

  4. Why not try market timing using moving average crosses (like 5-10-20) or RSI?

    • 9 MarketSci

      Hello Rich – if you just apply that kind of approach to equities or equity-like instruments, you introduce a lot of beta risk. That’s okay for the risk-seeking portion of your portfolio (see the uber risk I take on with volatility ETF trading), but not okay as a core wealth management approach for a portfolio of size.

      You could widen your net and include non-equity-like instruments like Treasuries or gold to try to diversify away some of that risk, but you’ll find that those kind of approaches have not worked well historically on UST or gold (not to mention the limited upside remaining in UST that I talked about in this post). michael

  5. 10 Carlos

    Hey Michael,

    Good post, as always.

    I just finished reading Nassim Taleb’s “Antifragile” (which I recommend). In it, he talks about an approach that involves putting something like 90% of your assets in a more or less risk-free category (cash or something like it), and then taking the remaining 10% and doing some fairly risky stuff with it.

    If the financial world comes to an end, so to speak, you’ve only lost 10%, but if your risky investment pans out, you still generate decent long term results.

    The main advantage of this is preserving your capital, not maximizing your gains. I think capital preservation is a reasonable long-term goal to have in mind for this type of program.

    Just food for thought….


    • 11 MarketSci

      Hello Carlos – very good “outside the box” comment. Personally I wouldn’t take that approach b/c (a) it’s inherently impossible to model quantifiably, and (b) risk free returns are guaranteed to be near nothing for the foreseeable future, but I think the concept has merit. michael

      • 12 Carlos

        You’re right on both counts, Michael. But keep in mind that Taleb’s over-riding theme is that we absolutely *cannot*, under any circumstances, predict long-term “black swan” type of risks. So he would argue that your point about not being able to quantify the risks is exactly why you should do this.

        As to your second point, if you earn next-to-nothing on 90% of your assets, but still own 90% of your assets after a big dive, you’re still in relatively good shape.

        All of the above notwithstanding, I’m not personally deploying my capital in this manner (at least not yet), but it is something to keep in mind.

      • 13 MarketSci

        Hello Carlos – I totally agree and for the reasons you mention I think the idea has merit.

        Summarizing the choices we’ve just talked about: either giving up all effort at quantifying expected return/risk but having the protection of a fixed max loss (Taleb’s approach), or making a best effort at quantifying expected return/risk, accepting the fact that that’s never truly possible and max loss is always theoretically total (the usual beta-driven approach).

        A little foreshadowing of what I’ve been doing: there’s a third choice. A strategy that makes a best effort at quantifying return/risk, but does it in a way with an inherently built in fixed max loss (think pairs-trading, options strategies, etc.) It goes without saying: easier said than done.


  6. Nice post Micheal. I applaud your practicing TAA for 18 months. I have sat out of the market since July 2007. Yes, that’s right. Now that new highs are being hit, I’m thinking of getting back in.

    But seriously…

    My frustration has been like that of most quant oriented investing folks, how to generate consistently high returns. I recently found a salve that is going to help me with this chronic condition going forward. The salve is in the form of a book by Michael Mauboussin called “The Success Equation.” Basically he examines the role of the Luck-Skill continuum in investing returns (and sports, and business). A predecessor version of the book which covers basically the same concepts can be found for free here

    So, now my goal is not to generate consistently high returns. Yes, that’s right. My goal now has a few steps:

    a) identify what returns are realistically skill based (this will be an educated guess, but I have some thoughts on how to proceed) vs luck based (tails)

    b) develop quant systems to maximize the quality of skill based trading decisions, and thus maximize available skill based returns (within my ability to apply those skills – ie quant systems)

    c) identify the unknowns (if and when possible) that cause returns to be more luck based, and try to make those unknowns into knowns (off topic example – lawyers now focus heavily on jury selection when in the past not so much), and account for them by moving the new knowns to the quant/skill side of the toolset.

    d) accept a set of returns within a ‘luck-driven” variation around my targeted skill based returns, and even use bad luck to help me: eg mean reversion trades around the skill core.

    This is esoteric right now, but after reading the book/article I have a some quant insights to get there.

    Net net, I want to share the frustration on TAA, but also share that skill vs luck can fool us, and take us off what otherwise might be a good path.

    Philosophy discussion over.


  7. 15 steve

    michael, haven’t read this paper but I’ve heard about it. have you checked out:

    • 16 MarketSci

      Hello Steve – that one is out of Optimal Momentum – I looked at it when OM first released it – if memory serves, it’s another variation on these momentum and/or TF-based TAA-like strategies. Both of the issues I talk about in this post would apply here as well. michael

      • Good topic, Michael. I know the feeling!

        Of course one of the poblems with Taleb’s approach is: “What hapens to your long term plan when you very quickly lose the “very risky” 10%, and then do it again, and then …?”

        Gary Antonicci’s Optimal Momentum paper was originally published in the NAAIM Wagner Award contest where it won first prize. You can read other papers at:

        The 2013 Active Investing Paper competition (our fifth annual) requires submissions by Feb. 28, 2013 to be eligible for the $14,000 in prizes ($10,000 first prize).

        To find out more go to:

      • 18 MarketSci

        Hello Jerry – very good point re: Taleb’s approach. Yet another drawback of an inability to (at least attempt to) quantify return/risk. michael

  8. 19 Don

    Look forward to hearing more about your solution. I’ve been trading TAA since 2006 and still doing ok with it, but I’m finding it to be more challenging. I’ve been thinking that perhaps it’s popularity is it’s downfall.

    • 20 MarketSci

      Hello Don – personally I think it’s just part of the natural ebb and flow of these type of long-term strategies – backtests always look good when viewed from 30,000 feet, but when you get down to individual months and years, they sometimes look very different.

      I say that only because of how long these TF/momentum plays have existed and how huge most of these asset classes traded are. Of course, that’s just a best guesstimate. michael

  1. 1 Friday links: knee jerk reactions - Abnormal Returns | Abnormal Returns

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s