Quants: More technical details, facts and discussion

Home Forums Logical Invest Forum Quants: More technical details, facts and discussion

This topic contains 23 replies, has 11 voices, and was last updated by  Vangelis 11 months, 4 weeks ago.

Viewing 15 posts - 1 through 15 (of 24 total)
  • Author
  • #12327

    Alexander Horn

    Quants: More technical details, facts and discussion



    How has the Maximun Yield Rotation strategy performance and drawdown changed in the backtest when the Adaptive Allocation strategy is applied. I’d like to see a backtested comparison between the original strategy and the new strategy.


    Michael Kiefer


    Frank Grossmann

    You have the new charts in the latest strategy blog: https://logical-invest.com/new-maximum-yield-rotation-strategy-backtest-charts/

    Here is the old chart for MYRS since 2010
    old MYRS chart since 2010

    The CAGR annual return is nearly the same. 56% for the old MYRS and 54% for the new strategy. The difference is small because in the past ZIV performed well over long periods and if this is the case, then both strategies invest 100% in ZIV. For such periods with a strong trend, rotation strategies with a 100% switch do perform very well and the new strategy will perform about the same because it is also 100% invested in ZIV.
    The difference between a 100% rotation and a rotation with variable allocation however is very evident when the trend slows down and goes over in a flat market like it is the case for 2014. In this case a variable allocation does much better. The volatility is reduced a lot and the new Sharpe ratio (1.15 for 2014) is double the old Sharpe ratio (0.61 for 2014). Since about March, the new strategy is holding about 50% Treasuries (EDV). The effect of this can be seen here:

    You have better return with a fraction of the volatility. Holding 100% ZIV for 2 weeks in such a volatile whipsaw market is like playing Russian roulette. The chance that the strategy is switching wrong is very high. With the new strategy I feel much safer now. It is a pity, that I can not back-test it for much longer periods with ZIV, but the results correlate nearly 100% with my new Universal Investment strategy which invests with a variable allocation in SPY-TLT. This one can be back-tested until 2002 and also shows much better results during market periods without a strong trend.

    Best regards


    Tom Gnade

    I don’t see any information on the SPY-TLT strategy. Do you plan to post that soon?



    It is already created, just in editing now. You will see it shortly.


    Don Krafft

    When volatility increases, eventually the position sizes will be reduced by the equal volatility weight calculation that was given in the paper “Adaptive asset allocation: A primer,” Exhibit 2, by Adam Butler, et al (allocation = 1/n * 1% / [observed volatility of the asset over the past 60 days]).

    For the new adaptive allocation strategies, you calculate return and sharp ratio on the weighted combination of two tickers. For the volatility position-sizing calculation, do you use the volatility of the combined two tickers and reduce the position size for each of the two tickers by the same amount? Or do you calculate the volatility for each of the two tickers separately and then reduce the position size of each of the tickers independently?

    Thanks very much,

    Don Krafft


    Alexander Horn

    Don, thanks for the question. Indeed in the original strategies the equal volatility contribution was like the 1/n approach above, which is very robust, but ignores the covariance between the assets as only volatility at the single asset level is used.

    The new adaptive allocation with the modified Sharpe ratio works a bit different, as it is used as criteria at the ranking level already, not a later optimization layer. By scaling up the weighting of volatility though the attenuator, the algorithm basically becomes a volatility ‘minimization’. As the ranking depends on the weighted combination of the assets, the optimization is at ‘portfolio level’. Now, this is a heuristic approach, e.g. there is no assumption made of the expected volatility and covariance going forward, it is a momentum concept. You might have noticed in the screenshots of our new backtesting tool that there is an input field for “volatility limit”, which would be a second layer to target certain volatility levels. However as the modified Sharpe ranking already does a great job in the first step, we are not using the second layer currently to keep things simple

    As I noted two days ago in this thread, we’re working on a new core strategy which would always hold several assets, optimizing their weighting by a full fledged minimum (alt. mean) variance optimization, e.g. employing the covariance matrix to optimize volatility at the portfolio level. In this strategy we will also use a volatility targeting layer (at the portfolio level) and offer three different options to choose from.

    This is inspired by the great AAA approach by Butler, et al which you mention above. As they highlight in their also great “Dynamic Asset Allocation for Practitioners” series (must-read for a rainy day), the evolution from risk parity to more complex concepts does not necessarily mean better results, as these assume robust input variables like for example the covariance matrix. So our current focus is on how to create an easy-to-digest format for this new strategy with robust and transparent assumptions anybody can follow and gain confidence in.

    P.D. Just by the length of my response you see this is one of my favorite topics, happy to discuss further.


    Don Krafft

    Thanks very much Alexander. I’ll go back and spend some more time with the “Dynamic Asset Allocation for Practitioner” series. I’m sure we’ll have more discussion.



    Hi Alex,

    You mentioned about modified Sharpe Ratio during your presentation last year in Silicon Valley. What is it about?



    My question is how can I quantify if it’s better to:

    1) Long TMF
    2) Short TMV
    3) Long 3xTLT on margin

    I’m using Interactive Brokers, I know my fees and borrowing cost. Still I don’t know how can I calculate what is best option for me. Perhaps there’s no single best answer as it depends on volatility, but one can assume some volatility and check some scenarios.

    Update: I’ve just noticed this article (very fresh): http://www.signalplot.com/the-definitive-guide-to-shorting-leveraged-etfs/ but it still not giving me a full answer.

    But am I right that (given “low” borrowing cost)

    1) Long TMF is better only when TLT is going up without much corrections?
    2) Short TMV is better in ranging market and downward market?
    3) I don’t know under which conditions TLT is better than 1) or 2)…



    There are many online resources that discuss leveraged ETFs and how they behave vs normal ETFs.
    Most of the time it is most efficient to short the inverse levereged ETF. In this case shorting TMV. There is an exception and that is if TLT goes straight up.



    Thanks for an answer. I know that there’s a lot of articles regarding leveraged and non-leveraged ETFs, but I haven’t found one that could give any kind of formula, rather than general guidelines like “if your borrowing costs are low [but how low?], it’s most of the time better [but what does it mean “most of the time”] to short TMV than going long TMF”.




    Leveraged ETFs experience decay and do not track, long term, their corresponding ETFs at the stated multiplier. So TMF does not return 3 times the returns of TLT over two or three years. The decay is larger as the leverage gets bigger. The decay gets larger for inverse etfs. So the decay of -3x instrument is biggest than all others (-2x,+2, etc). That being said, it is volatility that causes this decay. If an ETF’s price path is 100,101,99,101,99,101,100 the +3x equivalent will not end up at 100 (0%) and will end up loosing money. If the ETF’s price path is 100,101,103,104,105, the +3x equivalent will not return 3x (ie 3×5% = 15%) but actually more than that. See the second paper for more info.



    “Leveraged ETFs experience decay and do not track, long term, their corresponding ETFs at the stated multiplier. So TMF does not return 3 times the returns of TLT over two or three years. ” yes, I understand this. I looks like it’s not that easy to calculate the decay, maybe I’ll do some backtesting, how it performed in different market regimes. I found one backtest on the web, it says that using 3xTLT give 3.1% better results (long term) than TMF, so if my costs are lower than that it may be worth to rather use leverage on TLT.

    PS. What bothers me a bit is that in all those articles there’s a mantra: “Leveraged funds are not for long term investment”. And if I have MYRS and Nasdaq100 in my portfolio I will have constant, long term investment in TMF. Sometimes it’s bigger, sometimes it’s smaller, but nearly for all time there is some allocation to it (I have to check if there was any period of time in which both of those two strategies were not invested in TMF at all, I doubt it). And then we know that in the long term TMF is not performing better than TLT. So it’s more a painkiller for a portfolio (reduces short term volatility) than a cure. But that’s a bit different topic, I have discussed it before.


    Petr Trauške

    Hello Alex,

    I have inquiry according to excelent Interactive Seminer, which you provide last time. Would it be possible to publish link so it can be accessible source at any time?
    My second question is: Do you recommend frequency of rebalancing portfolio in 14 days or in month cycle. I am using your predefined strategies.
    My third question is:
    I use within the Custom Builder Portfolio Strategy “Lev 2: Max 15% Volatility”. Which period is the best for rebelancing for optimal allocation – monthly, 14 days or on daily basis with using of QUANT Trader or leave for a constant ratios defined in the Custom Portfolio Builder for the duration of such investments, e.g. 5 years?
    In the case of using QUANT trader at the monthly rebalancing of the lookback period which recommends that you set – 6 months (126 days) or three months (63 days) for optimal adjustment of individual items in the strategy of “Lev 2: Max 15% Volatility”.

    Thank you for you response.



Viewing 15 posts - 1 through 15 (of 24 total)

You must be logged in to reply to this topic.