By Lowell Herr
Most, if not all the links found within the Feynman Study posts are broken. Instead of attaching each Word file that goes with each blog post, here are all the associated documents in one zip file. See the link below.
http://ppl.ug/7tCvRAsP1Lo/
Subscribe to get the latest posts sent to your email.
Portfolios coming up for review. Kepler, McClintock, Huygens, and Bohr are the critical portfolios next up for review. Non-scheduled portfolios may be reviewed. If you are a new user, check the posts you missed. Links to Random Posts are found in the lower right-hand footer or just to the right of what you are now reading. Most popular posts are found in the lower left-hand footer.
Check the Forum for more detailed information. If you wish to begin a financial discussion, use the Forum.
This blog is free to all who signup as a Guest.
Subscribe now to keep reading and get access to the full archive.
Lowell/David:
Considering the extensive documentation of the Feynman, what are the trade offs of using this as a tradable strategy ? Thanks
Jack B.
John,
Think of The Feynman Study as a forerunner to the “Rutherford 10” and other momentum models. Last summer David, Ernie, and Herb spent a lot of time back-testing various models, discussing robust methods, determining how to reduce risk, etc. and out of that work emerged the Kipling Tranche 2.5.1 spreadsheet. There are a few key ingredients that make, what I call the Tranche Momentum Model work.
1. Select ETFs that cover the globe and all asset classes. Yes, we miss private equity, but we cannot corral them all. Our list of asset classes match almost one to one with those recommended in the book, Adaptive Asset Allocation.
2. Include a number of low correlated ETFs among the ETF group. This is why we include TIP, TLT, DBC, GLD and in many cases, BIV or some other bond equivalent.
3. The tranche option is introduced to avoid luck-of-review-day. Ernie pushed this idea.
4. Keep the look-back period confined to one year or less.
5. Stay away from ETFs if priced below their 195-Day EMA.
6. Our analysis indicates it is best to stay invested in the top two recommendations. The Tranche Model will usually include more, but focus in the current top two recommendations that emerge from the SS.
7. Use the Position Sizing worksheet as a way to control the overall risk of the portfolio.
Lowell
John,
I’m not sure which strategy you are referring to. The Feynman Study is basically a review of all the major strategies I have used as I have gone through the “Investment Learning” process. So, it includes Strategic Asset Allocation (SAA) models (Buy and Hold, maybe with periodic rebalancing adjustments), Mean Variance Optimization (MVO) models using Efficient Frontier analysis and Momentum Models. The study was designed to provide a comparison of the different strategies when applied to the same list of assets and over the same period of time. The comparative performance and pros and cons are covered so that you can make your own determination as to which strategy best meets your own personal needs. Not all investors will chose to use the same strategy for various reasons. My (personal) preference is to use a momentum strategy, recognizing that this does not always work as well as a different strategy might work – but unless I know ahead of time how the market is going to behave it is impossible to know which will be the “best” strategy over the next N months.
David
For members looking for access to the Feynman Study, including Appendices, all files can also be found in my Dropbox Folder at https://www.dropbox.com/sh/vimiajlv3359v75/AAC5Pnjaq9UAJjYpu5CwADO0a?dl=0
Although not particularly relevant to the Feynman Study, the folder also contains (some old) instruction files for the Ranking Spreadsheets for anyone that may have missed them and who may find them useful.
David
Lowell / David,
Thanks for the very informative responses.
My interest in the Feynman Study is primarily due to my focus on MaxDD and the longer available charts / data. I have been out of touch with investing/market for the last 15 months and I am just starting my catchup effort. Obviously, I have missed a lot of important updates and additions. Please point me to the necessary versions to catch up, including TLH if it still relevant.
Since I am 11 years into retirement my investment horizon is around 3 to 5 years. This DIY retiree cohort is likely to expand greatly in the near future as more of the Boomer Generation moves into retirement. I suspect they will will also rely heavily, if not exclusively, on MaxDD in guiding/creating their investment strategy. Thus, so far, the Feynman Momentum Portfolio with a MaxDD of -12.6 % (Section 5.2, page 25-26) seems most appealing, especially if this stat was calculated on daily prices.
Thanks for any catchup info you can provide.
Jack Brennan
John,
Although a 12.6% MaxDD is a reasonable number based on the backtests I would keep in mind the “luck of review date” effect that we have documented in a number of posts using Monte Carlo techniques to “stress test” the basic momentum system. I would suggest that 15% might be a safer expectation (with an absolute maximum of 18%) and that decisions should be based on these values – this assumes that the portfolio is always 100% invested.. Use of the Position Sizing sheet in the Tranche workbook is a reasonably easy way to “fine tune” this risk level and usually recommends a portion of the portfolio be held in Cash or “risk-free” asset. At the end of the day we need to balance potential risk and return.
Regarding the calculation of MaxDD in the Feynman Study, although momentum is calculated based on daily data the MaxDD is month over month DD (at the review date) – hence my suggestion to assume the possibility of a slightly higher day over day maxDD. Some of Ernie’s MC backtests may use day over day maxDD – I don’t remember the details of his code.
As I see it, the momentum methodology keeps us out of those significant drawdowns associated with major bear markets (e.g. 2008 Financial crisis).
David
Jack,
The two critical spreadsheet should be in your mailbox.
Lowell
David / Lowell,
Considering the possibility that market prices may not be a Normal (Gaussian) distribution, would a bootstrap methodology be a better mechanism for stress testing than Monte Carlo ?
Jack B.
Jack,
I can’t answer the above question as I don’t know what a bootstrap methodology is. Perhaps Herb, Ernie, or David can answer this question.
Lowell
Since the way the MC tests have been performed does not assume any distribution function – simply a random selection of start/stop dates I don’t think it matters whether a MC test or a bootstrap test is used.
More practically, it is probably easier to code a MC test than a bootstrap test – especially since we aren’t considering a re-ordering or sequencing of data points – just where we break the known historical sequence to take measurements.
I could be wrong here since I haven’t looked at bootstrap algorithms too seriously.
David
Thanks to Lowell for the 2 files and David for info on randomized Start / Stop dates. Once I read Ernie’s doc I will probably agree with David’s take. However, IF there is still a Turn-of-the-Month effect, the the method used for Start/Stop randomization might be relevant.
Jack B.