• Australian (ASX) Stock Market Forum

Hello and welcome to Aussie Stock Forums!

To gain full access you must register. Registration is free and takes only a few seconds to complete.

Already a member? Log in here.

Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypotheses

Discussion in 'Stock Market Nuts and Bolts' started by tech/a, Jul 16, 2015.

sentifi.com

Aussie Stock Forum Sentifi Top themes and market attention on:

  1. tech/a

    tech/a No Ordinary Duck

    Posts:
    18,961
    Likes Received:
    2,704
    Joined:
    Oct 14, 2004
    I'm on the search for any papers relative to any analysis F/A or T/A that has been rigorously tested
    The result positive or negative is un important. The veracity of the testing is however---must be Evidence based.

    If you know of any could you post up a link.

    Have a bit from Dr Bruce Vanstone and a few others.

    Howard Bandy??
    Deep State??
    Craft??
    Sinner??

    Anyone else interested in this stuff?
     
  2. craft

    craft

    Posts:
    1,999
    Likes Received:
    303
    Joined:
    May 3, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    http://papers.ssrn.com/sol3/results.cfm?RequestTimeout=50000000

    Just use the search facility and knock yourself out.

    Following the references in papers of interest becomes a great source for finding further papers.
     
  3. tech/a

    tech/a No Ordinary Duck

    Posts:
    18,961
    Likes Received:
    2,704
    Joined:
    Oct 14, 2004
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Yes thanks
    Know of that.

    But really hoping people will have come across
    ones which they have found to be of great interest.

    Recommended reading if you like.
     
  4. sinner

    sinner

    Posts:
    2,931
    Likes Received:
    1
    Joined:
    Oct 26, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Seminal paper (if you pick one it should be this): "Value and Momentum Everywhere" by Asness et all http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2174501

    Asness doesn't just research this stuff for fun, he is a big player who co-founded AQR Capital with like ~$120b FUM and good track record since 1998.

    But realistically, there are sooooo many papers which cover this stuff, it'd be impossible to build a list which does it justice. So I am picking a few random ones from my notes.

    Almost all of these papers are the basis for what happens at the worlds largest hedge funds and most of these names work at one or more of these funds as advisers:

    On the value front, nobody has produced more mountains of work than Fama+French, who have their own website full of all the data you could ever need to validate and reproduce their findings.

    I would say that allocators like James Montier, John Hussman Cliff Asness and so on represent the culmination of implementing all available information (TA and FA) into a single highly performant strategy. Then there are people like Eric Falkenstein who think everything is wrong and we should all just invest in low volatility (which is also hard to disagree with given his justification).

    "The Dao of Corporate Finance, Q Ratios, and Stock Market Crashes" by Mark Spitznagel on Tobins Q and valuations.

    "Cape Around the World: Update 2014 – The Relationship between Risk and Return" http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2470935
    Value Versus Growth: Australian Evidence by Gharghori, Stryjkowski and Veeraraghavan
    (you can google "ssrn value premium australia" to see a huge body of work on this)

    "The Cross Section of Expected Holding Period Returns and Their Dynamics: A Present Value Approach" http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2182628 (I really like this approach)
    "Enhancing the Investment Performance of Yield-Based Strategies" http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2051101
    On the T/A side:

    "Relative Strength Strategies for Investing" by Mebane Faber (and the various updates to this paper) e.g.
    "A Quantitative Approach to Tactical Asset Allocation"
    "Absolute Momentum: a Simple Rule-Based Strategy and Universal Trend-Following Overlay" by Gary Antonacci and related
    "Risk Premia Harvesting through Dual Momentum"
    "Does trend following work on stocks?" by Cole Wilcox and Eric Crittenden
    "Feasible Momentum Strategies Evidence from the Swiss Stock Market" by Rey and Schmid (academics, not traders)
    "Residual Momentum" by Blitz, Huij, Martens
    "MR Swing: A quantitative system for mean‐reversion and swing trading in market regimes" by Abrams and Walker (this one won some kind of award and AFAIK the system is still in use)
    "Momentum and Markowitz" by Keller, Butler, Kipnis http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2606884
    and so on. Basically, SSRN is your friend. Again, I would stress that this is merely a tiny random selection from my research notes, the amount of research out there is more than anyone could hope to read.
     
  5. sinner

    sinner

    Posts:
    2,931
    Likes Received:
    1
    Joined:
    Oct 26, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    It's important to add that

    "TA"

    and

    "FA"

    are crappy crappy labels to apply, since they both encompass a huge range of competing strategies.

    As an example, quantitative analysis of timeseries is not just about trend following and momentum, but alpha can be generated from mean reversion. Then there are differences based on time scale, trade size, etc. And again, differences across markets (some are trendier than others, FX, for example).

    Same for "FA", this is a huge field. It's not just about buying cheap stocks! The factors include, value, momentum (e.g. earnings momentum), volatility (e.g. earnings volatility), quality, liquidity, and so on. And again, differences across markets (you can buy EURUSD if you think it's undervalued on a PPP basis or something) and so on. Some fundamental shops do nothing but short fundamentally crappy companies (e.g. Sino Forest).
     
  6. ThingyMajiggy

    ThingyMajiggy

    Posts:
    1,627
    Likes Received:
    43
    Joined:
    Nov 12, 2007
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    This webinar is kind of along those lines tech, he goes on and shows that there is apparently no edge in fibs, S/R etc. and show's what there apparently is an edge in....plus other stuff. Might be of interest(if you can watch through an hour and 50min).

    Also goes on about randomness and how the markets are mostly noise and that even though we may know of some successful traders that could still be random/chance. Interesting thoughts....

     
    Last edited by a moderator: Dec 30, 2016
  7. sinner

    sinner

    Posts:
    2,931
    Likes Received:
    1
    Joined:
    Oct 26, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    I don't think there is much use in arguing about the effectiveness of S/R or fibs or whatever, since there is no replicable technique being applied here by traders. Some use different timeframes, some will trade the breakout of S/R, others the retrace, some will fade, some might use the 50% retrace, others might use the 50 day high or midpoint between 50 day high and low, etc. For each strategy, the chosen asset classes, leverage, capitalisation, position sizing, stop loss size, market regime and adaptivity, etc will play a huge role in long run robustness.

    I know and like an intraday strategy which trades breakouts of 10 bar highs/lows on 1 min charts. It works in 3 major assets, providing robust and statistically significant results over hundreds of walk forward trades. Without the position sizing algorithm implemented, it would not be robust over the long term. If I traded it on the daily it would die a long slow death and if I traded it on the 5 minute chart it would get chopped to pieces in less than a year.

    More thoughts from my own observations:

    Most trading strategies fit into broad groups, these broad groups can then actually be condensed into 3 major groups (4 if you have access to leverage).

    The return profiles for these 3 major groups look just like the return profiles for investing in a large cap company. They tend to go to the upper right hand of the equity chart over time, with some volatility along the way.

    These returns perform over the long term on a valuation basis, just like a large cap company might (because in aggregate, most businesses are long in the momentum or mean reversion of something or other - the most common thing being economic growth).

    That is to say, you will find that periods of underperformance in trend following strategies come immediately after periods of outperformance. If outperformance persists for a long time, then the proceeding underperformance will persist for a long time (and vice versa).

    As strategies begin to outperform, demand for the strategies increases, driving the future returns on those strategies down (e.g. crowded trade). As strategies begin to underperform, demand for the strategies decreases, driving future returns on those strategies up.

    I have definitely seen this unfolding in real time across different strategies. The best example is of course the GFC. At the lows of the GFC, demand for trend following/absolute momentum/relative strength/volatility adjusting/risk limiting strategies was at an all time high (witness the price of vol as proxied by VIX). Which strategy outperformed over the next 5 years? Buy and hold was certainly hard to beat! Then in 2013 after B+H continued to outperform, investors piled into low cost index funds (at valuation highs and volatility lows), is exactly when relative strength really took off providing crazy returns for sector and country RS investors. The current fad is figuring out how to stay leveraged long (in case bull run in large caps continues) while being protected from 2008 style losses, so my guess is that mean reversion strategies (e.g. 60/40 stocks/bonds, value, equal weight, permanent portfolio, 3 way system, etc) will outperform until demand picks up for those.

    It is possible to mix and match strategies for a very smooth equity curve. Forgive me for saying the name a lot but if you look at what John Hussman does in his stock fund is to buy value stocks (and did an excellent job at stock picking, outperforming Russell 2000 Value over >20y on raw and risk adjusted metrics) with a beta as close to the SP500 as 1, so that the portfolio can be hedged on a macro valuation/risk metric (breadth, credit spreads, etc)/volatility basis with index options. The goal being to optimise allocation across all 4 major groups (he uses options for leverage where appropriate).

    I have posted this link previously, but appropriate here:
    https://cssanalytics.wordpress.com/...et-allocation-lessons-from-perold-and-sharpe/
     
    Last edited by a moderator: Dec 30, 2016
  8. luutzu

    luutzu

    Posts:
    7,956
    Likes Received:
    1,051
    Joined:
    Apr 21, 2014
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Do we really need proof that buying $1 for 50 cents is profitable?

    The only difficulties is making sure it's actually $1. How do you make sure of that? Look carefully at the business.

    But then some would look at the business and say, it's not worth $1 now but it will be $1 in the future... So that's where all the fun comes in.
     
  9. sinner

    sinner

    Posts:
    2,931
    Likes Received:
    1
    Joined:
    Oct 26, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Everyone is making a bet one way or another. The idea is to derive information from data that reduces the possibility of risk while maximising possibility of reward. Some people do this using the numbers that come from financial reports, others use the numbers that come from the price of the asset (or related assets), some combine it all together. It's clear that alpha can be generated from all of those approaches.

    Evidence suggests that over the short term, price is difficult to predict and governed much more by investor risk preferences. However over the long term, investment securities can be expected to return a value approximate to their long term stream of cash flows (which can be forecast). So both matter.

    The point is not about proof, but rather about evidence.

     
    Last edited by a moderator: Dec 30, 2016
  10. tech/a

    tech/a No Ordinary Duck

    Posts:
    18,961
    Likes Received:
    2,704
    Joined:
    Oct 14, 2004
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Yes I think we do.
    in fact many many accepted and adopted trading and investing ideas and methods are used by just about everyone WITHOUT any evidence based validation.

    Just taking 2 widely accepted ideas.

    Buying Value
    Buying increasing Volume

    You me and 10000 others will trade the exact same signals multiple ways and offering just as many
    differing reasons for any failures.
    But where is the evidence that you/me or the other 10000 of us have a snowballs chance in hell to return a consistent profit.

    90% according the evidence I have on trader success-----fail---longer term.
    So those who adopt the two ideas above would be in that lot.
    I presume---(although I don't have the evidence) that some of the 10% also reside there.

    So If I'm going to be in that 10% or have ambitions of being there
    Id want Evidence---not just a belief that has been touted around for years---works.
     
  11. craft

    craft

    Posts:
    1,999
    Likes Received:
    303
    Joined:
    May 3, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    You'll only find evidence of regularities - not immutable laws. Unless you can predict the future you will not know how long the regulatory will be stationary. Some results with the appearance of regularities just result from the data mining process and are unlikely to persist beyond random others may have a cause but every man and his hedge fund is in a race to harvest the causal regularities which will inevitably arb them away. The hardest to harvest or most obscure will likely persist the longest.

    I suspect what defines your 10% is they "know" they are dealing with uncertainty and lack of evidence and come up with solutions to still cope. Continual checking of their beliefs against reality and adaption when reality dictates it necessary.

    The best that evidence based approach can achieve is a regularity that has cause rather than arising from randomness. But because the regulatory may not be stationary in the future as people try to harvest it you still have to manage uncertainty.
     
  12. KnowThePast

    KnowThePast

    Posts:
    457
    Likes Received:
    52
    Joined:
    Jun 3, 2013
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    craft, more and more I come to exactly this line of thinking. :xyxthumbs
     
  13. luutzu

    luutzu

    Posts:
    7,956
    Likes Received:
    1,051
    Joined:
    Apr 21, 2014
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Buffett gave a lecture in 1984 titled "Superinvestors of Graham-and-Doddsville".
    I just googled it and the copy from Columbia uni have the charts as evidence of the results achieved by the 9 students of Graham and Dodd that he personally knew of and worked with.

    Of the 9 investors, Buffett showed their results achieved over 20 to 30 years, and all nine managed to produced compounded annual growth of some 20% to 28%.

    What's more, they all achieve it buying different stocks. Some of the fund buy entire businesses as well as smaller numbers of stocks, some just buy stocks and took no control or influence over management.

    I don't think you can look at results like that and call it luck, or that those few guys are just geniuses no one else can hope to emulate.

    ---

    As an aside... there could be a misperception that Value investing is pure data driven, just the figures and no quality or qualitative analysis. The biggest misunderstanding, to my mind anyway, is the belief that the reported figures as collated by the research houses are all right and true and so effort should be towards predicting the future - be that interest rates or momentum or whatever.

    It's very comfortable to believe that since value investing is about the figures and all figures are the same as reported and so all that's needed is to scan for ratios and mechanically hit to big time.
     
  14. luutzu

    luutzu

    Posts:
    7,956
    Likes Received:
    1,051
    Joined:
    Apr 21, 2014
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Not really.

    While it's true that the market price cannot be predicted in the short term, it also follow that the price cannot be predicted in the long term too. Just logic.

    It's wrong to say that over the long term, the value will be corrected. That's nonsense.

    Over the long term, perhaps the market may come to agree with your assessment and upgrade or "correct" their pricing much more aligned to the fundamentals of the company - reflecting its true value in the long run.

    But that is very different from saying that over the long term, we'll be right, or the market will be right. It may, it may not.

    ---

    That is, when you estimate the approximate value of the company, your estimate is not that the price will be X in two years' time or 5 years' time etc. Your estimate is that its value is X to Y right now... and in time, the market will see that - you hope.

    When the market sees and agrees with you, say in 2 years time... the value of the company maybe have changed, may stay the same, or may have deteriorated. At that point, you must look to determine which condition it is in.


    Example. Say I value company A at $10. It is now selling at $5. So I buy it for $5.
    In two years' time the price went to $10. It does not automatically follow that $10 is fair value, or that my value has fully realised. The company may have won a few big contracts, may be become more efficient, new markets or sanctions have been lifted and it's expanding... So no, $10 is not fair value.

    The reserve could happen. That condition is terrible and it's about to be known by everyone. So $10 is really pushing it. etc. etc.


    To think that in the long term the value will be realised tend to lead to buying now for unknown future possibilities.
     
  15. sinner

    sinner

    Posts:
    2,931
    Likes Received:
    1
    Joined:
    Oct 26, 2008
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Completely regardless of what price you value some security at, the long term (hint, long term means a lot longer than 2 years) returns on any given security will be approximately equal to the stream of cash flows associated with that security.

    Just logic.

    The way you misinterpreted my statement and got into an argument with yourself about the misinterpretation is very cute.
     
  16. galumay

    galumay learner

    Posts:
    2,036
    Likes Received:
    596
    Joined:
    Sep 17, 2011
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    As sinner says, 2 years is not a realistic time scale, none the less you seem oblivious to the obvious point that the very events you postulate impacting upon the price do so because in the long term they will change the free cash flow to the business, which as we know is what detirmines a companies value in the long term!

    Obviously value is dynamic and its necessary to revist valuations and check for potential impacts both positively and negatively on free cash flow.

    I also personally think of value as range rather than a point, I am not sure how anyone believes its possible to calculate value of a company to a specific point, given the variability of a number of the assumptions required.
     
  17. luutzu

    luutzu

    Posts:
    7,956
    Likes Received:
    1,051
    Joined:
    Apr 21, 2014
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    What does "say, 2 years" mean? 2 years? Or any point in time, but let say it's two year for sake of argument.

    What does "say, $10" mean? Exactly 10 dollars even though I was using it in an example of an unknown company?

    Ok then.

    ---

    Cute or not, a smart person wouldn't pay for today what is essentially tomorrow's prices. Not literally "tomorrow" yea?

    I know it's a hard concept to understand, with discounting back to the present and all... but you do realise that, even with DCF, you are discounting back to the present... to the now. Given that that's what you're trying to do, how then can you go and say the price you're working out from DCF or anything is not meant for the now, but for the long term because in the long term it will work out as you predicted.

    Then in the long term it work out as you predicted, but then you say not really because things change.

    Like I said, if I'm a consultant or a hired help... I'll believe these nonsense too. Not only that, I'll tel the boss I need a couple more consultants to look at the US market, the European market, then the EU central banks, the Fed... Then build a team, then report weekly on prices I've forecast.

    Wow man... I'll be really impressed when people tell me they've put hours and hours into predicting tomorrow's prices, today... then by lunch time tell me tomorrow's prices is now this lunchtime's price... then when tomorrow come tell me the price is actually tomorrow tomorrow...
     
  18. luutzu

    luutzu

    Posts:
    7,956
    Likes Received:
    1,051
    Joined:
    Apr 21, 2014
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    So you spend countless hours predicting cash rates and all that to come to a range of value. Then if it turn out right or wrong or really wrong or really right... how do you know what causes it to be right or wrong? You got like half a dozen variables that could have cause it - and that's just on your modelling alone.

    The earnings might be up, why? Interest rates? maybe. Greek crisis? No? New competitor? All of them? To what degree?

    So DCF asks you to be precise, but then you can't be. But let's do it anyway.
     
  19. howardbandy

    howardbandy

    Posts:
    838
    Likes Received:
    92
    Joined:
    Jun 13, 2007
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    Hi Tech/a, and all --

    I am one of the people in your short list.

    I can reaffirm my strong interest in "evidence based" trading system development. A search for the definition of "evidence based" returns references mostly to medical studies and medical practice. So I'll go with your first request -- "rigorously tested."

    Rigorously tested, as applied to analysis of trading systems and to my thinking, implies application of high quality modeling and simulation techniques, including out-of-sample testing. As such, rigorous analysis begins with statements of intent and metrics by which performance will be measured, definitions of risk tolerance limits acceptable to the trader, analysis techniques that provide many out-of-sample data points, and guidelines for trading management.

    I believe you have copies of some or all of the five books I have written recently, all devoted to trading system development, all showing examples that are rigorously tested. Over 60,000 copies in circulation as of January 2015. Each book has its own website. Begin here:
    http://www.blueowlpress.com/WordPress/2-column/

    The best single book is my latest:
    "Quantitative Technical Analysis"
    http://www.quantitativetechnicalanalysis.com/
    or
    http://www.amazon.com/Quantitative-...?ie=UTF8&qid=1437055735&sr=8-1&keywords=bandy

    I have made several presentations to organizations including Australia Technical Analyst's Association, International Federation of Technical Analysts, and Market Technician's Association, all describing rigorously tested development theory and techniques. Many have been posted.
    Try this specific search:
    Google Bandy YouTube
    or this more general search:
    Google Howard Bandy
    or begin here:
    http://www.blueowlpress.com/WordPress/grid/

    Speaking specifically to your search for "papers." Papers, to me, implies academic quality publications. I am a retired university professor and dean. The academic community has been very slow to accept that trading is a reasonable alternative to long holding period investing, and papers related to technically based trading that pass academic filters are scarce. There are papers related to use of fundamental analysis in investment, but my own research and experience suggests to me that none of those techniques pass validation. Consequently, papers describing applications that are useful to us retail traders that have rigorous analysis will be scarce.

    I read in earlier responses in this thread to techniques such as Fibonacci, Elliott, Gann, etc. These are all ill defined, require subjective evaluation, and are prone to revision as additional data is received. None pass rigorous testing or statistical validation.

    Long term holding exposes the position to risk greater than most trades can tolerate. (My QTA book discusses risk related to holding period. My July 15, 2015, presentation to the IFTA entitled "The Four Faces of Risk" will be posted to YouTube shortly.)

    In my opinion, techniques that have the highest possibility of passing rigorous testing have the following characteristics:
    Are based on clear rules.
    Trade frequently.
    Hold a short period.
    Have a high ratio of winners to losers.
    Adjust position size dynamically as the model and data move in and out of synchronization.

    Best regards,
    Howard
     
  20. galumay

    galumay learner

    Posts:
    2,036
    Likes Received:
    596
    Joined:
    Sep 17, 2011
    Re: Looking for papers on Evidence Based Results for T/A and F/A methods/ideas/hypoth

    I think craft is right, you just dont read what other people write, then you reflect back something entirely different, and argue against that. Its very odd!

    I am going to follow his example in the other thread and leave you to it, good luck with what ever it is you do!

    On topic, I think evidence based paper that show methods dont work are more common than the reverse, so might be slim pickings. Nearly everything I have seen is not of rigorous academic standard, its just dressed up to look like it!
     
Loading...

Share This Page