Mean reversion system

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
LeviF
Roundtable Knight
Roundtable Knight
Posts: 1436
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Mean reversion system

Post by LeviF »

I have been working on a mean reversion system to compliment my trend following systems. I put one together that tested very well over the last 15 years, except it has basically blown up since August during the "financial crisis". My back testing & optimization data included through December and I started paper trading last month during a notable draw down. It is a short term system so I have already made 20 or so trades. The equity curve just keeps heading down, down, down. My initial thoughts are that mean reversions systems work until they don't, and that is why we use trend following systems...However, its also hard for me to believe that this "crisis" as of late has been that different from the Pound Collapse, Peso Crisis, LTCM, SE Asian Crisis, etc.

Does anyone here trade a mean reversion system? Have you seen similar performance ever? Any input?

btw- I'm trading forex.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

I have always found that it is hard to put a meaningful stop loss for mean-reverting systems - you know, the type of stoploss that reduces risk, boosts reward/risk ratio, while not compromising reliablity. Since the entry of a mean-reverting system is aganist the immediate orderflow of the trend that you try to capture, if you put a too tight a stoploss, chances are you will be stopped out; if you put a wide stoploss, then the reward/risk ratio is compromised; finally, if you don't put a stop loss, you are risking ruin. Also, continuation patterns seem to be a lot (2 to 3 times) more reliable than reversal patterns.

I previously traded a few mean-reverting systems, all resulted in heavy losses in the end - many mutliples of the whatever small profits they generated in the beginning. So now I just stick to trend-following systems.

The trend is your friend!
Last edited by TrendsCatcher on Tue Jan 13, 2009 7:00 pm, edited 1 time in total.
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

My experience goes in the same lines. Yet, sometimes, I still feel tempted to try, research and experiment with some "mean-reversion" or top-bottom-picking system. The lures of having the power to call the next move... of feeling god-like... They never go away.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

It's probably an innate human bias to call the "counter trend". Time to reread William Eckhardt's chapter in the New Market Wizards.
LeviF
Roundtable Knight
Roundtable Knight
Posts: 1436
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF »

I have no desire to be able to "pick bottoms", i just want something that works when my trend systems are not...
DeanoT
Roundtable Fellow
Roundtable Fellow
Posts: 52
Joined: Sat Apr 15, 2006 7:37 am
Location: Melbourne, Australia

Post by DeanoT »

The difficulty with being a trend follower is that you will never be the "genius" who picked the bottom of the bear market, or the top of the bull market. Only the weaker get traders will get to make that claim (on their fifth, tenth or twentieth attempt). Those same weaker traders will then ask you how you made 20%, 50% and 100%+ in years where the markets were considered "horrible". And when you tell them that you follow trends, they look at you like you are some kind of crazy person.

For me, it ultimately depends on whether your motivations are to make profits from your trading, or to demonstrate how "smart" you are, and often these two outcomes are mutually exclusive. I think we all possess these motivations, but to different degrees, which ultimately determines whether we are willing to put our ego aside, and let the weight of statistical evidence demonstrate that following trends is where the money is, or bury our heads in the sand and continue to predict, knowing that eventually we will be right, and hope that everyone forgets all our poor predictions.

The latter method doesn't do much for the bottom line, however.
rabidric
Roundtable Knight
Roundtable Knight
Posts: 243
Joined: Mon Jan 09, 2006 7:45 am

Post by rabidric »

from a certain point of view, trailing exits are a form of counter-trend system overlaid onto a slower trend following system. you can dissect and recombine a system into all sorts of component parts that vary in how they add and reduce size in various directions. typically though the total equity curve will rise during periods of trending in markets. this labels the meta-system as "trend-following"...

er i don't know where i am going with this, but i think i am trying to say, even if you combine all sorts of trend and counter trend systems on different timeframes etc, you still end up with an equity curve that has drawdowns, and that with hindsight these drawdowns will seem like they could be diversified against, but the act of implementing a sub-system(counter-trend) to hedge those drawdowns simple raises exposure to a different kind of weakness to a slightly different form of market behaviour. you end up chasing your tail...

the worst is when your "trend following" meta-system goes into a drawdown, and your backer/investors turns around to you and says:
"why not use some of your capital to run a counter-trend system that would hedge your drawdowns"
...and you know that there is little chance you will be able to get across the relevant arguements to support your position without it going over their heads or them being unreceptive to it cause their "other fund/manager/CTF made new equity highs in these markets, so why can't you?".

*le sigh*
LeapFrog
Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

Post by LeapFrog »

Keith Fitschen (of Abberation fame) gave a presentation a few years back where he presented a study on trend versus countertrend trading approaches. He used over 7,000 equities from 1980 to 2002 and concluded that trend following worked better on commodites and counter trend worked better on equities (I don't remember the metrics he used for commodities).

But, if "markets" are choppy or sideways around "70 percent of the time" as is often said, it would stand to reason that a counter trend approach during those times would be the way to go - during those times.

I confess though, even though I will trade intraday and one/two day swings, I still go for breakouts - simple things like b/o from compressed ranges, breaking of pivot points in the direction of the longer trend, that sort of thing.

I'd rather not though. I'd rather just put my orders in at night and go ride motorbikes all day (maybe I should be designing more trading systems).

Read that book "Millionaire Trader" http://www.amazon.com/Millionaire-Trade ... 0470049472 recently. Not very well written and no believable backup evidence of trader's results (one of the them subsequently was charged with fraud) but a couple of chapters were interesting. One fellow is a scapler doing hundreds of trades a day (holy crap) and went at it very workman like - claims to make a good six figure income with only 10 losing days a year. Here is an interview with him:-

http://www.youtube.com/watch?v=lK58oT5GTGE

I know his style isn't what most of us around here would be interested in, but it is another approach.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

Ed Seykota has shed some light on this issue on his website:

"Trading against imbalances presupposes an "equilibrium value" and a negative feedback loop that intends to restore the state to equilibrium.

You might consider the following processes:

micro-organism growth in a culture broth
virus spreading in a population
avalanche breakdown within a dielectric
nuclear chain reaction
multi-level marketing
Moore's law
compound interest
spread of cell phones

These processes proceed, in the early stages, according to the exponential growth characteristic of positive feedback loops.

Some People Like to Bet

on negative feedback

and on a return to the norm.


Others Like to bet

on positive feedback

and on Trends."

This discussion of the nature of trends and counter-trends is the best I've seen. Interestingly, Soros also had similar discussions about "positive feedback" mechanisms in the market in his books and his various interviews. the Book "Fooled by randomness" also has an excellent discourse on this feedback mechnism.

My personal experience is that counter-trend is psychologically easy to accept, but seems to have negative expectancy (in the long run), trend-following is hard to do, but seems to positive expectancy: but one needs to be extremely selective with stocks when applying trend-following techniques though, the vast majority of them in the vast majority of time are not worth playing at all, must be in a bull market, top leading stock from a superb group, etc. Only in that type of enviorment, stocks demonstrate beatufil trends as in AAPL in 2004-2008, TASR in 2003-2004, HANS in 2005-2006, QCOM, JAVA, SCHW and slew of other in 1999, DRYS in 2007, POT in 2006-2007), stocks in general are not as "tendy" as commodities, but there are a lot of them, and some of them can easily go up 10 times, some 100 times, and a few 1000 times (eg, CSCO, DELL) - they are more volatile and has a lot higher implied volatility compared with commodities. Since trend following is essetnially simulating longing a call option (or put is shorting), and you self-broke the options's pricing (by using wide or tight stop losses), in stocks, you can "buy" the option a lot chepter than in commoditis, and the reward/risk can be dramatically higher (only in extremely selected situations), think about having a stock going up 10 times (10 bagger) while risking only 3-5% of its starting value, Reward/Risk = 1000%/5% = 200! It is hard to find this type of muliple hundred R-muptile trading opportunities in commoditions. On the other hand, leverage is great in commodities, and once a major trends is in place, you can add, add, add, until... you know what I mean.

Positive feedback situation doesn't occur most of the time in all the market, it is the few situation that they deveopment in some of the markets made trending following rewarding in the long run, whereas while counter-trending trading is rewarding psychologically (becuase the high win/lose ratio), eventually, there is a "tide" (Soros's terminology, also Buffet's favorite), the counter trend traders, especially those without a fail safe mechanims in the means of a stop-loss ("naked swimmers?) will be carried away, and years of winning, and typically more (I have been there before), can be taken back by the market in one wallop.

I don't claim expert on this, and I am still thinking about all these all the time, so any opinion, reinforcing or contradictory, is welcome.
adamant
Roundtable Knight
Roundtable Knight
Posts: 118
Joined: Mon Aug 25, 2008 1:20 am
Location: Norway

Post by adamant »

Trendscatcher:

I very much agree with you that at first, countertrend thinking was easier to accept. But after reading Soros on the subject, I have found a philosophical basis for embracing trend following as well, and to me, it helped to see a philosophical basis for trend following also. Because that is what reflexivity is, in my view.

Maybe some people out there don't need this sort of basis, and are content to observe through empirical testing what the best approach is. If so, I give them credit for this very rational approach. Personally, I have found readings on Soros' theory of reflexivity to be a good supplement on the philosophical side of things.
CRMCM
Full Member
Full Member
Posts: 20
Joined: Sat Apr 04, 2009 8:16 pm
Location: New York

Post by CRMCM »

I definitely agree that positive feedback loops drive trends - the big massive trends that make everything worth it. It's surprising that more people don't see the "obvious". All the better for those that do ;)
Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat »

CRMCM wrote:I definitely agree that positive feedback loops drive trends - the big massive trends that make everything worth it.
Can you elaborate on that? What does "positive feedback" mean when applied to a market?

Let's take crude, going from 70 to 150 in 13 months, then from 150 to 50 in 5 months. I thought that, broadly speaking, in the 13 months demand was greater than supply, and demand less than supply afterwards. It takes a long time to fundamentally change demand or supply in commodities, that's what makes big massive trends to my mind. Additionally it helps, that for some (like crude) demand is not very price sensitive (in the short and mid term).
CRMCM
Full Member
Full Member
Posts: 20
Joined: Sat Apr 04, 2009 8:16 pm
Location: New York

Post by CRMCM »

My take is that feedback loops cause big trends but not all big trends have feedback loops. Some trends may or may not have a positive feed back loop but many do. When I see a trend develop I look to see if there is a positive feedback loop in place, if so, I want to participate. This is how I'm different than many others that look at price alone. I like to see fundamentals corroborate a trend whereas, from my observations, most people on this forum look at price alone. Call me a big trend hunter I suppose.

A good recent example of a feedback loop is the financial stocks. As the price goes down customers lose faith, as customers lose faith they "diversify" who they do business with which deteriorates the fundamentals, and away we go. There are a slew of other feedback loops with the financials so I'm just highlighting one (willingness to loan <=> value of collateral is a big one). This was a massive trend that was easy to spot if you saw the feedback loop at work. You need something to fundamentally stop the feedback process, alas Government stepping in to stop/slow the process. That was time to cover, the feedback loop played out.

As for your question on oil and what the feedback loop was. There's lots of speculation on what caused this (no pun intended). There was a really good post on here somewhere about how "buy and hold" allocators have distorted commodity prices to the upside, basically large institutional investors that are allocating money to the commodity asset class and hoarding per say - this shifted the demand curve only for it later to be shifted due to the economy. This could be the mechanism that caused such a run (as price goes up more want to allocate), however, there is a big debate about this so it's just a hypothesis that many share. I would love to hear others thoughts as to whether a feedback loop was at work.

I hope this helps clarify.
LeviF
Roundtable Knight
Roundtable Knight
Posts: 1436
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF »

I am still working on this project. Recently, I have put a counter-trend system together that by itself has negative expectancy, but when combined with my trendfollowing systems, it improves the overall performance. I guess this is basically what I was looking to do, but I dont really like the idea of trading a negative expectancy system...
Michael
Senior Member
Senior Member
Posts: 31
Joined: Wed Dec 31, 1969 7:00 pm
Location: Vienna

Post by Michael »

In my experience chances are high that the negative correlation of the counter trend system will turn positive in the future and you end up with just the negative performance
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

It seems to me that CRMCM's need for a "positive feedback loop" to confirm a trend involves a disctrionary decision (no criticism intended). One that I would be unable to make; for me it would destroy my purpose in following a mechanical system.

On "negative expectancy" I have much the same reservations about trading short. It is true that my system has profited (just) from trading short over time thanks to the exceptional periods like that we have just had. However, I have recently been testing a LTTF system which produces very similar measures of goodness without the need for trading short, which I rather like. I don't like adding barely profitable systems just for the sake of straightening the equity curve and share the fear that such systems can easily tip over into severe negative performance.
CRMCM
Full Member
Full Member
Posts: 20
Joined: Sat Apr 04, 2009 8:16 pm
Location: New York

Post by CRMCM »

AFJ Garner wrote:It seems to me that CRMCM's need for a "positive feedback loop" to confirm a trend involves a disctrionary decision (no criticism intended). One that I would be unable to make; for me it would destroy my purpose in following a mechanical system.
No offense taken, its all a matter of perspective.

I should probably clarify what I'm meaning. I don't require a feedback loop, but if I see one I'm much more likely to participate in the trend, so it's not so much a requirement as it is a desirable feature. I do use discretion and you will never hear me claim to use a "fully mechanical system". I view systems as filters that I know are giving me trade ideas that have positive ev.

I understand that if you use miss the big trends performance can be hampered (the argument against discretion). However, most big trends are caused due to fundamentals so I should catch most of those but limit some of the whipsaws. I still let the trend run and use systematic trade management, it's the entry where I use discretion.

I have a background that has led me to blend different disciplines (trend following and fundamentals). I started out trading stocks when I was 13 and shortly thereafter purchased the Turtle System, I've been a trend follower ever since. Did Big 4 Audit, M&A and now work at a hedge fund (value investors that do prop research, invest anywhere in the capital structure). So I'm sort of acclimated to being a trend follower but also using discretion and fundamentals. I believe there are a lot of institutions that believe in trend following but still can't quite part with their money unless there are fundamentals involved (on the equity/debt side), for better or worse.

Back to feedback loops, Gold would be one currently on going. As the US issues more currency, other governments issue more currency to keep trade balanced, gold increases due to the debasing, people realize gold is going up and lose more faith in the system, and around we go.
TrendsCatcher
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Thu Jan 10, 2008 3:28 pm
Location: Southern California, USA

Post by TrendsCatcher »

I think that one of the reasons that some pure trend-followers (eg, Richard Dennis and his cohort) do not want to miss even a single big trend is that big trends are relatively rare. I found that these people mostly trade only futures markets, which do not have that many markets, especially when you factor in correlations, you have no more than a dozen markets, each individual market may give you a good trend once in a decade, so missing one is devastating if one trades ONLY these markets.

This premise is not valid anymore if you trade invividual stocks, in a clear bull market, there are so many tradable stocks that double, triple, or go up 10 times, a few 100 times, nobody can catch them all, just need to trade a few correctly. Miss one or two or a dozen, no big deal, who cares if you missed the big trends in AAPL, TASR, TIE, DRYS etc in the last stock bull market, if you have caught the just-as-big trends in HANS, RIMM, POT?

So, I can see the benefit of adding in fundamentals in the process. Disctrionary decisions can be used to select the vechicles that can potentially move a lot, especially up a lot. True, you will increase type II error, but in return, you decrease type I error.
CRMCM
Full Member
Full Member
Posts: 20
Joined: Sat Apr 04, 2009 8:16 pm
Location: New York

Post by CRMCM »

TrendsCatcher wrote:I think that one of the reasons that some pure trend-followers (eg, Richard Dennis and his cohort) do not want to miss even a single big trend is that big trends are relatively rare. I found that these people mostly trade only futures markets, which do not have that many markets, especially when you factor in correlations, you have no more than a dozen markets, each individual market may give you a good trend once in a decade, so missing one is devastating if one trades ONLY these markets.

This premise is not valid anymore if you trade invividual stocks, in a clear bull market, there are so many tradable stocks that double, triple, or go up 10 times, a few 100 times, nobody can catch them all, just need to trade a few correctly. Miss one or two or a dozen, no big deal, who cares if you missed the big trends in AAPL, TASR, TIE, DRYS etc in the last stock bull market, if you have caught the just-as-big trends in HANS, RIMM, POT?

So, I can see the benefit of adding in fundamentals in the process. Disctrionary decisions can be used to select the vechicles that can potentially move a lot, especially up a lot. True, you will increase type II error, but in return, you decrease type I error.
I agree. After doing some research and valuation you can say whether you are comfortable with a given equity and that it's not fundamentally overvalued (I like GARP but would allocate more to one that had a large margin of safety) and that you like the sector given your outlook. If you miss some, not a killer when judged against your benchmark.

For commodities, and this is not my expertise, but I would think it's much harder to say whether something is under or over valued (unless arb is possible). What is the right price for corn? I think supply and demand is better reflected in commodity prices than equities; in equities you often get price dislocations, misunderstood securities, and behavioral biases that influence prices.

I wouldn't use discretion if I were trading commodities for your exact reasons.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

TrendsCatcher wrote:I found that these people { pure trend followers } mostly trade only futures markets, which do not have that many markets, especially when you factor in correlations, you have no more than a dozen markets.
Modern practice among managed futures firms using trend following strategies, often includes large portfolios containing >100 and occasionally >300 futures markets. Have a look at Transtrend, Winton, etc. In my own personal futures trading, I "trade everything including the kitchen sink" too. Why? Among other reasons, because my backtest simulations show that results get better and better the more markets I add. Even when I add markets that human intuition would label "obviously redundant" -- like adding Brent Crude and ICE WTI Crude to a portfolio that already includes NYMEX Crude Oil -- I find that the results are improved. Really.

Let me suggest some experiments you can do in your own home, that may produce a few Eurekas. First, measure the correlation coefficient(s) of the daily changes in PRICE of a few futures markets that you believe are highly correlated. (like: the Soybean Crush, or all the Grains, or Crude + HeatOil + Gasoline, or all the different US Treasury Note futures, ...) Now stream those historical price data series through a mechanical trading system and measure the correlation coefficients of the DAILY RETURNS from trading them with your system. (Since the Markowitzean "assets" being combined are in fact the profits and losses which are the output of the trading system). Compare the correlations among prices, versus the correlations among trading system returns. Say "Eureka" if the correlation among trading system returns is less than the correlation among prices.

The trading system's outputs are less correlated than its inputs? True or False? ...insert remarks about entropy here...

Next, rummage around in the Blox Marketplace and find the Random Portfolio manager donated by Nickmar and Jake Carriker. Use it with a large universe of at least a hundred futures markets, to (A) run a trading system 1000 times, on 1000 different randomly-chosen portfolios of 20 markets; (B) run that same trading system 1000 times, on 1000 different randomly-chosen portfolios of 40 markets; (C) run that same trading system 1000 times, on 1000 different randomly-chosen portfolios of 80 markets. Look at the means of the three different attempts; which one has the biggest mean Sharpe Ratio? the biggest mean R-Cubed? the smallest mean Longest Drawdown? Say "Eureka".

Also see (ref.1) , (ref.2) .
Post Reply