What’s causing the underperformance in my R2G Ports?

I have certainly been disappointed with the performance of my Ports. Much of the underperformance was caused by the 3 market timing whipsaws since they were launched. For example, the 3 whipsaws in my R2G SmallCap Port ( https://www.portfolio123.com/app/r2g/summary/1045239 ) were June 24 > July 15, 2013; Oct. 14 > Oct. 21, 2013; & Oct. 20 > Oct. 27, 2014. During these 3 periods my port was in cash the S&P 500 gained 6.96%, 2.2%, & 4.26% respectively for a total of over 14% lost gain for the R2G Port. The effect of the whipsaws is even more pronounced than the missed S&P gains. After a quick check of the sell rules a few of the 5 stocks would not have been sold since launch. The 3 whipsaws caused 5 stocks to be sold and later replaced by 5 new ones 3 times causing extra fees and slippage expense while losing out on those stocks excessive gains.

My approach to develop a sound market timing system in 2009 was to avoid most of the >50% drawdowns of the S&P to save capital while not hurting annual return very much. In hindsight, we can see that the rules I developed then resulted in these small whipsaws that caused part of the Port underperformance since it was launched. There were also 3 small whipsaws in the Port between the time I developed the timing system in March 2009 and the launch of the Port in March 2013. They were in June 2009, Nov. 2011, & Dec. 2011. However there were also 2 times that the timing system saved significant losses in the same time period, a small saving of 3.5% in Oct. 2012, and a much larger savings of a 16.47% drawdown in July > August 2011.

The biggest cause of the underperformance is that for this ranking system, these value stocks need some time for additional momentum traders to buy in for the prices to really grow. The ranking system first finds pure value stocks and then looks for the beginning of investors &/or momentum traders buying in, moving the price and volume. When that occurs the Port buys the stocks and holds them until they have run their course by falling significantly in rank. However, when the stocks are sold too soon because of market timing, they don’t have time to grow.

The problem with any timing system that saves drawdowns during deep corrections and recessions/depressions, but loses small gains during small corrections due to whipsaws is that the subs only see the out of sample performance with their real trades. Since there has not been a deep correction, recession, or depression since the port was launched, the market timing rules are negatively effecting the out-of-sample performance compared to the S&P. That also causes most of the subs to bail to better performing Ports (I don’t blame them for that). If there had been a deep S&P drawdown (I don’t wish that on anyone) and the Port had saved all the subs a LOT of money, everyone would be applauding my conservative trading system. Alas, I am stuck with poor out of sample performance.

Another cause of underperformance is the ultra-conservative way P123 calculates the R2G Prices for buys & sells. P123 uses the Hi + Lo + 2*Close)/4 +or- slippage. That caused significant underperformance relative to buys & sells near the open, which is what my original Sim & Port were based on (before P123 changed the rules), and what I have discussed in the Forum over the years. The differences between running the Sim with next open, Avg of Hi & Low, or next close is significant. I re-ran the Sim starting on March 2013 (the date I launched the Port) through today. The total return using next open was 27.5% with a Max DD of -13.6%; the total return using Avg of Hi & Lo was 25.8% with a Max DD of -14.3%; and the total return using next close was 19.5% with a Max DD of -15.1%. So the excessive underperformance using next close vs next open is 8% in 21 months. I always buy and sell close to the open.

Now, if I re-run the Sim starting in March 2013 using next open AND turn of the market timing rules to remove the whipsaws (which also allows the Port to hold the stocks longer) The total return becomes 68.7% with a Max DD of -15.1%. That is more than three times the 22.8% total return shown for my R2G Port, and more than double the S&P since launch.

The bottom line is that this Trading System, minus market timing, significantly outperform the market during the out of sample period using next open instead of the P123 buy & sell prices. What I must do now is rethink the timing rules to reduce the whipsaws while trying not to significantly increase drawdowns during the recessions. That, is not a trivial effort without data mining.

I do not know if anyone would use this. But I would like developers to have the option of submitting an R2G port without market timing and then submit the same port with market timing and not be penalized for the second port as far as the number of ports allowed.

This could be useful for different developers for different reasons. Certainly, giving more options and transparency to the subs would be helpful and probably generate increased interest.

A port’s (out-of-sample) performance is what it is. It is based on its rules set out. There it does not help to mention a better performance for the case that the rules would have been different.

And it also does not help, to offer potential subscribers a second (x-rd) variant, and to point out the good performance of the respective best one.

However, the creator of each R2G should indicate that the outperformance of the backtest in the worst case, for example, occurred only after 3 years.

Denny,

  • Market timing is the weak link in many ports. Indeed some market timers offered in R2G are pure exercise of curve fitting and hence have no predictive value whatsoever. Getting whipsawed out was my exact case against ma crossovers in the discussion we had a few weeks back.
  • The R2G execution price is very strange. Why would the close price be counted twice? Including the average of high and low is also unfavorable for volatile stocks. Micro/small caps ports have been hammered because of this. I wish the execution price would be calculated as average of open and close, since these are the times with most volumes. This would surely be more realistic.

Talking about market timing I am also wary about these new models who short SPY some or all of the time with as much money than they have on the long side, effectively making a 100% long 100% short portfolio, which I have never heard of in this industry. It seems very risky to me.

aurelaurel,
that was the discussion of my MAC-US model where you could not find any predictive value from the system. Other people like Denny disagreed, and here is an in-depth analysis of an outsider. In October 2014, Jeff Swanson of System Trader Success investigated the question: Is the MAC-System Overly Optimized?
http://systemtradersuccess.com/mac-system-overly-optimized/
He concludes:
[i]"It appears this trading model is not overly optimized. Through my simple testing I was able to demonstrate that the core of the MAC trading system shows profitability over a wide range of values. In fact, some combinations produce better results. The default values do not appear overly optimized. Note, many different values appear to work just fine.

It’s my opinion the MAC system will likely work into the future. It works because it positions a trader into bull trends as seen from 2009 to today (late 2014) while more importantly, it gets a trader out of the market during prolonged bear markets such as seen in 2007-2008. The exact parameters don’t matter! You can use a 260 SMA or a 180 SMA for the major trend filter. You can enter the market using a 40 EMA or a 20 EMA. You can use a factor of 1.001 or .0098 as the buy threshold. They all produce positive results which speaks to the robustness of the system."[/i]

Georg

I think it is the courtesy of the model designer and his motivation to build trust to disclose information about his market timing rules, if the model heavily relies on that.

People who curve-fit will always have short-term advantage over other model designers. But curve-fitting can be achieved with any factor.

But still - disclosing too many information should not be obligatory as we don’t want to restrict ourselves too much.
I think that “the market” (designers, subs, the community) will take care of sorting the wheat from the chaff in the long run.
Instead, more attention should be directed towards oos data which we get more of everyday.

Best,
fips

I apologize if it sounded like I targeted the MAC system in particular. Ma systems do well for setting a ceiling on the max drawdown. They do this well since they will go to cash if the market is in free-fall. In terms of absolute returns or even risk adjusted returns it’s a different story, especially for the former.

I believe MA systems have no predictive value because their logic is very weak and is very dependent on the parameters used.
The first argument behind MA systems is “buy what goes up and sell what goes down, it’s proven to work”. This is of course very wise but stocks that go up can go down temporarily and vice versa, thus the short-term random fluctuations can cause a collapse in the strategy. Even with parameters such as 200 and 50, there is no guarantee. Stock prices can remain erratic long enough to whipsaw any ma strategy.
The second argument is “a lot of people do it, it’s a self-fulfilling prophecy”. This is a very simplistic view of capital flows in the equity market. First off, do trend followers account for a significant portion of total capital flows? Of these trend followers, how many uses ma crossovers? Of these how many use your parameters? What reasons do we have to think that these folks will not be cancelled out or outweighed by people believing in mean reversion? Jeremy Grantham for example, who is to be taken seriously with over 100 billion AUM, is swearing by mean reversion. How many hedge fund managers are swearing by ma crossovers?

Denny,

I’m sorry to hear about your whipsaws. I find them to be more painful than corrections.

A combination of the below may accomplish your goal of minimizing whipsaws and avoiding damage during bear markets and corrections. However the downside is that holding a fixed allocation to 2) will reduce your short term returns during bull markets and thus make your system less attractive to performance chasers on r2g.

  1. A slow market timer to reduce the probability of debilitating financial damage during bear markets.
  2. A small fixed allocation to bonds, commodities, and or currencies (ex: 20-30%) chosen to have an inverse correlation to stocks during times of market stress to reduce the pain from corrections as well as to buffer any sub optimal performance from your slow market timer.

Scott

[quote]
Talking about market timing I am also wary about these new models who short SPY some or all of the time with as much money than they have on the long side, effectively making a 100% long 100% short portfolio, which I have never heard of in this industry. It seems very risky to me.
[/quote]Why does a market neutral strategy have to be risky? Most of the risk is taken by the designer since if it underperforms the market in a bull market such as 2012-2014 or the late 1990s then investors will fire the manager. Perhaps you mean the risk–not of losing money–but of opportunity cost, that is, the lost money that investors could theoretically have made elsewhere?

Market neutral strategies are not used in the industry because very few funds (if any) have a dependable way to produce more than a couple of percent of alpha. That means that a 100% long/100% short strategy will only make a couple of percent a year on average before fees. So it doesn’t make sense for them. But for R2Gs, if you believe that you have a reliable way of beating the market by a significant amount [some models probably have that ability] then going market neutral may make sense under certain conditions.

BTW, do you mean 100% long and 100% short, or 50% long and 50% SH?

[quote]
A combination of the below may accomplish your goal of minimizing whipsaws and avoiding damage during bear markets and corrections…

  1. A slow market timer to reduce the probability of debilitating financial damage during bear markets.
  2. A small fixed allocation to bonds, commodities, and or currencies (ex: 20-30%) chosen to have an inverse correlation to stocks during times of market stress to reduce the pain from corrections as well as to buffer any sub optimal performance from your slow market timer.
    [/quote]Am I the only one who would like to see the ability to switch the ETF used for the hedge as well as having the ability to vary the size of the hedge depending on market signal(s)?

aurelaurel/chipper - The beauty of being market neutral is that you are invested at all times. MN is not a new concept as I believe the first “hedge fund” was essentially market neutral. You don’t have to precisely time when to jump in to catch the big bull run (as in 2009), when doing so could just as easily cost you your shirt. In my opinion, attempting to time the market is synonymous to gambling. Get it right and you are a hero. Get it wrong and you are a bum.

What is even better than one market neutral fund is several that have a low correlation, preferably by different model providers.

I firmly believe that the best strategy is to minimize the volatility of your equity curve (which is what MN does). The more volatile your equity curve, the more likely you will lose in the end.

Steve

I also want that Chipper. I want a dynamic hedge modality. Hedges composed of two or more ETFs could be a start.

By the way, every investor and trader in existance is “timing” the market as they position themselves whether they know it or not.

:slight_smile:

Hi Denny:

I’ve worked on timing systems for 15 years and you’ve summarized the problem well. Here are my conclusions.

  1. Any timing system (that I can make) will, over the long term, ALWAYS reduce my absolute gains. Maybe someone else can make a timer that improves absolute gains over the long term, but from experience I know I can’t. Of course, like eveyone else I can design curve fit timers that produce beautiful back tests, but experience has taught me that is no guaranttee for the future. Theoretically I have come to believe I’ll never discover a new timing system that the big guys have not already found first with their far more extensive computer resources being guided by math and physics PhDs. Once the big guys have found a timer that worked in the past, they will exploit it in the present and thus make it significantly less helpful/profitable in the future. And worse still, some of the big guys may actively seek out timers that provide beautiful back tests and then trade against those timers (ie, trade against small guys like me who think we just found the holigrail).

  2. Yet I still use timers – beacause I now have a different goal – to improve the reward/risk ratio. I think (hope) that I can make a timer that will reduce drawdowns a lot but only reduce absolute gains a little. But I expect this will result in me making less money over the long term. Last year my son made more than I did because he ignored my timer and stayed in the market. Over the next 20 years I expect his absolute gain will be much greater than mine. The only way I’ll come out ahead of him is if he quits at bottom of a 60% draw down (or if I give up on timers entirely and trade like he does).

  3. Furthermore, I expect any timer I make will fail badly at least once every ten years (or more often). After 2008, I developed a timer that uses SPEPS estimates combined with an MA crossover for the SPY. From comments in some of your posts it looks like one of your timers may be similar. Any way, this timer largely failed me in 2011 (ie, it did not reduce the draw down all that much). I still use it, but I don’t expect it to always work. I don’t make it my only line of defense. I make sure my trading systems will “work” without a timer, by “work” I mean the system is profitable and the draw downs are tolerable without a timer. Then I add a timer to the system with the expectation (hope) that some of the time it will reduce draw downs without giving up too much profit.

Best regards,
Brian

I find that not buying any new positions when the benchmark is below a certain moving average is a good compromise between being fully invested at all times and being all in or all out. I use a percent gain stop loss in my sell rules, so as the market moves down my losing positions are gradually paired down and those holding up allow me to stay in the game if the market suddenly moves up. This allows one to avoid 2008 like events while not getting whipsawed excessively during slight market corrections. It’s not perfect but it’s the best compromise I have come up with.

I use for Timing:

Buy: (close(0,#spepscy)>ema(75,0,#spepscy)) or (close(0,#bench)>ema(75,0,#bench))
Sell: (close(0,#spepscy)<ema(75,0,#spepscy)) or (close(0,#bench)<ema(75,0,#bench))

The reason for this is, that I believe that there is only one Major factor why stocks rise: rising earnings.
Every other factor (interest rates, Industry, great Business modell etc.) are influencer, the key role Play earnings.

Though I have to admit that (close(0,#bench)<ema(75,0,#bench) is an optimisation.

It serves me well, drawdowns of 20-25% are normal and from my Point of view can not be avoided without
hurting Overall Performance to much. The Goal is it not to take the Monster hits of 50-70% that wait for
high beta small caps.

Regards

Andreas

Andreas,

Agree the total ‘value’ of the business matters a lot, and earnings is one of the major drivers of value.

But, there are many reasons stocks rise:
a) PE or valuation multiple expansion is one of the major ones. Robert Schiller just had a paper on this.
b) Total fund flows into / out of the market is another.
c) Huge technological shifts / labor productivity is another.
d) Increases in share buy backs or dividend payouts is another.
e) Inflation rates is another.

Attached a simple paper on this as well, decomposing historical return drivers of US equities (one author’s attempts). See the chart on page 3 for how things have looked by decade. Frequently, in many decades, it’s not earnings growth that appears to be driving returns at all - but rather valuation multiple expansion and contractions.

I don’t think primarily earnings related timing rules will hold up all well over time and don’t use them in my personal trading heavily. Using it as one component of a multi-factor market attractiveness ranking system makes a lot more sense at the tactical level, but requires a lot of patience.

Best,
Tom


DriversofEquityMarketReturns.pdf (448 KB)

What if Andreas is right?

In case no one has noticed SPEPSCY is in a decline.

Here in December SPEPSNY is in a huge decline. You can make your own judgements on global growth, oil demand, high yield bond defaults, possible excesses in consumer sentiment etc.

In a few months, we may all be saying Denny’s market timing is pure genius. In any case, it is always hard to know if abandoning a system during a decline is a good idea or just a retail investor mistake. Only time will tell. BTW, I’m not making any bets either way on this. I’m just saying I don’t know.

Okay I am making a few bets: I’m a little bit hedged now but less so than in the past and I will probably be hedging even less in the future. However, I’m not cutting any of my hedges now, under the present conditions. Full disclosure. Most accurate to say no new bets.

Andreas, should you really use an AND Boolean in your sell rule?

I would agree that the expectation of future earnings per share is what drives stock prices. The things Tom highlights all play into that. I do agree that using EPS related factors in a ranking system does not seem to work well. For me, using improvements in free cash flow or operating income gives a better indicator of earnings direction than using EPS itself.

Jim,

Earnings estimates for S&P 500 stocks are falling but those for the S&P 1500 are still rising. I suspect that a few mega oil majors are pulling down the average for the 500. What does that mean for the economy? I don’t know. I would guess that the economy is still okay, but I am keeping an eye on it. Theory says that low oil prices are good for the economy. Theory also says that there is no difference between theory and practice, but practice says that there is. My earnings based market timing system that I developed after studying 75 years or so of data and that I use in a couple of my R2Gs is still in the market, but I am watching developments carefully to see if the economy overall is staying healthy.

I use US Government bonds as a hedge as that is currently the worlds safe haven asset and tends to spike with volatility. Between my Bond R2G and another private portfolio I am invested > 40% in US Treasuries.

My best guess for the cause of falling oil prices is that the Chinese economy is hurting much more than the Communist party is letting on. Oil prices don’t usually fall so fast due to more production, but they do tend to fall off a cliff under certain types of recessions. China is now a significant chunk of the world and can affect the world. A Chinese recession may not affect us here in the U.S. because the U.S. are consumers while the Chinese are producers. A Chinese recession may not affect us here in the U.S. because the US are consumers while the Chinese are producers. However, much of the growth in jobs since 2008 in the U.S. has been in the energy sector, and with prices falling some of those jobs may not be maintained.

All,

Over the past 75 years or so in the U.S., every time that the economy went into a recession it took the S&P 500 (large caps) down with it by at least 20% (which is the accepted definition of a bear market). Those recessionary bear markets were generally long and grinding. The S&P 500 did go down by 20% without a recession a few times (such as in 1987) but it recovered relatively fast. It stands to reason that any earnings based model should have gotten you out of the bulk of those bear markets, but different models may have started a little earlier and/or ended a little later.

MA timing would also have gotten you out of of the bulk of the recessionary bear markets but would have gotten whipsawed on many of the quick corrections, because by the time a down trend started to form it began to bounce back. It would take a lucky MA based system to get out of those quick corrections in time. My conclusion has been that earnings based timing is better than MA based timing.

Some may argue that there are other things that can bring the market down besides recessions and therefore MA based timing feels more secure for them. If it works for you then fine, earnings based systems have fewer false positives.

I did my testing on large caps. It is possible that earnings estimates based systems don’t work as well in the small cap space. Small caps tend to fall much more often than large caps especially when relative valuations are higher as they are now.

(Disclaimer: I don’t have access to the data right now and this is all from memory).