High Probability ETF Trading

Quote from smilingsynic:

An annualized gain of 10.06% with a 16.87% drawdown is not particularly impressive.

Drawdown is arguably the most important thing to consider, since it helps one guess how much leverage to use (or not to use).

I agree with you on a few fronts.

10.06% is not particularly impressive. But within the context of a robust strategy, it fits nicely versus a buy and hold investor that suffered a 50% drawdown. You will note it just went through its most difficult period as this bear market rally gained steam the past few months. Given that the rally appears to have run into some headwinds, I might surmise the short side of this strategy may bear some fruit in the next few months...

I couldn't agree with you more on the drawdown argument that it is the most important thing to consider. No offense to trend followers, as I use some of that in my work as well, but how a guy like Richard Dennis can basically take $400 to $200 mil, then give a good chunk of it back, well...that's not for me. Drawdowns of 50%+ are just insane. I get that pit in my stomach when one of my strategies hits 10%...
 
As a fairly new backtester - thanks for your posts. Much appreciated.

Quote from bentedges:

In a previous post I stated I would show a simple strategy that can give one an edge based off of ideas from other's strategies. One of the great things about learning from others is that it helps you to think and branch out into other ideas that very well may be profitable (provided one is willing to put in the time, effort and energy.)

A few things to note that I feel are important:

1. The KISS principle. Simple IS better. If you have 47 parameters or some such nonsense, the strategy will NOT be robust and there is a strong possibility it will blow up at some point in the future.

2. Generally speaking, a strategy should work across multiple time frames and securities (with minor adjustments.) This goes back to #1 and being a robust system. For example, if one uses a moving average in their strategy of say, 150 days, but they then try to optimize it by going to 168 days and the results are a bit better, but then they test 135 days and the system deteriorates badly, then they need to look at the structure of the strategy and if it is sound.

3. Risk management, and in particular, position sizing, are very important in the big picture. A guy that throws his whole wad into one great strategy, or for that matter into each trade, is bound to blow up. The inevitable drawdown will kill them, which leads me to...

4. The toughest thing with any strategy is following the rules and not letting emotions or ego get involved. When I was working at a hedge fund some years back, there was a dedicated programmer/trader who found a great system to trade the Q's. (Aside - he used to work for Ned Davis Research, and has since returned to that organization. This is a research firm that funds pay tens of thousands to get their insight, but I digress...) Anyway, so this system performed well in the backtesting and out-of-sample testing, so they started trading it. Long story short, after a few months it went into the inevitable drawdown with like 4 losing trades in a row. One of the principals said, "We're not taking any more signals from that system." Yep, you guessed it - the strategy went on a fantastic run that would've (helped) make their year. Again, if one has did the work, it's time to check the emotions and ego at the door. All too often a trader wants to be 'right' when all they should worry about is trading with discipline. If the strategy is sound, the profits will follow.

Now, onto this very simple system that anyone can follow without having to be a programmer. It doesn't make many trades, but when it speaks, I listen. Again, it isn't the holy grail, but as part of the bigger picture in the context of a total, diversified long/short portfolio, it will add alpha. Oh, and yes, this did come from ideas from others...in particular, Zweig and Connors. This strategy was designed for ETF's, but is able to work with individual stocks...I personally use other strategies for stocks. And I would be surprised if something similar hasn't already been put in the public domain.

RULES for LONGS:

1. The closing price is ABOVE the 200 day moving average.

2. Enter the trade on the close when the price is 3% or more BELOW the 5 day moving average.

3. Exit the trade at the close when the closing price is ABOVE the 5 day moving average.

RULES for SHORTS:

1. The closing price is BELOW the 200 day moving average.

2. Enter the trade on the close when the price is 3% or more ABOVE the 5 day moving average.

3. Exit the trade at the close when the closing price is BELOW the 5 day moving average.

Very simple, right? And purely price structure with no fancy indicators necessary.

Here are the results for some commonly traded ETFs going back to their inception (at least 5 years of data.) :

QQQQ

# Trades = 109
Winning Trades = 83 (76.15%)
Losing Trades = 26
Long Avg Win = 2.44%
Long Avg Loss = (3.26%)
Long Combined Avg = 1.07%
Short Avg Win = 3.82%
Short Avg Loss = (6.09%)
Short Combined Avg = 1.47%
Largest Win = 12.89%
Largest Loss = (21.86%) (DIVERSIFY!)
Max Consecutive Winners = 12
Max Consecutive Losers = 3
Winning Avg Held = 3.14 Days
Losing Avg Held = 7.15 Days
Profit Factor = 2.17

SPY

# Trades = 29
Winning Trades = 23 (79.3%)
Losing Trades = 6
Long Avg Win = 2.7%
Long Avg Loss = 0 - no losers
Long Combined Avg = 2.7%
Short Avg Win = 3.39%
Short Avg Loss = (4.36%)
Short Combined Avg = 1.18%
Largest Win = 11.18%
Largest Loss = (11.05%)
Max Consecutive Winners = 5
Max Consecutive Losers = 1
Winning Avg Held = 3.09 Days
Losing Avg Held = 9.0 Days
Profit Factor = 2.71

EEM

# Trades = 35
Winning Trades = 25 (71.4%)
Losing Trades = 10
Long Avg Win = 2.32%
Long Avg Loss = (3.72%)
Long Combined Avg = 0.19%
Short Avg Win = 5.5%
Short Avg Loss = (3.91%)
Short Combined Avg = 3.41%
Largest Win = 20.34%
Largest Loss = (8.39%)
Max Consecutive Winners = 5
Max Consecutive Losers = 4
Winning Avg Held = 3.04 Days
Losing Avg Held = 6.7 Days
Profit Factor = 2.90

GLD

# Trades = 15
Winning Trades = 11 (73.3%)
Losing Trades = 4
Long Avg Win = 2.46%
Long Avg Loss = (1.87%)
Long Combined Avg = 1.59%
Short Avg Win = 4.36%
Short Avg Loss = (2.38%)
Short Combined Avg = 1.66%
Largest Win = 4.71%
Largest Loss = (3.61%)
Max Consecutive Winners = 3
Max Consecutive Losers = 1
Winning Avg Held = 3.64 Days
Losing Avg Held = 6.50 Days
Profit Factor = 3.79

SMH

# Trades = 85
Winning Trades = 63 (74.1%)
Losing Trades = 22
Long Avg Win = 2.47%
Long Avg Loss = (1.88%)
Long Combined Avg = 1.31%
Short Avg Win = 4.20%
Short Avg Loss = (4.04%)
Short Combined Avg = 2.08%
Largest Win = 10.07%
Largest Loss = (23.05%)
Max Consecutive Winners = 10
Max Consecutive Losers = 2
Winning Avg Held = 2.78 Days
Losing Avg Held = 5.45 Days
Profit Factor = 3.02

OIH

# Trades = 72
Winning Trades = 49 (68.1%)
Losing Trades = 23
Long Avg Win = 2.20%
Long Avg Loss = (2.72%)
Long Combined Avg = 0.81%
Short Avg Win = 4.45%
Short Avg Loss = (5.74%)
Short Combined Avg = 0.96%
Largest Win = 17.91%
Largest Loss = (15.34%)
Max Consecutive Winners = 6
Max Consecutive Losers = 2
Winning Avg Held = 3.35 Days
Losing Avg Held = 6.43 Days
Profit Factor = 1.69

I Included OIH because a trade just triggered on the close on the last day of the week (Thursday.) It is a long trade, and the closing price of OIH was $92.62. Given that the last 2 trades in OIH were losers, I am thinking this one stands a reasonable possibility of coming in the black...
 
Quote from bentedges:

Not statistically significant? I disagree.

I don't care if you disagree or not. You show that you have no idea of what you are talking about. Less than 200 short term trades in 20 years is NOT a statistically significant sample. You basically have no idea of what statistically significant means. In 20 years you need at least 1000 trades to have a level of significance to even start thinking trading the system.

I can show several of systems that test well across many markets, actually across most US futures markets but when you begin to trade them you get excessive drawdown because the test results were not significant. Try a 200 MA combined with some filters and time exits and you get superb returns. Then try to trade it. Before you know you get your drawdown.

Ron
 
Quote from ronblack:

I don't care if you disagree or not. You show that you have no idea of what you are talking about. Less than 200 short term trades in 20 years is NOT a statistically significant sample. You basically have no idea of what statistically significant means. In 20 years you need at least 1000 trades to have a level of significance to even start thinking trading the system.

I can show several of systems that test well across many markets, actually across most US futures markets but when you begin to trade them you get excessive drawdown because the test results were not significant. Try a 200 MA combined with some filters and time exits and you get superb returns. Then try to trade it. Before you know you get your drawdown.

Ron

And I couldn't care less what you think. I am just thankful there are guys like you to take the other side of my trades.

I notice nothing to say on the S&P basket I showed. Oh, my bad...if one isn't a momentum monkey, then they're doing something wrong. Good luck with that, pal.
 
I saw a few comments on statistical significance. Here's a
general rule to determine if a method has any statistical
edge:

profit factor * SquareRoot(number of trades) > 20

So if a strategy has a PF of 2 after commission and slippage,
and there were 100 trades...

2 * 10 = 20

There is merit for further research. Make sure to filter
out strategies that win 100% of the time, as this would
result in an infinite number.

I would love if every strategy had 1000 trades, but many
do not.

Here's a great example. A long-term strategy has 38 trades.
Price is not a factor...just some cash flow data. The
strategy wins 91% of the time with only one optimized
variable. Do we throw this out because there are only
38 trades in 15 years? Profit factor is > 20.

D
 
Quote from danzman:

Here's a great example. A long-term strategy has 38 trades.
Price is not a factor...just some cash flow data. The
strategy wins 91% of the time with only one optimized
variable. Do we throw this out because there are only
38 trades in 15 years? Profit factor is > 20.

D

The problem is not whether you throw out the startegy. The problem is whether you have what it takes (money and discipline) to follow it IF it continuous performing like that. Even one optimized variable can do the damage IMO. In the past I considered several simple strategies that showed excellent results on paper with just one variable. One of them worked well for only six months.

I think long-term performance based on a few trades is an artifact of testing, an illusion that is. Signals for long term positioning are not robust at all and conditions change constantly. If you apply prudent risk and money management, your returns will come close to market returns. This is reasonable assumption but it is also confirmed empirically.
 
you can backtest 1 million systems and I guarantee you will find a few dozen that look golden.

with todays computer power this is trivial.


you could even write a book after you found them.
 
Quote from intradaybill:

The problem is not whether you throw out the startegy. The problem is whether you have what it takes (money and discipline) to follow it IF it continuous performing like that. Even one optimized variable can do the damage IMO. In the past I considered several simple strategies that showed excellent results on paper with just one variable. One of them worked well for only six months.

I think long-term performance based on a few trades is an artifact of testing, an illusion that is. Signals for long term positioning are not robust at all and conditions change constantly. If you apply prudent risk and money management, your returns will come close to market returns. This is reasonable assumption but it is also confirmed empirically.

I suppose you never know, but it's worked for the past 9 years with no problems. Battle tested in two major bear markets.
 
Back
Top