The Case For Short-Term Trading

|
Includes: SPY
by: ETF ProTrader

Summary

Analysis of Annual, Monthly, Weekly, and Daily Trading Systems.

Monte-Carlo Simulation for gains using SPY.

Time Span and Compounding are Major Contributors to Gain.

Over the last several years I've developed and back tested a number of models to trade major indexes and commodities. Through that process I've developed a number of insights. First, I've studiously avoided comparing results where compounding is used; I believe that compounding is everyone's free lunch and it tends to seriously distort results, especially if one cherry picks a starting point or compares different time frames. Over time I've come to believe that only four things matter when looking at any trading system, methodology, strategy, set of rules, scheme, or whatever drives your trading decisions. The 4 things that matter are:

  • The time span between entry and exit points (holding period)
  • The percentage of winning versus losing trades
  • The average gain for winning trades
  • The average loss for losing trades

Now some purists would also want to know what the distribution of daily gains and losses is, but for now we'll leave that alone. What we will do to assuage that concern is to simply use a real-world distribution, namely the distribution of winning and losing trades for the S&P 500 Index (or the SPDR S&P 500 Trust ETF (NYSEARCA:SPY)). Traders also might add another item of concern; commissions, and for those who still believe that commissions are an important aspect of trading I have one word for you: Robinhood.

First let's look at some simple examples. There's a large number of articles on Seeking Alpha that describe various schemes for selecting and/or rebalancing portfolios on an annual basis. So let's suppose you had the world's best crystal ball; suppose you could look into your crystal ball every January 1st and predict, with 100% accuracy, whether the S&P 500 was going to be up or down at the end of the year. I'm sure most traders would pay a lot of money (and they do in the form of various newsletters) for such a crystal ball.

So how good would that be? For all of the examples in this article we'll assume we start with $100. You'll understand why later. So to establish a baseline, if we had $100 at the inception of SPY back on January 29, 1993 and we bought 2.276 shares of SPY at $43.94 we would now have $464 in our account at the end of 2015 with SPY at 203.87. And most investors would be happy; that's better than a 4-fold gain (not even counting dividends) on your investment albeit over a long 23 year time span.

Now let's see what having a 100% perfect annual crystal ball would do for you. Since you know, a year ahead of time, whether the market will be up or down for the year you simply either go long or short the S&P 500 (SPY) on January 1st and wait until the end of the year. Using your 100% accurate annual crystal ball your $100 initial investment would grow to $2,995. I think most people would be overjoyed with that kind of growth. You're beating the S&P 500 by 6.45 to 1 over the period and investors pay homage and lots of money for stock picking advice from a guru who advertises that he beats the S&P 500 by 3 to 1. Now suppose you wanted to trade your annual crystal ball in for a new and improved model (2.0); and the new and improved version 2.0 crystal ball worked on a monthly basis, and predicted, again with 100% accuracy, whether SPY would be up or down for the month at the beginning of the month. How much would $100 grow to now? Well, the novice would probably think something like 12 times the annual result, or 12 x $2995 = $35,940. But remember, we're compounding here, we invest the entire proceeds of each month's results into the next month (and why wouldn't we, after all our crystal ball is 100% accurate!). The correct answer is $814,493 not $35,940. Welcome to the power of compounding, what Einstein called the 8th wonder of the world.

So now you have your 100% accurate annual crystal ball and you want to trade it in for a monthly crystal ball. You go to the nearest flea market and see what's available; and what you find is that no one is selling a 100% accurate monthly crystal ball yet, but there are some 80, 85, and 90 percent accurate monthly crystal balls for sale at various prices. Let's say you wanted to get a new monthly crystal ball that would give you that 12x gain ($35,940) over your current annual 100% accurate crystal ball's return ($2995). What's the minimum % accurate monthly crystal ball that you should be looking for to trade in?

Now we're in the realm of statistics and Monte Carlo simulation. So let's look at the statistics for SPY on a monthly basis:

Total Number of Months = 275

Positive Gain Months = 167 (60.73%)

Negative Gain Months = 108 (39.27%)

Average Monthly Gain, Positive Month = +3.29%

Average Monthly Loss, Negative Month = -3.43%

We'll look at the 275 months and pick, at random, X% of the time from the distribution of positive gain months and pick, at random, (100-X%) of the time from the distribution of negative gain months. We'll do this 1000 times and find the average return.

The answer is 84%. If you have a monthly SPY crystal ball that's 84% accurate, on average you'll accumulate $35,536 over the 24 year period, which is about 12 times what you would get with a 100% accurate annual crystal ball which is already 6.45 times better than good old buy-and-hope (aka Buy-and-Hold).

No sooner do we start trading with our shiny new monthly 84% accurate crystal ball than we start to see advertisements for version 3.0 weekly crystal balls. So again we would like to trade in our monthly crystal ball for the latest weekly crystal ball but the question is the same; what percentage accuracy do we need in a weekly crystal ball to produce the same or higher total accumulation as our current 84% accurate monthly crystal ball?

Here's the statistics for 5-day windows in SPY:

Total Number of 5-day windows = 5801

Positive Gain 5-day periods = 3292 (56.75%)

Negative Gain 5-day periods = 2509 (43.25%)

Average 5-day Gain, Winning periods = +1.67%

Average 5-day Loss, Losing periods = -1.83%

If we run the simulation we find that now we only need a weekly crystal ball that's accurate 67.5% of the time to accumulate $40,456 over the 23 year period.

Well hopefully you can see where this is going. So let's look at the daily statistics for SPY and see what sort of accuracy we would need in a daily crystal ball to continue our quest to make even more money over time.

Here's the statistics for SPY daily:

Total Number Days = 5801

Positive Gain Days = 3130 (53.96%)

Negative Gain Days = 2675 (46.11%)

Average Gain, Winning Days = +0.78%

Average Loss, Losing Days = -0.84%

Running that simulation 1000 times we find that we only need a daily crystal ball that's 58.8% accurate to accumulate $56,325 over the 23 year period.

Well I certainly don't know how to build a SPY model that's 100% accurate on an annual basis, and I'm pretty sure I couldn't create a model that was 84% accurate on a monthly basis. I also seriously doubt that I could create a SPY model that's 67.5% correct on a weekly basis, but I do have SPY models that are 59% correct on a daily basis.

So of the four things that I believe are important in trading, I hope you see how important the time span between entry and exit (holding period) is. The moral of the story is "Make a little money in a very short time span, but do it consistently and repeatedly over time". You don't need to bend the odds that far in your favor (58.8% vs 54%) to dramatically affect your long-term compounded gain ($56,325 versus $464).

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.