S&P Intraday Price Study

[jswanson 10]

In this post I would like to explore the intra day action of the market to determine if we can find an edge. I'm going to be using the S&P E-mini futures market, but the principals here could be applied to any market. In particular, I'm interested how different times of the day affect different trading strategies. The trading day, as defined by the U.S. open and close, for the S&P is seven and a half hours. That translates to 390 minutes. When can thus break the day up into three 130-minute periods. Lets also explore the times before and after the regular day session. To do this we can add a 130-minute period before the market open (Pre-Market) and a 130-minute period after the market closes (Post-Market). We have broken a trading day into the following periods:

 

Six Market Sessions

 

1. Pre-Market, 06:20 - 08:30
   
2. The Open, 08:30 to 10:40
   
3. Midday, 10:40 to 13:50
   
4. The Close, 13:50 to 15:00
   
5. Post-Market, 15:00 to 17:10
   
6. Trend Following and Mean Reversion
   

 

Now that we have these various times defined, I want to create a simple trading system to see if I can find an edge. The system should be very simple. I'm not looking to create a trading system but, I am attempting to locate clues as to what might work during a particular period. To keep this simple, I'm going to explore trades on the long side only and will be looking at the trading days between 1/1/2006 and 12/31/2011.

 

In general terms, we can attempt to trade the market with either a trend following type strategy, or a mean reverting strategy. Trend following implies the market is moving in a direction and we expect it to continue. On the other hand, a mean reverting strategy means we expect the market to reverse soon, thus we position trade opposite of the current market direction. To determine which type of trading strategy would work better during these time periods I'm going to create two simple entry methods.

 

For the trend following entry rule I will use the momentum calculation and open a long position if the momentum is above zero. For the mean reversion entry rule I will use a 9-period RSI and enter a long position when the value is below 25. The system will not have any stops. A single trade is entered and the position is closed at the end of the period. We are allowing only one trade per day. Slippage and commissions are not deducted.

 

 

Trending System Entry = 12-period Momentum is greater than 0

Mean Reversion System Entry = 12-period RSI < 25

 

Market Mode Filter

 

I often will make use of a market mode filter when I'm developing a trading system. This filter breaks the market into two distinct modes: bull or bear. We will use a 200-period simple moving average on a daily chart to determine if we are in a bull or bear market. A bear market is when price is below the 200-period moving average. A bull market is when price is above the moving average. Using this filter will limit our long trades to only be taken when price is trading above the 200-period moving average on a daily chart. In other words, we will be taking trades in the direction of the overall market trend.

 

 

Bull Market = price > 200-daily simple moving average

 

Testing Trend Following Characteristic

 

BULL MARKET: I created an EasyLanguage strategy to first test the trend following characteristics of the five different periods. Below is a bar graph representing the net profit for each period. Period 1 starts on the left-hand side through period 5 on the far right-hand side. What this shows is during a bull market there is an edge with momentum trading during the Midday period and The Close period. This seems contradictory to what many may think. According to these results, momentum trades don't seem to do well during the morning. Instead they do better during the lunch time and at the end of the day.

 

 

BEAR MARKET: Lets switch things up a bit. Lets now only take long signals during a bear market. In other words, we will now only take our long trades when price is trading below the 200-day SMA. This is opposite of the test we just performed above. It is also counter to the prevailing major market trend. Below is a bar graph representing the net profit for each period. This makes a huge difference. During a bear market there seems to be and edge for momentum trades during the last minutes of the trading day. How many times have you seen price rally into the close during a bear market? I've seen it many times and this study shows it.

 

 

Both bull and bear market exhibit late day momentum bias to the up-side. This phenomena is even more pronounced during a bear market. Trying to build a trading system to exploit this edge may be worth investigating.

 

Testing Mean Reversion Characteristic

 

BULL MARKET: Now I will test using our mean-reversion trading system rules over the five different periods. Below is a bar graph representing the net profit for each period. The clear winner here is The Close period.

 

 

BEAR MARKET: Now I will test using our mean-reversion trading system rules over the five different periods during a bear market. Below is a bar graph representing the net profit for each period. The clear winner here is the Pre-Market Period.

 

 

Here we see our best edge can be found in the pre-market hours during a bear market. Below is a graph that combines all our results into a single image.

 

Conclusions

 

The graph below may look a little confusing at first. Focus one the bars above the zero line. That's what we are interested in since they show the most net profit. Ask yourself, which bars are the largest?

 

If looking to create an intra-day trend following system there is one clear choice, locating an opportunity to go long during The Close period while within a bear market. It's seen as that large green bar during period four which almost hits $10,000. This conclusion seems counterintuitive but that's often the nature of the markets. For an intra-day mean-reversion strategy, a good place to start would be during the Pre-Market period while within a bear market. This can be seen with the large red bar during the first period. Also note how many bars are below the zero line. This suggests that shorting during the day may have a slight edge vs. going long.

 

Is it not strange the two largest edges for both mean reversion and trend following strategies are going long during a bear market? How odd is that! Yet, it's not so strange when you take into account other studies. In fact, this article's strange conclusion is also supported by a previous study that demonstrates much of the points gained in the S&P market over the past 10 years have occur during the overnight session. In other words, the up-side edge during the day session is not overly strong. It seems we are seeing this over-night affect within this study as well.

 

This is a rather simple study, but such studies can often point you in the right direction when gathering ideas on trading systems. Of course it is possible to discover a intra-day trend trading system that will do well. But this study is a great starting point on where to look for potential edges. In short we are asking, where might be the low hanging fruit?

 

This type of study can also be applied to any other instrument. Likewise, we only explored longs during this study. Adding a shorting study would also help build a more complete picture. Other ideas could include, using different indicators for our trend following and mean reversion trading systems. Likewise, it would also be important to vary the look-back period for our indicators. We used a 12-period-look back for both indicators. By testing a range of values, say between 6-14, we can demonstrate the robustness of the edge.

[BlueHorseshoe 10]

Focus one the bars above the zero line. That's what we are interested in since they show the most net profit. Ask yourself, which bars are the largest?

 

An interesting study. However, I'd be much more intrested in other metrics (profit factor, maximum drawdown, symmetry of long and short performance) before focussing on Net Profit as you suggest.

[jswanson 10]

An interesting study. However, I'd be much more intrested in other metrics (profit factor, maximum drawdown, symmetry of long and short performance) before focussing on Net Profit as you suggest.

 

Those would be interesting as well. I'm simply providing a starting point...

[steve46 12]

So item 1

 

Do you believe this sample size is significant......I can see the time period, but what is the actual number of data points? and what leads you to believe that you have a "a significant data sample".

 

Do you account for the element of stationarity (if so.....how)

 

Your classification of bull vs bear markets uses the 200 period MA (bull market above/bear market below)....what evidence leads you to believe that markets exhibit symmetrical behavior?

 

 

Thanks

Edited January 19, 2012 by steve46

[bobcollett 10]

Hi jswanson

Your post is very interesting

I would like to test your ideas in my market (South Africa )

This market is open from 8.30 am to 5.30pm + a low volume overnight market.

Do I still divide by three given these longer trading hours?

Were your 130 minute periods simple convenience or is there some reasoning behind the choice.?

regards

bobc

[Moyyim 10]

I'm going to be using the S&P E-mini futures.... The trading day, as defined by the U.S. open and close, for the S&P is seven and a half hours..That translates to 390 minutes

 

Is it ?..Does it?

 

ES RTH is 9.30am to 4.15 pm est = 6.75 hrs (405 mins)

 

anyway, 7.5 hrs = 450 mins...(390mins =6.5hrs)

 

What am I missing or am I just tired ?

[jswanson 10]

So item 1

 

Do you believe this sample size is significant......I can see the time period, but what is the actual number of data points? and what leads you to believe that you have a "a significant data sample".

 

Do you account for the element of stationarity (if so.....how)

 

Your classification of bull vs bear markets uses the 200 period MA (bull market above/bear market below)....what evidence leads you to believe that markets exhibit symmetrical behavior?

 

 

Thanks

 

The sample size may be significant or it may not. A rolling test over a longer period may be interesting as well so see if we can find any ebb and flow on how the market evolves.

 

The 200 SMA has a lot of weight behind it. I've done numerous studies over the years that show the value of this MA. I don't think I posted any here however.

 

No account for stationarity.

 

Remember, this is not a trading system. It is a market study, one of many that may help to highlight important market characteristics.

[jswanson 10]

Hi jswanson

Your post is very interesting

I would like to test your ideas in my market (South Africa )

This market is open from 8.30 am to 5.30pm + a low volume overnight market.

Do I still divide by three given these longer trading hours?

Were your 130 minute periods simple convenience or is there some reasoning behind the choice.?

regards

bobc

 

The division of the market is largely arbitrary. In general, I tried to break the U.S. day session into the open session, lunch time session and end-of-day session. But there is nothing to stop you from testing larger or smaller intervals.

[jswanson 10]

Is it ?..Does it?

 

ES RTH is 9.30am to 4.15 pm est = 6.75 hrs (405 mins)

 

anyway, 7.5 hrs = 450 mins...(390mins =6.5hrs)

 

What am I missing or am I just tired ?

 

Error on my part. That should read 6.5 hours.

 

From 15:00 I include in the Post-Market period.