Momentum Trading with Pre-Market Trends

Author Image Ivan Struk

Ivan Struk

One way signs and compass next to stock market line chart.

Anyone that has ever stared at a stock chart on their monitor in anticipation of the opening bell will know that stocks don’t open at the same price that they close. Most charting software doesn’t incorporate after-hours and pre-market price action, and therefore most charts look like this:

Netflix is notoriously active in the Q2 reporting period pre-market session.

As soon as one starts diving further into the realm of extended-hours trading they start asking a lot of questions. Does a stock that is increasing in price during pre-market continue to increase during the day? Should I short Tesla if it has a bad earnings report? Will the S&P500 stay positive if it’s already increased in pre-market?

We can answer this by studying historical pricing data using Python. All the code from this post is available on Github.

Objectives

Let’s try to discover if pre-market prices have any predictive qualities. Is it possible to build a successful trading system using stock performance in extended-hours.

  1. How important is pre-market price action relative to intra-day performance?
  2. What is the probability that a select stock increasing in pre-market will also increase throughout the intraday session?
  3. What’s the probability that a stock trending in pre-market will continue to trend in the same direction during intraday trading?
  4. What effect does earnings season have, are pre-market trends following announcements stronger or weaker indicators of intraday prices?

Data Collection & Methodology

The goal is to examine whether pre-market movements are indicative of intraday price action, to what extent can price action inform our investment decisions. For this we will need to define certain conditions, determine which variables we will analyze, outline our methods, and ultimately build a program that will output some results.

Defining Pre-Market

Pre-market and after-hours trading are two sessions that occur before and after the main intraday session, respectively. Traditionally, we would look at these two sessions individually, because they are separated by an 8-hour “dead” period, where no trading occurs. Pre-market runs from 4:00 to 9:30, after-hours runs from 16:00 to 20:00, and between 20:00 and 4:00 there is no trading. Pre-market and after-hours are very similar, with the same market participants and liquidity profiles. They also both play host to earnings announcements during company reporting cycles, which means this is when most price-action will occur following an announcement.

Instead of examining what happens within these sessions on an individual level, we like to think of the pre-market holistically as the time between the day’s close, and the next day’s open; 16:00 – 9:30. Regardless as to whether the company reports in after-hours or pre-market, the intraday session always follows later. We will therefore be examining the pre-market holistically, and analyzing the reactionary price action in the following intraday session.

Data Collection

For data collection we used Ramaroussi’s Yahoo! Finance Python library, which allows us to fetch historical price data for individual stocks. We made the sample of stocks the S&P 500 and Nasdaq 100 constituents, which aren’t as thinly traded as Russel 3000 components.

def get_hist(symbol):
    target_tickers = []
    target_tickers.append(symbol)
    fetch = yf.download(tickers = 
                target_tickers,
                period = "5y",
                interval = "1d",
                group_by = "ticker",
                auto_adjust = False,
                prepost = True,
                treads = True,
                proxy = None)

We used the historical data based for the trailing 5 years of each security. It is possible to complete this analysis without including pre-market and after-hours price data, by simply using the difference between the market open and the previous day’s close, so any EOD data source would be acceptable.

A component of this analysis considers the price action differences associated with reporting periods. À la post announcement drift, stocks tend to exhibit higher volatility and returns in the time leading up to and shortly following an earnings report. It was important that we find a data source that would allow us to quickly identify reporting periods within our five year sample period. For this we used the U.S. Securities and Exchange Commission’s EDGAR database. While the SEC offers a free API that can be used to fetch results, you can also scrape filings using Beautiful Soup 4, which is what we have done.

Juxtaposing the reporting dates over the 5 year performance allows us to analyze the price-action resulting from shifts in information crucial to understanding the price-discovery process. For the purpose of this study we ruled out companies that have IPOed in 2019, because of the small sample size.

Results

For the first test, we examine how much does pre-market activity contribute to total (close-to-close) price action.

Since we want to focus on the SP500 and Nasdaq 100 constituents we will initiate the first test by passing a list of symbols through the function.

stocks = qqq() + spy()

test_sample_contribution(stocks)

We find that on average pre-market trading contributes 36.68% to absolute daily price action in the sampled stocks.

The histogram below offers a visual representation of the attributable pre-market price action of the individual stocks. This clearly shows that a significant part of price movements occur outside of regular trading hours.

The table shows the probability distribution of the contribution that pre-market trading makes to absolute stock returns. The average is 36% and there are few outliers.
We can see that the data shows little deviation from the mean, with few outliers. Suggesting that pre-market trading contributes to a significant amount of price action across all SP500 and Nasdaq 100 component companies.
This table shows the stocks that depend more on pre-market action the most.  In descending order (symbols); ASML, FTI, JPM, USB, PRU, LNC, C, MET, MS, BK.
Using this table we can see that the stock ASML (ASML Holding) actually has more price movements occurring in pre-market than during intraday.
This table shows the stocks that depend more on pre-market action the most.  In descending order (symbols); AMCR, NKTR, EXR, VTR, NRG, HCP, LW, PSA, FE, DLR.
This table illustrates that even on the lowest end of the spectrum, pre-market trading still contributes to total price action.

Moving further, we now aim to determine whether a stock that is increasing in value during pre-market will continue to increase through intraday. After all, this is the determining factor in whether you should buy a stock that has popped overnight.

We use the function below and pass the SP500 and Nasdaq stocks as arguments in a list.

test_sample_long(stocks)

We find that on an aggregate level, a stock increasing in value during pre-market has a 50.77% probability of maintaining it’s price action during through the regular session. This is interesting because it suggests that there is close to no correlation, and the odds that stock will continue to increase are as good as flipping a coin. Some would argue that this is a great testament to the efficiency of a market in that this showcases an anti-persistent property.

The distribution of probabilities for positive pre-market price action as an indicator for positive intraday performance shows that most of the time a stock increasing in pre-market has a 50/50 chance of continuing to increase through intraday.
In assesing the seminal research question of “Should you buy a stock that is going up in pre-market”, we can see that the data suggests close to even odds.

With that said, there are still stocks that lend themselves to more predictability, as presented below.

The table shows the ten most predictable stocks by pre-market increases (a stock increasing in pre-market that continues increasing in intraday), in descending order; RMD, DLR, ED, ACN, PKI, VLO, CLX, O, VTR, IDXX.
Even the stocks that show a higher probability of increasing in intraday following a pre-market increase don’t boast stellar predictability. Depending on the sampling period that you use in your own tests you may not encounter probabilities higher than 60%.
The table shows the ten least predictable stocks by pre-market increases (a stock increasing in pre-market that continues increasing in intraday), in ascending order; BIDU, CMCSA, LBTYK, JD, FOX, COG, CTVA, AMCR.
Interestingly, on the bottom end of the test results we find two Nasdaq listed Chinese companies; JD and Baidu.

The previous test took into account only positive price action, and now we must test whether pre-market price action is at all indicative of intraday performance. This means we want to consider events where the pre-market action and intraday performance were both either positive or both negative. Similar to how investors use fair-value trading strategies with pre-market futures.

We apply the next function, passing the same sample of stocks as an argument.

test_sample_indication(stocks)

We find that a pre-market activity is indicative of intraday performance 49.65% of the time. Which means that it is actually statistically more likely for the market to do the opposite of what it’s done in pre-market.

This chart shows the probability distribution for pre-market price action as a forecast for intraday performance. In other words, if a stock is going up or down in pre-market, does it keep going in the same direction during intraday? We see the results tend the 50th percentile, showing on an aggregate level there is no way to tell based off pre-market action.
Testing all-time predictability based off pre-market yields results that are almost academic in nature. We see that on an aggregate level pre-market movements are not indicative of intraday performance.
This shows the top 10 most indicative stocks based on pre-market action. If a stock increases or decreases during pre-market it is likely to continue intraday performance in the same direction. Here is the list in ascending order: RCL, BA, CF, BWA, PVH, ALB, IBM, IP, FLS, AMCR.
In this table we see Amcor (AMCR) as the most indicative stock, however since it appears in the previous test as not positively indicative this points to a strong downside correlation. If AMCR decreases in value in pre-market, it is very likely to continue falling through the intraday.
These stocks are more likely to do the opposite during intraday, rather than follow their own pre-market trend (descending order - least indicative); CTVA, INFO, CHKP, LULU, RE, CMCSA, TMUS, CBOE, COG.
Those stocks that show an indicative probability of less than 50% are more likely to reverse on their pre-market price action throughout the intraday.

This doesn’t leave us with any actionable results. Examining individual markets shows little deviation from the mean, and there are few stocks that show any indicative behavior in pre-market.

Surely, there must be time when stocks move in a more predictable way with respect to pre-market. In the final test, we aim to run through the same statistics, but we will change the sample to include only the reactionary period that follows the quarterly reporting period.

test_sample_earnings(stocks)

Using the function above we will now scrape the EDGAR database for reporting dates, and pull historical returns from the following trading day.

S&P500 & Nasdaq 100 stocks increase in both pre-market and intraday, following an earnings announcement, 27.10% of the time.

Out of the four possibilities of price action between the pre-market and intraday session, 27% is in line with our expectations, given the previous results we have seen. We expect to see more correlation during reporting periods because of event’s significance warranting more investor attention. Therefore, it can be inferred that under the pressure of higher trading volume there would be more momentum in price-action, causing prices to move in the same direction.

In this image we see a graph of the distribution of the frequence of increases in pre-market AND intraday during reporting periods. In other words; how many times does a stock increase in value in pre-market and in intraday.
The histogram above illustrates how most datapoints lie within the 24% 28% range.

There is a noticeably higher dispersion in the data following this test. It is important to remember that for certain stocks the number of reporting periods is less because the companies are newer to the exchange, and this plays a large role in constraining the sample.

Below we can see the top and bottom ten stocks by frequency of increases in both pre-market and intraday. A high frequency not only demonstrates that the company has performed better in the sessions following previous reports, but there is an inherent price-action consistency, and we can expect more causal attribution in further tests.

This table shows the frequency of the top 10 stocks that have had simultaneous increases through the pre-market and intraday trading sessions, following a reporting period.
Disney, and Dow Inc., both showing a frequency in positive pre-market and intraday price action.
This table shows the stocks that have had the smallest frequency of concurrent positive price action in premarket and intraday.
On the other hand we are presented a list of stocks that rarely experience simultaneous positive returns across both sessions.

The probability of a stock increasing throughout the intraday, after having increased in pre-market is 52.4%.

Similarly to the 5 year aggregate test, we see a close to fair result in whether pre-market price action is indicative of intraday action. However, like with the previous test, there is a higher dispersion, and we are able to identify companies that lend themselves to significantly more predictability.

The image shows a histogram of the probability that a stock continue to increase in intraday after increasing in premarket shortly following it's earnings announcement.
The histogram shows that the odds are slightly in favour of a positive intraday session following a positive pre-market session. Additionally, we can see that several stocks clearly show exceptional predictability, albeit possibly influence by sample size.

In the case below we are presented with 7 companies that have shown to have a perfectly indicative pre-market price action. In other words, if Broadcom Inc. (AVGO) or Disney Inc. (DIS) reports earnings and increase in pre-market, they are going to increase intraday as well.

In this table we show stocks that have indicative pre-market behavior, in descending order; APTV, TFX, WDAY, TDG, AVGO, DIS, DOW, MAS, HII, PAYX.
Disney inc. (DIS) is a great example of a stock that has a considerable reporting history, a high level of liquidity, and could be considered for actionable tests.

Conversely, anytime Corteva Inc. posts earnings and increases in value in the pre-market, you would be better off shorting it.

In this table we show stocks that have low indicative pre-market behavior, in ascending order; AMCR, CTVA, NXPI, WU, FLT, CNP, COP, AIZ, HAL, GPC.
Stocks low down on the spectrum actually showcase high negative correlations to pre-market price action.

Final Remarks

In testing whether pre-market action is indicative of intraday performance we can conclude several things.

  • Pre-market price action plays a significant role in all sampled stocks and should not be ignored.
  • On an aggregate level pre-market price action does not dictate intraday price action.
  • Select stocks have shown historically to have a more indicative pre-market price action.
  • During reporting periods stocks move with more volatility, however pre-market price action is only slightly more indicative than on a five-year historical level.
  • In the case of select stocks, reporting periods can bring out indicative pre-market behavior.
  • Specific stocks should be tested for larger time periods in order to assess the true predictability of pre-market price action, however some companies are too young to have enough reporting periods to analyze.

Keep in mind that these tests are subject to misinterpretation and price action may still be driven by unique fundamentals within the stocks that exhibited highly-indicative pre-market behavior. Additionally, past returns can be useful in approximating future price action, but do not guarantee future performance. Finally, remember that the market changes constantly, and strategies that work today may not work tomorrow, however using the takeaways from these methods of pre-market analysis can help drive better decision making.

Disclaimer: All investments involve risk, and the past performance of a security, industry, sector, market, financial product, trading strategy, or individual’s trading does not guarantee future results or returns. Investors are fully responsible for any investment decisions they make. Such decisions should be based solely on an evaluation of their financial circumstances, investment objectives, risk tolerance, and liquidity needs. This post does not constitute investment advice.



Related Posts