People have been asking about performance of sentiment data over the past few weeks. We track performance for all asset classes multiple ways. A lot of recent volatility has been overnight. The chart below illustrates performance of S-Factor S-Score data relative to pre-market close sentiment S-Scores. Criteria for this chart is as follows.
Minimum $5 filter.
S-Scores are snapped at (3:40 PM EST). 20 minutes before close.
We multiply S-Score with the subsequent Close-to-Open returns to calculate ‘Stock Score’.
For example. AAPL has an afternoon Pre-Close S-Score of 3.0 and has a Close-to- Next Open return of 10%. Its score would be a .30 = (3.0*.10))
Positive ‘Stock Scores’ are hits and negative ‘Stock Scores’ are misses.
We aggregate all ‘ Stock Scores’ each day to create a daily sum. The Sanity Graph is the plot of aggregating the daily sum of Stock Scores.
The below chart illustrates cumulative performance. As you can see recent large spikes in the chart illustrate the predictive power of S-Score for next days open. We are seeing a significantly higher number of positive Stock Scores over the last weeks. Sentiment data is proving to be an even more important predictor of stock moves during this period of volatility.
StockTwits data is below. Both charts are consistent.
During this period of market volatility Social Market Analytics S-Score data can be an effective aid.
At Social Market Analytics (SMA) we create predictive
signals by aggregating the intentions of professional investors as expressed on
Twitter. We have accumulated eight years
of out-of-sample data illustrating the predictive nature of the data. We publish sentiment metrics to illustrate the
tone of the current conversation relative to historical conversations. One of our key metrics is S-Score. S-Score is effectively a Z-Score, the measure
of deviation from the mean. An |S-Score|
> 2 means the current conversation is two standard deviations from the mean
over the predefined lookback period.
In the prior blog we
explored the benefits of SMA patented machine learning algorithms on return
characteristics. In this blog we
incorporate rolling back tests on the predictive signal to select portfolio securities.
SMA data is predictive across sectors
and industries but as with any factor there are securities that react more
predictively than others.
For chart below we use a rolling one-year accuracy metric for
predicting subsequent O-C return. The
faded lines are S-Score values only. Bolded
lines represent a theoretical portfolio with accuracy filters overlaid with
S-Score values. Only select |S-Score|
> 2 securities that have moved in the predicted direction 60% of the time over
the last year (bold lines). This is all
S-Score > 2 return values are very similar for accuracy filter
and S-Score only. S-Score < -2 had a large
impact from the accuracy filter. Securities
reacting negatively to negative Twitter conversation as measured by S-Score continued
to underperform relative to sentiment only.
This is another example of using sentiment combined with other metrics
leading to statistically significant predictive signals.
To see how sentiment can be used in your models ContactUs@SocialMarketAnalytics.com.
Social Market Analytics has extensive Intellectual Property in three distinct areas: Topic model creation, account filtering and natural language processing (NLP). I have written blog post about SMA topic model creation capabilities and the impact of our account filtering algorithms. This blog answers the question – “Do your machine learning algorithms really add value to the NLP process?”. Answer -> Yes. The chart below illustrates the statistically significant benefits of Social Market Analytics Machine Learning Algorithms in isolation.
Start date for this analysis is 11/20/2018 and the end date is 4/30/2019. This period was chosen because of the significant market draw down in December. We use dictionaries with three distinct rule sets. We use a static dictionary as of the start and end dates and compare resulting predictive returns with a point-in-time dictionary (production). Our patented NLP scores Tweets using the dictionaries at each time, S-Scores are calculated from the generated Tweet scores. The point-in-time dictionary represents word additions, phrases, and grammatical logic as they are made.
We isolate the impact of our NLP process by turning off account filtering applied to the Twitter stream. To ensure we are pulling Tweets only discussing companies and securities, we are using our topic model filtering algorithms. We regularly publish our full return charts to illustrate the impact of our entire process.
Let us start by defining the lines in our chart.
Red Line = Tweets are scored using our dictionary of words and phrases as of 11/20/2018. This illustrates the performance with no machine learning applied on a go forward basis. This is the base case. This line represents the least amount of learned information.
Black Line = Tweets are scored using words and phrases
applied Point-In-Time. This is the
production feed SMA customers receive. We
use Supervised and Unsupervised Machine Learning. There are impacts from both during this
Line = Represents the Perfect Information scenario. Take the most up to date
dictionary of words and phrases (4/30/2019) and apply them backwards. All information learned during the volatile
period is included. This represents the
values expected to be received on a go forward basis.
The charts below represent the cumulative Open to Close return of securities selected based on S-Score 20 minutes prior to market open. S-Score measures the tone of the current conversation relative to historical benchmarks. We select securities with an |S-Score| > 2. Securities with S-Score > 2 are purchased on the open. Securities with S-Score < -2 are sold short on the open. SMA Chart lines represent a theoretical long/short portfolio. Isolated long and short sides are available upon request.
For comparison purposes S&P 500 open to close chart for the analyzed period is below.
The chart below illustrates the cumulative O-C performance illustrating the impact of our ML algorithms. As expected, the lowest performance is the red line representing the dictionary at start date. The back line represents SMA production data and green line represents the perfect information case.
Sharpe and Sortino Ratios
for our test period are below. SP Return
= 3.3%, Sharpe = .58 and Sortino = .96.
Again, this only looks at the impact of SMA NLP and does not include account filtering. At SMA we believe it’s not just what is being said but who is saying it. We employ a twelve variable algorithm to score and filter all Twitter accounts Tweeting about companies/securities to identify our approved account universe. As you can see SMA NLP is a learning system with demonstrable impact. To learn more please contact us at contactUs@SocialMarketAnalytics.com.
Coin Metrics and Social Market Analytics (SMA) announced today a partnership to incorporate SMA’s Crypto Currency Data Feed into the Coin Metrics Market Data Platform.
Alternative data such as social media platforms and data feeds have become a vital source of information for traders, particularly in the Crypto Currency Markets. The SMA Crypto Currency Sentiment Feed will offer the Crypto Currency community a tool for including social media sentiment data in their trading and portfolio strategies and expand Coin Metrics market leading Crypto Asset market and network data products.
“As the Crypto Investing market continues to mature, institutional investors are demanding data from trusted partners. These institutions are looking to make data-driven decision by accessing sources of data that they understand from their legacy investing frameworks. We believe that the power of combining sentiment data with granular network and market data is fundamental to building a deeper understanding of crypto assets. Coin Metrics is excited to partner with SMA, who has a long history of providing sentiment data to traditional capital markets participants and share Coin Metrics’ principles and values. The ability to provide an all-in-one Crypto Financial Data solution is a huge convenience for institutions.” Comments Tim Rice Co-Founder and CEO of Coin Metrics.
“Artificial intelligence and Natural Language Processing are moving into our everyday lives at light speed, and perhaps into financial markets even faster than that. We feel strongly at SMA that participants in Crypto Currency markets will benefit from our unique process in this emerging field, both in its approach to filtering social media data and in the analytical methodology used to develop our proprietary metrics. We’re excited to partner with the Coin Metrics team to offer this service through a versatile industry leading platform” said Joe Gits, Co-Founder and CEO of SMA.
About Coin Metrics
Coin Metrics was founded in 2017 as an open-source project to provide the public with actionable and transparent network data. Today, Coin Metrics delivers market and network data, analytics and research to its community and wider industry. https://coinmetrics.io/
About Social Market Analytics, Inc.
Social Market Analytics quantifies social media data for traders, portfolio managers, hedge funds and risk managers using patent pending technology to detect abnormally positive or negative changes in investor sentiment. SMA produces a family of quantitative metrics, called S-Factors™, designed to capture the signature of financial market sentiment. SMA applies these metrics to data captured from social media sources to estimate sentiment for indices, sectors, and individual securities. A time series of these measurements is produced daily and on intraday time scales. For more information, including a User Guide to S-Factors™, please visit www.socialmarketanalytics.com
Social Market Analytics aggregates the intentions of professional investors as expressed on Twitter. We identify these professional investors using our proprietary twelve factor ranking system. One factor is the forward accuracy of Twitter accounts. If a Twitter account is Tweeting bullishly based on our patented NLP process and the security subsequently moves higher over specified periods that account is deemed to be accurate over that period. Overall accuracy is aggregated across time for each account. We have been tracking account accuracy out-of-sample for the past seven years. – it is impossible to recreate this data. SMA is the only provider with out-of-sample account accuracy. We found significant variability in account accuracy for supposed professional investors. Social Market Analytics account scoring algorithms are extremely effective in excluding non-professional professionals.
SMA’s Accurate Account algos aggregate expectations from the most accurate Twitter accounts for individual securities for a specified time period: 1-Day, 2-Day, 1-Week, and 1-Month holding periods. Definition of ‘Accurate’ – correctly identifying directional movement of the security’s price. We do not include size of move – their sentiment is positive and the security moved higher.
We calculate consensus expectations of these accurate accounts on individual securities. Accurate account universes differ across holding periods. Some accounts are more accurate in the short-term (Day trades), while others are more accurate for longer holding periods (up to one month).
Securities with significant consensus for both long and short are available through our API’s, Widgets and in Reports. Below is a widget identifying securities with the most positive and negative consensus. In this example, SMA’s accurate account universe is currently 100 bullish on MCO over the next 24 hrs. Positive, negative and neutral are identified separately.
To discuss getting access to these or any other SMA data feed or widget please contactus@socialMarketAnalytics.com
Social Market Analytics aggregates the intentions of professional investors as expressed on Twitter. We apply our patented filtering and natural language processing(NLP) to Tweets to proactively select Twitter accounts to use in our predictive metrics. We track several metrics to gauge the predictive nature of our dataset. For this blog I am going to illustrate one of these metrics.
2018 was a rough year for the SP500, it lost about 9% (rolling one year). Given market loss and the high volatility we thought it would be an ideal dataset over which to run an experiment. Two questions we get regularly are: How would your data perform in a bear market? And what is the benefit of your NLP and account ratings systems? This blog will answer both questions from the perspective of 2018 market performance.
The table below illustrates performance of six theoretical portfolios. These portfolios represent stocks with Social Market Analytics S-Scores of 2 or higher (Long signal) or Social Market Analytics S-Scores of -2 or lower (Short signal). S-Score compares the tone of current Twitter conversations with average tone of Twitter conversations over the last twenty days. Social Market Analytics has multiple baseline for multiple prediction periods.
Each security in our universe represents a proprietary Topic Model. Each Topic is a collection of rules used to include or exclude specific Tweets from security buckets. For example, if you are looking for Tweets about Ethan Allen furniture (ETH) you do not want to include Tweets about Ethereum Crypto Currency (Also symbol ETH) conversations.
We created portfolios with our account filtering algorithms and compared them with portfolios of all twitter accounts discussing our Equity Topic Models. The purpose of the run was to quantify the ability of our patented account filtering algorithms to identify professional, and hence more accurate, investors. Spoiler alert: Our account filtering improved the long/short return by 50% (18.73 for 2018 versus 12.53 NLP only)
NLP applied only:
The NLP only portfolios illustrate the power of our NLP process to accurately identify and fine grain score Tweets discussing securities and companies. Our patented process reads each Tweet multiple times to identify if and how strongly someone is voicing a view of expected future performance. The NLP only portfolios illustrate the predictive power of our NLP in isolation. When you apply the Account filtering you get a predictive boost.
Account Filtered + NLP applied:
Account Filtered plus NLP portfolios illustrate the benefit of applying our account filtering metrics. Early in the life of Social Market Analytics we learned its not just what is being said on Twitter but who is saying it. We developed proprietary metrics to identify investors more likely to be correct about the future direction of a security. When the conversation of these professional investors is significantly more positive than the average conversation over the last 20 days those securities significantly outperform. When the conversation of these professional investors is significantly more positive than the average conversation over the last 20 days those securities significantly underperform.
Portfolios are constructed of securities with an S-Score of 2 or higher (long) or -2 or lower (short). All portfolios are equally weighted. A negative value for a short portfolio denotes a positive return to that portfolio. Short portfolios are supposed to move lower. All securities are entered on the Open based on a 9:10 am Eastern time S-Scores and exited on the Close. There is no overnight exposure.
We use SP500 as our performance benchmark. SP return is calculated from open to close in the same manner as the selected securities. Using open to close performance the SP500 returned -16.89% for comparison. As you can see from the table the S-Score > 2 outperformed the market and negative S-Score securities significantly underperformed the market (generating positive alpha). The L/S portfolio with NLP only returned +12.54%, NLP plus account filtering improved that performance by 50% to +18.73%. We do not illustrate this as a single factor model but removing 10% a year for slippage and commissions still significantly outperforms.
Please contact us with any questions or to see how SMA’s NLP and filtering capabilities can be used in your investment process. ContactUs@SocialMarketAnalytics.com
Social Market Analytics, Inc. (SMA) has been the leading provider of predictive quantitative signal in the alternative data space for 7 years. Over the years, we have developed patented algorithms that use machine learning and natural language processing to provide content that generates alpha. Our NLP is unique because of our proprietary processes that tag sentiment weights based on the language used in finance per asset class. The processing of tweets for futures and commodities is different from what we use for equities.
There have been a lot of conversations around Natural Gas futures in social media lately. Through our partner CME Group , people have been monitoring social sentiment from SMA in real time on their active trader site.
Recently, we did as study on using SMA signals to create a Long Short trading strategy using Natural Gas futures. In the example here, we are using front month futures contract and trading daily using a combination of S-Score (a measure of unusual sentiment) and SV-Score (a measure of unusual twitter volume activity).
The strategy buys contracts when the sentiment (S-Score) is positive. We scale up the long position when the sentiment is positive, and the volume of tweets is also significantly high (SV-Score). Conversely, when the sentiment is negative, we sell contracts and go short. We sell more contracts when sentiment is negative with significantly high number of tweets. For this study, we use a maximum of 100 contracts when going long and 100 contracts when taking short positions.
The sentiment strategy performs significantly better than the Natural gas prices, returning 87.42% YTD. We also avoid the volatility in the price and get a Sharpe of 3.53 on the strategy.
A PnL curve of investment of $300,000 in 100,000 contracts on Jan 1, 2018 is shown in the chart below.
The strategy is profitable throughout the period, and the maximum drawdown is only 7.19%, which is significantly better than the 29.72% drawdown in the natural gas prices YTD.