Hello and welcome to a brief *special edition* of Market Mood:

If you’re new here, consider checking out our recent posts to get your feet warm:

• Market News Roundup (4/4-4/10): Good, but not Great 🤷

• Market News Roundup (3/28-4/3): Mixed Views 🌦️

1. one year of analyzing market news

About a year ago I began writing my weekly reviews of stock market news coverage here on Substack, each week parsing through a few thousand news articles and analyzing the sentiment about each stock in conversation, with the hope of maybe conquering the markets for everyday investors like myself. There’s been some good and bad, some ups and downs, some sadness and euphoria (Billy Joel reference), and now a year later, it feels like a good time to sit back and reflect a bit on what this whole process of writing each week has taught me. I’d summarize everything I’ve learned in the following two bullets:

• the market answers to no one, no matter how much you think you know — luckily, everyone else also knows very little in the grand scheme of it.

• done correctly, analyzing market news sentiment can provide a slight edge, by weighing out the collective insights and opinions of everyone else and synthesizing the consensus mood about what the markets will do.

First, I definitely cannot say that we’ve conquered the markets; we’re still a looong way off from that one, and perfection is just an illusion at the end of the day. But I will say that we are a lot better now than we were when we started, and that’s the magic of it all. My first 6 to 9 months of writing weekly blog posts were very rough; I was basically spraying and praying, trying to figure out how to actually quantify the sentiment of stock market news articles (summarized in this blog post) and probably not providing very much value to the people who read them. Through consistent incremental tweaks, we have slowly improved our ability to leverage news coverage at scale to forecast stocks to some degree, and over the past few months, I’d say we are beginning to find our groove. Below, I’ll walk through how the stocks we’ve reviewed over the past few months have panned out, and how we’ll use this knowledge to continue improving.

2. how well our picks have performed

Rating overview:

If you’re unfamiliar with the format of our weekly Market Mood roundup, here’s a quick outline of how it works: we scrape a few thousand news articles from financial sources like Yahoo Finance, Benzinga, MarketBeat, etc. From here, we review the top 10 trending stocks (ie. those whose news coverage has increased the most over the course of the week), along with the top 10 most bullish and most bearish stocks (ie. those with the highest — or lowest for bearish stocks —average sentiment expressed across sentences written about that stock in news coverage).

We then select one stock from each of these three top 10 lists (1 trending, 1 bullish, and 1 bearish, usually the stock that ranks #1 in each of the respective lists) and review their news sentiment to assign each a simple “bullish”/”bearish” rating for the week ahead; generally, stocks with a bullish sentiment score receive a bullish rating (ie. buy it), stocks with a bearish sentiment score receive a bearish rating (ie. sell), and trending stocks receive either/or based on their sentiment.

Results:

Looking over the past 3 months, we’ve reviewed and assigned ratings to roughly 35 stocks (shown below vs. the performance of the $QQQ). Of these stocks, the news sentiment rating has correctly predicted the stock’s direction over the following week 58% of the time (21 out of 35) — which is slightly more than half. But the intriguing part here is the magnitude of those direction changes: when the news sentiment rating misses, it has tended to miss small, but when it hits, it’s tended to be more correct. On average, the ratings “misses” have lost approximately -3.32% over the following week, while the “correct” ratings have gained approximately +5.58% over the following week (ie. the ratings have performed roughly 1.7x better than the misses). A few of our best-performing news sentiment ratings over the past 3 months:

• Moderna Inc. ($MRNA) 1/9/22: bearish rating🔴, -18% following week🔴

• PayPal Inc. ($PYPL) 1/28/22: bearish rating🔴, -25% following week🔴

• Roblox Inc. ($RBLX) 3/13/22: bullish rating🟢, +17% following week🟢

On the whole, from the time of each individual rating through today, these picks (both buy or sell) have seen a total percent return of +58.71%. Now, a majority of this return can be attributed to the bearish picks (which would have a total return of +78.82%), while the return on only our bullish picks would have actually lost a decent amount of money (-20.12%) — in hindsight, this makes sense given the overall market corrections we’ve seen so far this year; over that same period the S&P 500 ($SPY) has returned -4.77%, the Invesco QQQ ($QQQ) has returned -8.88%, and the VIX Volatility Index ($VIX) has returned +12.47%.

Overall, if we were to have invested $100 into naked calls or puts for each of these 35 stocks ($3500 total capital deployed) based solely upon their news sentiment ratings on the day that we wrote about them (which we would NOT advise you to do, and would honestly recommend against ourselves at this point as we’re still tuning our rating system), holding those positions through today would have made us roughly $5,554.78.

3. what’s this all mean? what now?

There are certainly a few caveats to all of this showboating, if you can call it that. The markets so far in 2022 have been particularly weird, and basing our trust on a sample of only 35 news sentiment reviews is admittedly a bit nearsighted and maybe even reckless — the stock market is inherently random, and some will argue that a monkey could have guessed its way to better performance than we did (or in this case a goldfish named Frederick that literally made a profit by randomly swimming around his tank over the same period). And even if our algorithms and process for rating stocks based on their news sentiment were perhaps immaculate, the sample size and window on this test is just too small for us to know for sure at this point.

HOWEVER, if there’s anything to glean from this little exercise, it is hope. Through trial and error, lots of failures, iterations, and a bit of patience, we have what seems to be a half-decent way of beating the market, by synthesizing the mood about stocks in their corresponding news coverage. For us, this is a mere starting line — point A, if you will. From here, we will continue to refine our news sentiment algorithms, continue reviewing and rating stocks based on their news sentiment, and ultimately keep trying; as the old saying goes:

Going forward there are plenty of things that we are working on adding to this framework to improve it. Some ideas for future additions on top of our current structure and approach:

• adding social media and other forms of stock “conversation” into coverage.

• building out better “credibility” and/or “importance” weightings for individual news articles (after all, not all news sources are created equal)

• revising the language/criteria used to classify the sentiment of a given text as “bullish” or “bearish” about a given stock

These ideas seem promising, but I’m open to others as well. In fact, I find it extremely helpful to hear your feedback on these, so PLEASE feel free to reach out with any comments, questions, concerns, compliments, or cheap shots by emailing team@babbl,dev, replying to this email, or commenting on this post below. And with that I leave you — thank you all for reading this far if you did, and I wish you all the best, both on the markets and in your everyday lives. 🙏

Source cited: