Major collapse of social media sentiment with Bitcoin at 8000 USD

Bitcoin social media sentiment, calculated with machine learning from social media posts has recently collapsed to levels last seen when Bitcoin was around 6000 USD.

See our tweet with charts:

Positive sentiment has exhausted itself and we could see possible corrections on bitcoin and crypto in general.

Social Media Sentiment Signal calling intermediate bottom for Bitcoin at 3700 USD

BittsAnalytics provides its users with real-time social media sentiment, calculated with machine learning from social media texts about individual cryptocurrencies. Social media sentiment has proved as a very valuable signal many times in the past and again in recent downturn, effectively signalling intermediate bottom at 3700 USD.

There were several days of severe negative social media sentiment for Bitcoin during recent correction, as evident from this picture, showing daily social media sentiment for Bitcoin:

Similar picture was seen for social media sentiment on hourly level.

After the first surge on October 22nd the price remained first flat and then fell after retreat in sentiment. Yesterday however came a second even stronger sentiment surge when Bitcoin was at around 3700 USD and this sentiment surged to new intermediate heights. Unlike previous surge, the positive sentiment trend persisted this time and Bitcoin price also followed afterwards.

Social media sentiment was thus again proven very valuable as a market sentiment indicator. It is available as part of our Advanced subscription, more information is available at:

Social Media Sentiment also flashed warnings with Bitcoin at 6500 USD, you can read more in our previous blog:

Warning from our market sentiment indicator was again correct – flashing warnings with Bitcoin at 6500 USD

Warning from our market sentiment indicator was again correct – flashing warnings with Bitcoin at 6500 USD

In the beginning of 2018 we have introduced a market timing indicator based on social media sentiment of cryptocurrencies, where the sentiment is determined with machine learning from texts of social media posts.

For more detail and some of past warnings see e.g. our blog post from March of 2018:

Our warning Sentiment Market Indicator was now correct for the fourth time in a row

The market timing indicator was again correct recently, in flashing warnings before the latest fall in cryptocurrency prices. As you can see in the picture below:

it actually gave two warning signals in a short period of time. In a way the latest warnings were even more worrisome. The market indicator gives a warning when all among the top 10 cryptocurrencies in terms of market cap have a positive daily change in sentiment. In a sense it measures short-term excess optimism. But unlike most of previous cases this time was no major price jump before the warnings. In other words, despite a lot of excess optimism about bitcoin and sector in general (due to positive news on Bakkt and many others), there was almost no price change. If the price does not move on such optimism then it is likely that the upside is limited and downside can be large on potential negative impact news.

This is indeed what happened subsequently after the warning signals, a sharp drop in bitcoin and other crypto prices:

Our market sentiment indicator is part of our subscriptions, learn more about it at:

Sentiment again a very useful warning signal for Bitcoin price

In this post we are looking at recent social media sentiment data for Bitcoin. Sentiment is determined with machine learning from analysis of texts in social media posts about Bitcoin.

By analysing first the sentiment on the daily basis we notice that it had a very strong spike a few days ago:

For those who are reading our blog posts over the last year, this usually means a major price decrease in the near future. For previous such examples of Bitcoin falling after reaching high levels of sentiment see e.g. one of our previous blog posts on this topic:

Bitcoin again falls after reaching sentiment high – for the fourth time the warning was correct

We have recently introduced an even more detailed sentiment analytics, on hourly level. If you examine the picture of hourly sentiment:

you will notice that although the prices were relatively high the sentiment was really on the lower part of historical range for the last 10 days. Just before the price fell, it was again near the trough. Especially concerning was also the development days ago when the hourly sentiment reached record levels without any major movement in price itself.

Our sentiment analysis tool was thus again proven extremely valuable for market timing. If you want to know more about it read our other blog posts or subscribe at:

Introducing real-time social media analytics for cryptocurrencies

Next week we will be introducing a new feature, real-time version of our social media analytics tools (our subscriptions currently offer social media analytics data aggregated only on daily basis). Before going to a an example of analyzing real-time data on the case of today’s Bitcoin jump let us start with a quick look at example screenshot of the dashboard:

As the next step we will look at the bitcoin development today, because it was interesting both in terms of social media data and price. If we closely inspect the above chart we see a big hourly surge in tweets mentions of Bitcoin at 16:00 UTC time (this value is the sum of all Bitcoin tweets mentions between 15:00 and 16:00 UTC time):

The change with respect to the same value a day ago but in same hour slot is +110%. The value of this jump with respect to historical population has a Z score of around 3. The increase is thus historically around 3 standard deviations from the mean, making it a significant signal. Note also from the second chart that we got a sentiment reversal to the positive in this time period, which is a supporting bullish sign for the signal:

(Sentiment is determined with machine learning from the texts of tweets about Bitcoin). The robust surges today in price and social media space look like a good sign (weekend day was probably also a factor in terms of resonance).

If you are interested in our upcoming real-time analytics tools send us an email before 7th September to receive an early bird coupon which allows you to get the subscription at 30% lower price when we launch the service. Or you can subscribe to one of our powerful analytics tools already today at and you will get the 30% reduction automatically as an existing subscriber.

Technical patterns and Bittsbands of great help in trading recent rally of Bitcoin

Both technical patterns and our proprietary BittsBand indicated a resistance level at around 7700 USD and much clearer path above this price level. See our tweets:

As we said in the tweet there was a higher chance of seeing a rapid rise once the 7700 USD level is broken. The optimal way of trading this would be to set up e.g. limit buy orders at 7750 USD. And what we conjectured has today indeed happened. Once the 7750 USD was broken the price rapidly rose for a return of +7% in less than a day.

You can get analytics like this and much more data and analytical tools by subscribing at

Double top pattern recognition worked greatly for the second time in a row flashing warning for Bitcoin at 9800 USD

BittsAnalytics provides its users with a wide array of Big Data and advanced analytical tools to produce market timing and trading signals. Trading signals from BittsAnalytics can be based on:

  • sentiment analytics of social media posts (we use machine learning for sentiment classification)
  • social media mentions analytics
  • price resistance bands, BittsBands, obtained with deep learning
  • technical analysis of chart patterns (we have automated detection of chart patterns with sophisticated machine learning algorithms)

In this blog post we would like to show how even technical analysis works surprisingly well in identifying turning points. See the chart of double and triple top patterns for Bitcoin (the chart was automatically generated by our AI which does it every day for all cryptocurrencies):

As you can note there were two cases where the Bitcoin established a top pattern and then fell on that top line. First the pattern was established by points 1 and 2 at around 11900 USD. Bitcoin then fell after reaching the level for the third time (point 3). More recently the next top pattern was established with points 4 and 5. And after reaching this level of around 9900 USD (point 6) the Bitcoin again fell and has been down almost 30% since that price.

Top and bottom patterns can thus be a very useful tool for assessing the potential of cryptocurrencies for rise or fall in price. We automatically generate chart patterns like this every day with our AI algorithms and make them available at our platform. You can subscribe at our platform here:


Our Sentiment Market Indicator was flashing warning at 9800 USD – correct for the fifth time in a row

Market Sentiment Indicator is an indicator that we use to assess the sentiment of overall cryptocurrency market and is part of our framework of warning signals (* see detailed definition of indicator below). The latest warning signal from indicator came on 5th May with Bitcoin as a reference for the whole crypto market at around 9820 USD. The signal turned out to be again correct, crypto markets went in correction with Bitcoin reaching 7150 USD for a loss of almost 30%.  This is the fifth consecutive time our warning signal has been correct, i.e., it was followed by loss of at least ten percent. See the picture below and statistics for previous warnings (updated):

We believe that making profits in cryptocurrencies markets is becoming harder and the winners will be those with better data and better analytical tools. That is what we provide our users with our BittsAnalytics platform and if you want to start using our market sentiment indicator and many other data and tools you can find more information about becoming our user at

* Market Sentiment Indicator or MSI is based on sentiment of individual cryptocurrencies, with sentiment determined with machine learning from tweets mentioning the particular cryptocurrency. It is defined as the percentage of 10 largest cryptocurrencies whose sentiment was positive on a given day. For comparison chart is plotted with Bitcoin open price as a reference for general market.

You can read about our some of our past warnings from market sentiment indicator in our previous blog posts:

Our warning Sentiment Market Indicator was now correct for the fourth time in a row


Our deep learned BittsBands have proven excellent for trading in recent Bitcoin price developments

We are using deep learning on vast sets of different data to assess where could be possible price bands of resistance for cryptocurrencies. Let us explain how users of our platform can use our BittsBands for better anticipating possible price paths. We will use Bitcoin price movements in the last days as an example.

The more red the particular band is (see picture below) , the higher is the probability that:

1. the price will consolidate within the band or rebound from it when falling towards the band from above (see e.g. point 3 below)

2. or that it will be “repelled” by the band downwards when approaching the band from below or consolidate within the band afterwards (see point 5).

If the band is particularly weak, i.e. almost white, then there is a higher probability of price rapidly moving through such price band. This has indeed happened as you can see in points 2 and 4 .  Of course not every price movement will follow this pattern as often there are other drivers, e.g.. fundamental ones but even in those cases we often see effects as described above. We regard BittsBands as one the best analytical tools on our platform.

If you want to include these deep learned data in your cryptocurrency analysis you can find it at our BittsAnalytics platform: Platform also has a lot of other advanced data such as social media analytics and  chart patterns automatically detected with AI as well as sophisticated analytical tools.

End of Wild West in cryptocurrencies occurred in January – the tale of correlations, Part 1

In 2017 the cryptocurrencies had emerged as a new asset class with full force, never before has an entirely new asset class been established in such a short period of time. The year 2017 was also peculiar in that the individual cryptocurrencies were still very much different from each other in terms of relative returns or performance. There were some huge winners, also some losers and they were all correlated to Bitcoin to greater or lesser degree but the correlations were still lower than what is typical in other classes, for example stocks.

See for example rolling 20-day correlations with Bitcoin 5 months ago (on 6th November) for main cryptocurrencies:

A lot of them were actually inverse correlated with Bitcoin. In December of 2017  (data on 6th December) correlations increased but were still lower than typical in other classess:

Now contrast those correlations with values in April 2018:

So when was the regime change or phase transition to borrow from physics? Clues can be find in the following chart showing rolling 20-day correlation between Ethereum returns and Bitcoin returns:

As you can see there was a drastic change in correlations both in magnitude as well as in persistence in month of January. While the chart is shown for Ethereum other cryptocurrencies display almost the same evolution. While we would not like to speculate about the possible causes for this phenomenon one explanation could be emergence of new large financial players who employ arbitrage or some other classes of quant strategies.

So how does one adapt to these new times? Well, the bar has certainly been raised and simple approaches will work less well in achieving superior relative returns. But as we have shown in many blog posts in last months and that you can read by yourself, there are still possibilities for substantial performance but it requires better data and more sophisticated tools. One of the reasons we built our platform BittsAnalytics (learn more about it at was just to prepare for these kind of times. It was inevitable that they would eventually come. Perhaps it was just surprising how long it took for that happen.