Crypto Senti-Meter V2: A Derivatives Sentiment Index

Motivation

Derivatives markets aggregate and process a wealth of information in order to accurately price products. As a result, they encode information about market sentiment in their pricing data. For example, a volatility smile that is skewed towards puts indicates a high demand for OTM puts relative to OTM calls, and hence suggests a bearish sentiment among traders. Our index uses several key derivatives metrics that hold significant information about market sentiment and aggregates it into an easily interpretable measure of sentiment in cryptocurrency derivatives markets.

Building upon methodology we have developed previously, as documented in our previous report, we now introduce an updated second version of the Senti-Meter index. In this second version, we revise and explore the concept of market sentiment: how do we transform derivatives market positioning into an interpretable measure of changes in sentiment?

To begin with, we construct a measure of the current positioning in derivatives markets by evaluating metrics from three different instruments.

• Futures-implied yields. Positive values indicate that the future price trades above the spot price, indicating that they are willing to pay a premium for leveraged long exposure, while negative values indicate that the futures price trades cheaper than spot price.
• Perpetual Swaps funding rate. If positive, traders holding a long position need to pay a periodic rate to short positions for the privilege of the long exposure. If negative, then short positions pay long positions. Hence, a consistent positive funding rate expresses positive sentiment for the underlying, as traders are willing to pay a premium instead to unwind their positions.
• Implied volatility smile skew. The spread of implied volatility for OTM calls above the implied volatility of puts that are the same distance OTM. Positive values indicate higher relative demand for calls, hence the market points at a bullish sentiment, while negative values indicate higher relative demand for puts.

Each of these signals express the same relationship with sentiment – a positive value in each indicates growing positive sentiment, while a negative value indicates bearish sentiment. Therefore, we take the weighted average (with weights informed by historical data) of this information to produce a score: a single real number which can then be further manipulated. As an example we show what the score looks like for the past year in Figure 1.

When the score grows, it indicates that traders are taking on more bullish positioning. The faster that this change is, the more euphoric we judge the market to be. The contrary, which occurs when this score plummets rapidly, indicates that the market is in panic.

If the score is range bound instead, it can be interpreted that the market sentiment is stable. Hence we shift focus to what our end-product index is tracking: sudden movements of the original score value. More formally, Given an historical set of scores as reference, we define the Sentimeter v2 so as to track the relative position of today’s score with respect to the predetermined set of previous score values. This effectively tracks changes in outlook on the market, and hence changes in sentiment.

‍

Methodology

How is the index put together?

We therefore have removed extra layers of complexity we calculated in V1 to obtain a direct, yet intuitive version 2. More formally, on each day we perform the following:

• Denoise. Data, and especially crypto-related data, is extremely noisy and volatile. To better perform statistical analysis and data manipulation we perform denoising techniques to carefully extract relevant signals vital for detecting sentiment.
• Compute PCA Projection. We modified the implementation of the Senti-MeterV2 to make the best use of the data at hand. We now use the full history of derivatives market data up to the most recent day to perform a PCA that aggregates the signals we identified in the previous index version to create the score, a single number encapsulating the level of sentiment across each market.  
• Compare live value with recent history. By fixing a time frame, a time window, this score is then compared to a short history of the previous scores. More specifically, we quantify the distance between the current score and the average score in the window.
• Normalise. We compute a final measure of sentiment by normalising this distance by the standard deviation of the scores within the selected window.
• Inflate changes in sentiment. Positive scores indicate a bullish market, and negative values indicate bearish sentiment. We therefore highlight the moment in which scores cross the line and move above 0 after a recent history of negative sentiment or drop below 0 after a recent history of positive sentiment by scaling the index values up by a constant factor.

The resulting index oscillates between positive and negative values alongside spot price movements as well as changes in sentiment that are not immediately reflected in the spot market. We can visualise the index compared to the score in Figure 2. The index is constructed using a 30 days lookback window, and given the short horizon, we clearly notice that this year has seen strong fluctuations in sentiment levels, while there has been consistent positive score values. However, It is clear that the index, while tracking the evolving derivatives market sentiment, also establishes a correlation with spot movements.

Interpreting the index

The resulting values of the Senti-MeterV2 are directly interpretable: a large index value, for example above 2, indicates that the current score is more than 2 standard deviations away from the average score observed in the window of recent scores. This indicates increasing bullish sentiment or euphoria, hence jumps in the positive direction indicate a strong and growing positive sentiment in the derivatives market. Values close to 0 on the other hand, indicate that scores are close to the average of recent historical data points and that we do not see significant movement in market positioning. Negative values, e.g. values that are below -2,  indicate that the derivatives market is now below the historical data, flagging that derivatives market positioning is becoming more bearish, showing panic. We therefore highlight these thresholds as indicative of extreme changes in the derivative’s market sentiment.

In order to better capture variations in sentiment, we also transform the index to better reflect when markets have flipped the direction of their positioning. In fact, the sign of the score holds evidence of sentiment, as positive values reflect positive market expectations and growing prices and negative values reflect the opposite. Hence we inflate the index every time the scores have crossed the 0 line after a period of positive historical values or have dropped to negative values after a period of bullish market positioning. We achieve this by multiplying the sentiment index by a factor of 2 at every occurrence.

At the same time we want to scale down when the scores drop but remain positive in value, and when they grow but remain within negative values. Hence, if the current score value drops and remains positive after a bullish history, or if the score rises within a bearish environment, the index is rescaled by a factor of 0.5, in order to deflate the recent move.

‍

The size of the window

A key parameter in the Senti-MeterV2 index that dictates its behaviour is the window size – or how much historical data we define as “recent” enough to compare recent changes against. Large values will consider the relative change compared to a long history of data, hence will be able to capture longer term sentiment in an overall bigger picture. Short historical horizons, on the other hand, will better capture short-term movements.

In order to visualise this effect, we plot below the results of the index construction with a lookback window of 30, 50, and 150 days. As it is expected, a smaller window size reacts faster to sudden changes, while longer windows have smoother movements.

‍

Historical Analysis

We can proceed to analyse a historical snapshot of the Senti-MeterV2 compared to spot prices for the underlying asset. For this scenario we will focus on BTC specifically with a fixed window of 30 days. We look at the first major rally in prices which happened at the beginning of 2021 which was followed by a swift crash.

In periods of rising spot, as seen for example between October and March in 2020, the green shaded area in Figure 5), the derivatives market sentiment is generally positive. The same can be highlighted for periods of falling prices, where the index plummets towards negative values, highlighted in red.

‍

The index therefore shows two behaviours: either the index reacts quickly, and strong movements are recorded against the previous history, or it undergoes slow increases or declines that tend to follow spot dynamics. Swift movements are more interesting as they indicate a quick shift in positioning for the different signals under study, which could impact trading strategies more significantly than slow changes. By keeping track of the relative movements with respect to a short history we have therefore pushed the focus greatly on these occurrences, as seen in 2024, when the scores went from extremely positive to neutral/negative very quickly.

In conclusion, we have created an updated version of the index that tracks the relative change in position of the derivatives markets with respect to previous historical sentiment values. How far we look back to determine the current value of the index will determine how quickly the index adapts to new changes in positioning. Longer windows are more indicative of the overall long-term picture, whereas an index constructed on a short history will reflect quick changes that happened only considering recent days.