A recent Deutsche Bank Quantcraft report ("Catching the Sentiment Waves", 22 Nov 2022), found strong predictability in the MarketPsych Analytics dataset for commodities, FX and fixed income quantitative strategies, outclassing the 12 other sentiment datasets assessed. The report indicates: “The predictability improves as we increase the smoothing window….” and “The results also suggest that combining MarketPsych News and Social Media sentiment enhances the predictability of commodity market returns.”

MarketPsych’s commodity sentiment datasets cover news and social media reports on 50+ commodities in real-time from media sources in 13 languages. Analytics provided include buzz, sentiment, and unique thematic scores such as SupplyVsDemand and AgDisease – all derived from textual mentions in the media and converted into time series for easy modelling.

“The degree of predictability is strong in Fixed Income, Commodities and FX ….”
~ "Catching the Sentiment Waves." Nov 22, 2022. Deutsche Bank, Quantcraft.

Supply and Demand Sentiment

Track Information Flow & Global Demand

When Covid-19 lockdowns were enacted, those lucky enough to continue working from home sought to build out their home offices. Coupled with the slump in supply from closed lumber yards, the lumber supply vs demand mismatch was well expressed by news & social media platforms and filtered into MarketPsych’s Lumber SupplyVsDemand score, which had already dropped significantly by the time the front month lumber contract spiked up to >$1400/kbft.

Quantitative Support

Embed Thematic Sentiment into Quantitative Strategies

Research from 2024 shows that across liquid futures contracts for commodities, media sentiment intensity contributes a unique 14.4% premium to commodity growth, which is pronounced for commodities with low media coverage (among other factors). This is one of over 100 academic papers written on MarketPsych’s Analytics dataset.

From arbitrage to long-term forecasting and drawdown prevention, MarketPsych has papers to support our clients’. Across the board, when these analytics are included in an existing pipeline or paired with price data for feature generation, transparency of market states and accuracy of model forecasts increases. In one such model using GMMs to forecast reversion of the CRU-NGS spread, we find that model stability improves over the backtesting period when our sentiment scores are added to the model. Copy the python code from this model or one of several others to enhance your returns with media analytics.