Papers, ranked by score

Ordered by a blend of empirical rigor (60%) and math complexity (40%).

Robustifying Markowitz

Härdle et al.

Markowitz mean-variance portfolios with sample mean and covariance as input parameters feature numerous issues in practice. They perform poorly out of sample due to estimation error, they experience extreme weights together with high sensitivity to change in input parameters. The heavy-tail…

Holy Grail Math 7.5 Rigor 7.0
Costs: ModeledOut-of-sample: Walk-forwardResolution: Daily Data
More details
Authors
Wolfgang Karl Härdle, Yegor Klochkov, Alla Petukhina, Nikita Zhivotovskiy
Evaluation
Trading P&L

Emoji Driven Crypto Assets Market Reactions

Zuo et al.

In the burgeoning realm of cryptocurrency, social media platforms like Twitter have become pivotal in influencing market trends and investor sentiments. In our study, we leverage GPT-4 and a fine-tuned transformer-based BERT model for a multimodal sentiment analysis, focusing on the impact of emoji…

Holy Grail Math 6.5 Rigor 6.0 capped Code
Costs: Not modeledOut-of-sample: HoldoutResolution: Daily Data
More details
Authors
Xiaorui Zuo, Yao-Tsung Chen, Wolfgang Karl Härdle
Evaluation
Trading P&L

Deep Learning and NLP in Cryptocurrency Forecasting: Integrating Financial, Blockchain, and Social Media Data

Gurgul et al.

We introduce novel approaches to cryptocurrency price forecasting, leveraging Machine Learning (ML) and Natural Language Processing (NLP) techniques, with a focus on Bitcoin and Ethereum. By analysing news and social media content, primarily from Twitter and Reddit, we assess the impact of public…

Street Traders Math 4.5 Rigor 6.0 capped
Costs: Not modeledOut-of-sample: UnclearResolution: Daily Data
More details
Authors
Vincent Gurgul, Stefan Lessmann, Wolfgang Karl Härdle
More data sources
Twitter / X
Evaluation
Trading P&L

Browse

All authors · Research topics · Papers with code · Download the scored dataset

📬 The Quant Space Weekly

One email a week: the most interesting quant finance papers, scored and summarized. No spam, unsubscribe anytime.