Quant LetterNo. 58
162 items across 10 sections, as sent to readers on 24 July 2024. Paper titles open their ML-Quant page; ↗ goes to the source.
Quantitative-finance and ML-for-finance preprints from arXiv.
23 items
The study shows that using S&P500 index options in systematic index option-writing strategies can yield higher returns than traditional methods, with the Black-Scholes-Merton model proving more effective than the Variance-Gamma model.
9 sharesSource ↗
The research compares deep hedging and delta hedging in a GARCH-based market model, suggesting that the difference between the two can be a statistical arbitrage if the risk measure doesn't adequately consider negative outcomes.
8 shares7 citations todaySource ↗
The study proves that the global balance index of financial correlation networks can effectively measure systemic risk, as confirmed by its application to real financial data.
8 shares10 citations todaySource ↗
The research indicates that existing contextual bandit frameworks can be effectively used in the securities lending market, generating at least 15% more total revenue than standard methods.
5 shares2 citations todaySource ↗
The study introduces a method for training neural networks for Dynamic Initial Margin computation in counterparty credit risk, which reduces dataset generation costs and eliminates the need for repeated training.
4 shares1 citation todaySource ↗
The research identifies a negative drift in market making models, particularly in limit order fills, using the 10 Year US Treasury Bond futures for empirical simulation.
4 shares5 citations todaySource ↗
The study expands the optimal investment framework in a market with transaction costs and search frictions, introducing a new asymptotic framework for small costs and frictions.
4 shares1 citation todaySource ↗
The research shows that weak convergence in generalized gamma convolution distributions implies mean value convergence, proving the robustness of the optimal portfolio under exponential utility function with hyperbolic distributions.
3 shares1 citation todaySource ↗
A deep learning framework is proposed for calibrating the Heston option pricing model, showing superior performance in calibration accuracy and computational time compared to non-differential neural networks.
3 sharesSource ↗
The paper studies efficient risk sharing in markets with counter-monotonic constraints, characterizing optimal allocations for different types of agents and confirming risk-seeking agents tend to invest more in risky assets.
3 shares5 citations todaySource ↗
The article refutes the existence of rational bubbles in the model proposed by Miao and Wang (2018), providing clarity on the concept.
7 shares2 citations todaySource ↗
The research on collective action games with incomplete information reveals that similar information can both facilitate and obstruct coordination, applicable to authoritarian governments and committee decision-making.
3 shares1 citation todaySource ↗
The article uses data on programming languages in open-source projects to estimate a country's software economic complexity, demonstrating its relevance to international differences in GDP per capita, income inequality, and emissions.
3 shares11 citations todaySource ↗
The research uses a new model to examine the growth of economic activities, considering the balance between agglomeration and congestion, using data from Italian municipalities' income from 2008-2019, and performs better than typical spatial economic models.
2 shares1 citation todaySource ↗
The research uses French job data to pinpoint issues in job transitions, introducing two measures, transferability and accessibility, to enhance understanding of labor market movements and improve mobility, offering a structure for effective retraining programs and labor market efficiency.
2 sharesSource ↗
The research creates a dynamic model of production, export, and capital investment, discovering that exporting plants significantly upgrade their technology, boosting total productivity and the productivity of both skilled and unskilled workers, especially for new exporters.
2 sharesSource ↗
A new Explainable Deep Reinforcement Learning (XDRL) method for portfolio management has been developed, combining Proximal Policy Optimization with explainable techniques for better transparency in investment predictions.
5 shares16 citations todaySource ↗
An online experiment with AI models DALL-E 2 and DALL-E 3 showed that as AI improves, users adapt their prompts to utilize new capabilities, with DALL-E 3 users giving longer, more detailed prompts.
4 shares15 citations todaySource ↗
A new mortality modeling approach has been introduced, focusing on individual vitality dynamics and defining mortality as vitality depletion to zero, providing a new perspective for mortality modeling in various research fields.
4 sharesSource ↗
A study of cryptoeconomics and tokenomics from an economic viewpoint suggests that integrating these concepts can be innovative, aiming to connect the contexts of economics and blockchain.
2 shares10 citations todaySource ↗
Algorand vs Ethereum 2.0: The study assesses the effectiveness of Algorand and Ethereum 2.0 in tackling the Blockchain Trilemma by examining their decentralization, security, and scalability.
3 shares18 citations todaySource ↗
The paper investigates why blockchain startups are shifting to Decentralized Autonomous Organizations (DAOs), citing financial goals, governance rights, and market, legal and social norms as possible motivations.
2 sharesSource ↗
The article uses a model to study the effect of cap-and-trade pollution regulation on competitive firms. It finds that the level of competition greatly affects the economic outcomes of such regulation.
3 shares9 citations todaySource ↗
Working papers in finance and economics from SSRN.
27 items
Horikawa and Nakagawa's 2024 study suggests that deep hedging can have a speculative aspect, which can be mitigated by choosing the right risk measure.
5 sharesSource ↗
The paper reconciles the conflict between risk premiums and belief uncertainty during economic crises using a model that includes leverage dynamics.
4 shares1 citation todaySource ↗
The research investigates the return spillovers between ETFs and stocks, revealing an industry pattern and emphasizing the importance of careful supervision during periods of high market volatility.
5 sharesSource ↗
The paper introduces a new method for updating Intrusion Detection Systems in Internet of Vehicles applications using Convolutional Neural Networks to speed up the detection of recent attacks.
3 sharesSource ↗
The research suggests a dynamic financial hedging strategy to spread a firm’s risk, demonstrating that reduced hedging costs result in increased investor value and firm investment.
3 sharesSource ↗
Big tech companies like Google, Facebook, and Amazon use online user behavior tracking for targeted advertising, which could disadvantage smaller firms and affect GDPR regulations.
4 shares1 citation todaySource ↗
A new tool that combines data science and machine learning can improve water management in large irrigation projects, ensuring fair distribution.
3 sharesSource ↗
The Opening Range Breakout day trading strategy can yield high returns when used with leverage or leveraged products in US public equity markets.
3 sharesSource ↗
Machine learning models, specifically the Extreme Gradient Boosting model, can accurately predict the shear capacity of certain strengthened beams, aiding in better design practices.
3 sharesSource ↗
Digital transformation can greatly enhance corporate diversification, as demonstrated by a mathematical model and data from China's A-share listed companies.
3 sharesSource ↗
The article introduces a faster, more accurate method for pricing variance derivatives using one-dimensional PDE.
12 sharesSource ↗
The article uses XGBoost to predict next-day volatility jumps based on over 1400 news topics, improving portfolio performance.
2 sharesSource ↗
The paper presents a robust method for calibrating the SVI model to volatility surfaces in cryptocurrency markets using advanced optimization techniques.
2 sharesSource ↗
The chapter explores the impact of big data on positivist field research in performance measurement, emphasizing the importance of qualitative field data.
2 sharesSource ↗
The study recommends using the implicit scheme for pricing and hedging options due to its stability, despite slower convergence.
2 sharesSource ↗
The article emphasizes the importance of open data for AI, but also warns about ethical issues and potential harms related to data availability.
2 sharesSource ↗
The chapter explores how subjectivity influences performance measurement, focusing on practices that improve incentive and behavioral alignment.
2 sharesSource ↗
The research uses machine learning to predict mafia infiltration in Italian local governments, offering a tool for early detection and understanding the causes of such crime.
2 sharesSource ↗
The research investigates four techniques to enhance the precision of volatility forecasts for the SP 500, including the GARCH model, an LSTM network, a hybrid LSTM-GARCH model, and an advanced hybrid model incorporating the VIX index.
2 sharesSource ↗
The study reveals that employees use stock options for consumption rather than diversification, with changes in home prices and new restricted stock grants affecting their choices.
2 sharesSource ↗
The article contrasts two models for pricing derivative contracts on a commodity index, emphasizing the difficulties of the microscopic approach and doubting the accuracy of the macroscopic approach.
2 sharesSource ↗
The paper presents a stochastic representation for the joint distribution of the out-of-sample mean and variance of portfolio rules, providing a toolkit for assessing portfolio performance and creating improved rules.
2 sharesSource ↗
The research suggests a method to gauge an investor's risk aversion using actual asset allocation, calibrating a risk aversion parameter and exploring differences across financial categories.
2 sharesSource ↗
Large U.S. dealer banks make substantial profits from their trading desks, which serve as financial intermediaries, without taking on market risk in the asset markets they mediate.
2 sharesSource ↗
Despite their lower investment sophistication, Robinhood investors significantly contribute to liquidity during earnings and M&A announcements, trading more based on sentiments and showing higher demand elasticity.
3 sharesSource ↗
ESG investing in India is scrutinized, comparing the financial performance of ESG funds to market benchmarks and using traditional risk-adjusted metrics to debunk misconceptions about ESG's impact on profits.
2 sharesSource ↗
The dividend risk premium (DRP) is studied at both portfolio and firm levels, showing it to be procyclical and trendstationary, with investor sentiment and interest rates significantly affecting DRP levels and magnitudes.
2 sharesSource ↗
Economics working papers from RePEc's NEP field reports.
30 items
A study reveals that market conditions pre and during COVID-19 did not affect arbitrage profits significantly, but high volatility and low liquidity during the pandemic made arbitrage unfeasible.
23 sharesSource ↗
From 2004 to 2021, multi-asset funds investing globally underperformed, but performed better during market crises, with bond-focused funds doing well in non-crisis times and equity-focused funds during crises.
22 sharesSource ↗
Long-term straddle momentum, implied volatility, and illiquidity are identified as key predictors of cross-sectional foreign exchange options returns.
19 sharesSource ↗
A new model enhances the prediction accuracy of large realized covariance matrices of returns by breaking down the return covariance matrix using standard firm-level factors and sectoral restrictions.
14 sharesSource ↗
The Factor Graphical Lasso (FGL) framework, which combines graphical models with the factor structure, consistently estimates portfolio weights and risk exposure, and outperforms several key competitors in portfolio allocation.
11 sharesSource ↗
The article discusses a new method that merges decision trees and hidden Markov models to predict NFL play calls using time series data and the expectation-maximisation algorithm.
23 sharesSource ↗
The study introduces a dual-level optimization model for retail banking credit portfolio selection, aiming to reduce risk and increase returns, tested using data from four retail asset classes.
20 sharesSource ↗
The paper emphasizes the need to differentiate between marginal and conditional feature importance in machine learning, proposing a method that merges the conditional predictive impact framework with sequential knockoff sampling for complex, mixed data.
15 sharesSource ↗
The study uses machine learning to identify factors affecting AI readiness in businesses, including scientific research output, internet infrastructure, and public consumption expense.
13 sharesSource ↗
An improved model for predicting oil pump inspection cycles is proposed in the study, which could increase production efficiency and pumpjack lifespan.
9 sharesSource ↗
The study creates a predictive model for forecasting financial distress in Pakistani companies, finding financial ratio indices to be more effective than individual ratios.
9 sharesSource ↗
The study suggests that adding cryptocurrency factors to a stock-bond portfolio can benefit investors, especially when using machine-learning for asset allocation.
23 sharesSource ↗
The paper presents a machine learning model to classify social media posts by risk perception, aiding in understanding human risk approach and informing communication strategies.
16 sharesSource ↗
The study uses sentiment analysis and machine learning to predict stock indexes, indicating that investor sentiments and exchange rate are key for short-term predictions of the Shanghai Composite Index.
14 sharesSource ↗
The research introduces a sequential random search method for hyperparameter tuning in machine learning, which can identify and remove poor parameter configurations early, reducing the number of evaluations needed.
13 sharesSource ↗
The paper uses machine learning to predict the results of standard battles in the Chinese solid-state lighting industry, with the random subspace-MultiBoosting approach proving most effective.
12 sharesSource ↗
The study uses a hybrid approach with historical financial ratios to predict US bank failures, showing better performance than existing methods with a low Mean Squared Error and high R-squared value.
10 sharesSource ↗
The paper uses double machine learning to estimate the impact of vaccination on COVID-19 mortality in the EU, finding that a 10% increase in doses significantly reduces deaths and that Moderna and AstraZeneca vaccines are more cost-effective than Pfizer.
10 sharesSource ↗
The study uses machine learning and text analysis to study the effect of CEO optimism and narcissism on a company's capital structure, finding a slight, nonlinear impact of narcissism and no sex-dependence for median values of optimism and narcissism.
9 sharesSource ↗
FinBERT and LSTM for Stock Price Prediction: The article discusses a new hybrid model that combines BERT and LSTM for predicting stock prices. This model also includes financial news sentiment analysis and technical indicators. The model outperforms existing methods, especially in predicting significant stock price fluctuations.
12 sharesSource ↗
Implementing machine learning algorithms for business trend prediction requires significant changes to current systems.
52 sharesSource ↗
Machine learning models can accurately predict future earnings changes, surpassing traditional models and expert forecasts.
93 sharesSource ↗
Three- or five-factor models can minimize market anomalies in equity mutual fund performance evaluation, with conditional methods offering more explanation.
14 sharesSource ↗
Machine learning can identify financial misinformation on social media, using a dataset of financial news scrutinized by the Securities and Exchange Commission.
32 sharesSource ↗
Bid-rigging cartels can be accurately detected using statistical screening and machine learning, but performance varies across different countries due to institutional differences.
24 sharesSource ↗
The study shows that portfolio returns based on long-short anomaly can predict overall market returns, due to asymmetric limits of arbitrage and overpricing correction persistence.
116 sharesSource ↗
The paper proposes a deep learning strategy using 146 characteristic-based factors, which yields high economic value and is robust even after accounting for traditional factor models and transaction costs.
52 sharesSource ↗
The research finds that mixed-frequency factor models are better at forecasting certain Chinese economic indicators than traditional models, except during the Global Financial Crisis period.
10 sharesSource ↗
The study examines the topics and sentiments of Ukrainian Telegram users during the first two months of war, emphasizing the need for social media analytics to detect and counter social manipulation.
9 sharesSource ↗
The research uses properties of the currency market to create a risk-efficient portfolio of individual currencies, discovering that at least 85% of risk in these assets does not impact their risk premiums.
21 sharesSource ↗
The general machine-learning papers the letter carried in 2023-25.
17 items
SlowFast-LLaVA is a new video language model that excels in capturing spatial semantics and temporal context in videos, surpassing other methods in various video tasks.
56 shares139 citations todaySource ↗
A comparison of KAN and MLP models shows that MLP generally performs better, except in tasks involving symbolic formula representation.
50 shares146 citations todaySource ↗
Bridging the Gap: ChatQA 2 is a model that improves long-context understanding and retrieval-augmented generation, matching the accuracy of top proprietary models.
36 shares57 citations todaySource ↗
Efficient Visual State Model: The Modulated Group Mamba layer is introduced for state-space models, effectively addressing scaling issues in computer vision tasks and improving performance in image classification, object detection, and segmentation.
31 shares23 citations todaySource ↗
Hard-Negative Mining: The article suggests positive-aware mining methods for fine-tuning text embedding models, with the NV-Retriever-v1 model outperforming previous methods in the MTEB Retrieval benchmark.
29 shares108 citations todaySource ↗
Novel View Matching: The article introduces MaRINeR, a technique that enhances 3D rendering using information from a nearby image, useful for mixed-reality applications and autonomous agent training.
22 shares4 citations todaySource ↗
The paper proposes a new method of discretizing mel-filterbank channels for speech data, called dMel, which outperforms other speech tokenization methods and aids in efficient speech and text modeling.
18 shares14 citations todaySource ↗
The study presents an Explainable Deep Reinforcement Learning approach for portfolio management, combining Proximal Policy Optimization with explainable techniques to improve transparency in investment decision-making.
17 shares16 citations todaySource ↗
Single Video 4D Reconstruction: The research presents a method for reconstructing dynamic scenes from single-camera videos, using SE3 motion bases and data-driven priors, achieving top performance in long-range 3D/2D motion estimation and novel view synthesis.
15 shares288 citations todaySource ↗
Stock Similarities Representation Learning: The paper introduces SimStock, a self-supervised learning framework for identifying similar stocks, which outperforms existing methods and can be used in various investment strategies, highlighting the potential of data-driven approaches in investment and risk management.
13 shares3 citations todaySource ↗
The article introduces EdgeSAM, an optimized model for edge devices that is 37 times faster than the original model and performs better in terms of speed and accuracy on various benchmarks.
417 shares71 citations todaySource ↗
The study suggests Activation-aware Weight Quantization (AWQ), a hardware-friendly method for quantizing large language models that reduces error and improves performance on various benchmarks.
146 shares1,800 citations todaySource ↗
The piece presents Dynamic Memory Compression (DMC), a method for compressing key-value cache in large language models that increases throughput and maintains performance while accommodating larger contexts and batches within a given memory budget.
106 shares133 citations todaySource ↗
The research examines the effect of near duplicate subwords on language model training, revealing that while duplication hinders efficiency, merging similar duplicates can also harm performance.
71 shares6 citations todaySource ↗
RL for Drug Design: ACEGEN, a toolkit for drug design using reinforcement learning, is presented and validated, demonstrating equal or better performance than other generative models.
30 shares41 citations todaySource ↗
The book offers a graduate-level overview of advanced Markov chain Monte Carlo algorithms, focusing on scalable methods relevant to machine learning and AI.
30 sharesSource ↗
The study explores if language models infer text meaning from training data patterns, discovering they can decode sentence relations, but the prediction test works contrary to the theoretical test due to text redundancy.
29 shares16 citations todaySource ↗
Papers that shipped their code, from the Papers with Code feed (2023-25).
10 items
The article emphasizes the need for Language Models to better understand buying intentions in Ecommerce to assist in various tasks.
27,386 shares
The article suggests using Guided Latent Consistency Distillation and a multistep Ordinary Differential Equation solver to solve convergence problems in LDMs with fewer sample iterations.
664 shares
The article presents DataComp for Language Models, a platform for controlled dataset experiments to enhance language models.
522 shares
The article discusses the latest developments in QwenAudio, a large-scale audio-language model that can analyze various audio signals and respond to speech commands.
379 shares
The article discusses recent advancements in virtual try-on technology using localized garment inpainting and latent diffusion models, improving the online shopping experience.
237 shares
Current research is exploring the principles of scaling in the field of Embodied AI.
218 shares
The article presents DiTMoE, an improved sparse version of the diffusion Transformer that offers optimized inference and scalability.
76 shares
The paper introduces semantic operators, a new programming interface that combines AI operations with the relational model for more effective data queries.
53 shares
The authors propose a single modality training method for E5V, focusing solely on text pair training.
48 shares
The study highlights the importance of sequence alignment in matrix parameterizations, improving the performance of matrix mixers and explaining the success of Transformers and recent SSMs like Mamba.
46 shares
Repositories the letter featured.
10 items
The article explores a complex equity risk model developed specifically for quantitative trading.
104 shares
Vizro, a toolkit for building modular data visualization applications, is introduced in the article.
2,481 shares
The article introduces FeatHub, a combined stream and batch processing feature store for real-time machine learning.
305 shares
The article recommends top training resources for learning quantitative finance.
367 shares
The article reviews the top free and open-source libraries for Tailwind CSS components.
32,329 shares
The article presents the RAG Framework, an open-source tool for creating GenAI Second Brains, which uses Langchain GPT 3 to interact with different document formats.
34,236 shares
The article provides a tutorial on how to integrate TaLib functions into Polars expressions using a Polars extension.
43 shares
The article provides a guide on how to automate a homelab setup from scratch to running services using a single command.
7,936 shares
Industry news: funds, hiring, markets and regulation.
20 items
Clear Street is planning to acquire Fox River from Instinet, a firm that offers algorithmic trading solutions in US and Canadian equities.
10 shares
Iresss QuantHouse will supply market data to Quod Financial to improve its AI-based trading algorithms and backtesting.
8 shares
According to Bloomberg, family offices with over 1bn in assets are increasingly opting for quantitative trading strategies over traditional hedge funds.
6 shares
Alex Gerko, a significant UK taxpayer, lost a legal dispute against a government decision to tax his income from foreign currency trading at GSA Capital.
4 shares
ISwiss Bank has established a new company in New York, approved by the SEC, to provide hedge fund investment strategies and bespoke financial solutions for institutional and private clients.
4 shares
Universa Investments' Mark Spitznagel predicts a major market crash that could cut stock values in half.
4 shares
Hudson Bay Capital Management is limiting quarterly withdrawals, following other hedge funds in extending investor capital lock-up periods.
3 shares
Chinese quant hedge funds experienced significant losses in the first half of the year, underperforming compared to traditional stock and global fund strategies.
3 shares
Jain Global, a new hedge fund firm, has recruited a portfolio manager from Caxton Associates for its London branch.
3 shares
Gates Capital Management, a significant shareholder in Vista Outdoor, is against the proposed sale of the company's shooting-related products division to Czechoslovak Group.
3 shares
Colin Heatherington, a former trader at Absolute Capital Management, has been jailed for three and a half years for manipulating penny stock prices to boost hedge fund profits.
3 shares
Hedge funds have significantly reduced their short positions against the yen, the biggest cut in over ten years, due to suspected market interventions by Japan to strengthen the currency.
2 shares
An employee at Syquant Capital is suspected of giving insider information to his brother, leading to profitable trades on stocks like ArcelorMittal and Alstom before major market announcements.
2 shares
EEX Group has seen substantial volume growth in the first half of 2024, with global power market volume increasing by nearly 50% and US natural gas volumes almost doubling.
1 shares
Billionaire Ken Griffin bought the largest known stegosaurus fossil, Apex, for a record $44.6m.
1 shares
JPMorgan's product division operates differently from standard investment banking teams.
1 shares
WorldQuant, once exclusive to Millennium Management, now manages $10bn in assets for various clients.
1 shares
Episodes on markets, quant methods and economics.
10 items
Laurent Bernut talks about the influence of market psychology on investments, Japan's economic situation, and the need for risk management in investment adjustments.
12 shares
Cameron Dawson discusses managing high-value portfolios, the significance of tax sensitivity, and the potential of alternative investments.
9 shares
Jonny Goulden and Saad Siddiqui discuss the impact of recent market developments on the EM fixed income asset class in a July 18, 2024 podcast.
8 shares
A July 19, 2024 podcast explores the potential effects of GOP control on the commodities sector following recent political events, including an assassination attempt on Donald Trump.
7 shares
In the podcast Confessions of a Market Maker, Anne-Marie Baiynd shares her knowledge on mastering technical systems and trading psychology.
6 shares
Bianco Research's Jim Bianco predicts a 3-4% inflation rate due to remote work, deglobalization, and political energy manipulation, cautioning against Federal Reserve rate cuts.
6 shares
Bridgewater to Wealthsimple: Ben Reeves shares his transition from Bridgewater to Wealthsimple, discussing allocation research, portfolio design, and the private investment space's challenges and opportunities.
6 shares
USD Trends: Patrick Locke and Arindam Sandilya discuss the impact of recent US political changes on the broad dollar and investor sentiment in the FX asset class.
6 shares
Credit Market Trends: The Spreadbites podcast features discussions on global credit market trends, with insights from Eric Beinstein and Amy Sze, heads of U.S. High Grade Credit Research and ABS Research respectively.
5 shares
David Marom of Panoply Business highlights the advantages of all-in-one data platforms, including efficiency, data accuracy, and cost reduction, sharing success stories from Panoply users.
5 shares
Posts from quant and economics blogs and newsletters.
5 items
The HighLow tool is increasingly popular in the financial trading industry for improving technical analysis.
3 shares
The Zigzag Fibonacci is a tool used in technical analysis to identify patterns in the unpredictable fluctuations of financial markets.
3 shares
If Joe Biden withdraws as the Democratic nominee, many election prediction models would require significant adjustments due to reliability and continuity issues.
0 shares
The Democratic party is experiencing an allpay auction, where all factions must pay their bids even if they don't win.
0 shares
The article explores different facets and impacts of Artificial Intelligence.
0 shares
Posts from quant researchers on X.
10 items
Arnold et al.'s research investigates the fluctuation in portfolio weights across different asset classes and hedge fund strategies, particularly event-driven market neutral and global macro strategies.
4 shares
SP Global's white paper suggests that their unique dividend forecasts can predict short-term returns, attributing this to 'data arbitrage'.
3 shares
Thierry Roncalli's lecture notes provide a comprehensive overview of sustainable finance, including ESG rating systems, impact investing, portfolio construction, and climate risk measures.
3 shares
Arakelian et al.'s article explores the concept of backtesting, its potential pitfalls, and provides a list of references for further study.
3 shares
Melissa Dell's paper examines various deep learning models and their use in the field of economics.
1 shares
Jiang's team successfully predicts sector ETF returns from 30 minutes to a week using stock return data and LASSO methods.
1 shares
The article analyzes the investment trends and themes in crypto strategies for Q1 2024.
1 shares
The impact of CrowdStrike's global outage on Microsoft's stock price is examined in the article.
0 shares
Lorenzo Ravagli of JPMorgan explores the potential of exploiting the skew risk premium in foreign exchange markets.
0 shares