Pub Date : 2025-10-01Epub Date: 2024-11-04DOI: 10.1016/j.ijforecast.2024.10.001
Spyros Makridakis , Evangelos Spiliotis , Maria Michailidis
The M6 competition aimed to identify methods that can accurately forecast asset returns and exploit such forecasts to make efficient investments. Specifically, the forecasting track of the competition required participants to estimate the probability that each of the 100 selected assets would be ranked within the first, second, third, fourth, or fifth quintile with regards to their relative percentage returns. Overall, less than 25% of the teams managed to estimate the probabilities more precisely than a benchmark that assumed equal probabilities for all quintiles. Moreover, those that did so reported inconsistent performance across the 12 submission points and minor forecast accuracy improvements. We identify price volatility as a key driver of forecast deterioration and show that avoiding overconfidence by assuming similar probabilities for symmetric quintiles can improve both forecast accuracy and portfolio efficiency. Interestingly, our findings hold true even when simple methods are employed to estimate the base predictions and investment weights.
{"title":"Avoiding overconfidence: Evidence from the M6 financial competition","authors":"Spyros Makridakis , Evangelos Spiliotis , Maria Michailidis","doi":"10.1016/j.ijforecast.2024.10.001","DOIUrl":"10.1016/j.ijforecast.2024.10.001","url":null,"abstract":"<div><div>The M6 competition aimed to identify methods that can accurately forecast asset returns and exploit such forecasts to make efficient investments. Specifically, the forecasting track of the competition required participants to estimate the probability that each of the 100 selected assets would be ranked within the first, second, third, fourth, or fifth quintile with regards to their relative percentage returns. Overall, less than 25% of the teams managed to estimate the probabilities more precisely than a benchmark that assumed equal probabilities for all quintiles. Moreover, those that did so reported inconsistent performance across the 12 submission points and minor forecast accuracy improvements. We identify price volatility as a key driver of forecast deterioration and show that avoiding overconfidence by assuming similar probabilities for symmetric quintiles can improve both forecast accuracy and portfolio efficiency. Interestingly, our findings hold true even when simple methods are employed to estimate the base predictions and investment weights.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1395-1403"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-02-25DOI: 10.1016/j.ijforecast.2025.01.005
Filip Staněk
This article describes the methods that achieved fourth and sixth place in the forecasting and investment challenges, respectively, of the M6 competition, ultimately securing first place in the overall duathlon ranking. In the forecasting challenge, we tested a novel meta-learning model that utilizes hypernetworks to design a parametric model tailored to a specific family of forecasting tasks. This approach allowed us to leverage similarities observed across individual forecasting tasks (i.e., assets) while also acknowledging potential heterogeneity in their data generating processes. The model’s training can be directly performed with backpropagation, eliminating the need to rely on higher-order derivatives, and is equivalent to a simultaneous search over the space of parametric functions and their optimal parameter values. The proposed model’s capabilities extend beyond M6, demonstrating superiority over state-of-the-art meta-learning methods in the sinusoidal regression task and outperforming conventional parametric models on time series from the M4 forecasting competition. In the investment challenge, we adjusted portfolio weights to induce greater or smaller correlation between our submission and that of other participants, depending on the current ranking, aiming to maximize the probability of achieving a good rank. While this portfolio strategy can increase the probability of securing a favorable rank, it paradoxically exhibits negative expected returns.
{"title":"Designing time-series models with hypernetworks and adversarial portfolios","authors":"Filip Staněk","doi":"10.1016/j.ijforecast.2025.01.005","DOIUrl":"10.1016/j.ijforecast.2025.01.005","url":null,"abstract":"<div><div><span>This article describes the methods that achieved fourth and sixth place in the forecasting and investment challenges, respectively, of the M6 competition, ultimately securing first place in the overall duathlon ranking. In the forecasting challenge, we tested a novel meta-learning model that utilizes hypernetworks to design a parametric model tailored to a specific family of forecasting tasks. This approach allowed us to leverage similarities observed across individual forecasting tasks (i.e., assets) while also acknowledging potential heterogeneity in their data generating processes. The model’s training can be directly performed with </span>backpropagation<span>, eliminating the need to rely on higher-order derivatives, and is equivalent to a simultaneous search over the space of parametric functions and their optimal parameter values. The proposed model’s capabilities extend beyond M6, demonstrating superiority over state-of-the-art meta-learning methods in the sinusoidal regression task and outperforming conventional parametric models on time series from the M4 forecasting competition. In the investment challenge, we adjusted portfolio weights to induce greater or smaller correlation between our submission and that of other participants, depending on the current ranking, aiming to maximize the probability of achieving a good rank. While this portfolio strategy can increase the probability of securing a favorable rank, it paradoxically exhibits negative expected returns.</span></div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1461-1476"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020564","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-02-25DOI: 10.1016/j.ijforecast.2025.01.006
Rakshitha Godahewa , Christoph Bergmeir , Zeynep Erkin Baz , Chengjun Zhu , Zhangdi Song , Salvador García , Dario Benavides
Forecasts are typically produced in a business context on a regular basis to make downstream decisions. Here, forecasts should not only be as accurate as possible, but also should not change arbitrarily, and be stable in some sense. In this paper, we explore two types of forecast stability that we call vertical stability (for forecasts from different origins for the same target) and horizontal stability (for forecasts from the same origin for different targets). Existing works in the literature are only applicable to certain base models and can only stabilise forecasts vertically. We propose a simple linear-interpolation-based approach to stabilise the forecasts provided by any base model, both vertically and horizontally. Our method makes the trade-off between stability and accuracy explicit, producing forecasts at any point in the spectrum of this trade-off. We used N-BEATS, pooled regression, LightGBM, ETS, and ARIMA as base models in our evaluation across different error and stability measures on four publicly available datasets. On some datasets, the proposed framework achieved forecasts that were both more accurate and stable than the base forecasts. On the others, we achieved forecasts that were slightly less accurate but much more stable.
{"title":"On forecast stability","authors":"Rakshitha Godahewa , Christoph Bergmeir , Zeynep Erkin Baz , Chengjun Zhu , Zhangdi Song , Salvador García , Dario Benavides","doi":"10.1016/j.ijforecast.2025.01.006","DOIUrl":"10.1016/j.ijforecast.2025.01.006","url":null,"abstract":"<div><div>Forecasts are typically produced in a business context on a regular basis to make downstream decisions. Here, forecasts should not only be as accurate as possible, but also should not change arbitrarily, and be stable in some sense. In this paper, we explore two types of forecast stability that we call vertical stability (for forecasts from different origins for the same target) and horizontal stability (for forecasts from the same origin for different targets). Existing works in the literature are only applicable to certain base models and can only stabilise forecasts vertically. We propose a simple linear-interpolation-based approach to stabilise the forecasts provided by any base model, both vertically and horizontally. Our method makes the trade-off between stability and accuracy explicit, producing forecasts at any point in the spectrum of this trade-off. We used N-BEATS, pooled regression, LightGBM, ETS, and ARIMA as base models in our evaluation across different error and stability measures on four publicly available datasets. On some datasets, the proposed framework achieved forecasts that were both more accurate and stable than the base forecasts. On the others, we achieved forecasts that were slightly less accurate but much more stable.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1539-1558"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2024-11-30DOI: 10.1016/j.ijforecast.2024.11.002
Spyros Makridakis , Evangelos Spiliotis , Ross Hollyman , Fotios Petropoulos , Norman Swanson , Anil Gaba
The M6 forecasting competition, the sixth in the Makridakis competition sequence, focused on financial forecasting. A key objective of the M6 competition was to contribute to the debate surrounding the Efficient Market Hypothesis by examining how and why market participants make investment decisions. To address these objectives, the M6 competition investigated forecasting accuracy and investment performance in a universe of 100 publicly traded assets. The competition employed live evaluation on real data across multiple periods, a cross-sectional setting where participants predicted asset performance relative to that of other assets, and a direct evaluation of the utility of forecasts. In this way, we were able to measure the benefits of accurate forecasting and assess the importance of forecasting when making investment decisions. Our findings highlight the challenges that participants faced when attempting to accurately forecast the relative performance of assets, the great difficulty associated with trying to consistently outperform the market, the limited connection between submitted forecasts and investment decisions, the value added by information exchange and the “wisdom of crowds”, and the value of utilizing risk models when attempting to connect prediction and investing decisions.
{"title":"The M6 forecasting competition: Bridging the gap between forecasting and investment decisions","authors":"Spyros Makridakis , Evangelos Spiliotis , Ross Hollyman , Fotios Petropoulos , Norman Swanson , Anil Gaba","doi":"10.1016/j.ijforecast.2024.11.002","DOIUrl":"10.1016/j.ijforecast.2024.11.002","url":null,"abstract":"<div><div>The M6 forecasting competition, the sixth in the Makridakis competition sequence, focused on financial forecasting. A key objective of the M6 competition was to contribute to the debate surrounding the Efficient Market Hypothesis by examining how and why market participants make investment decisions. To address these objectives, the M6 competition investigated forecasting accuracy and investment performance in a universe of 100 publicly traded assets. The competition employed live evaluation on real data across multiple periods, a cross-sectional setting where participants predicted asset performance relative to that of other assets, and a direct evaluation of the utility of forecasts. In this way, we were able to measure the benefits of accurate forecasting and assess the importance of forecasting when making investment decisions. Our findings highlight the challenges that participants faced when attempting to accurately forecast the relative performance of assets, the great difficulty associated with trying to consistently outperform the market, the limited connection between submitted forecasts and investment decisions, the value added by information exchange and the “wisdom of crowds”, and the value of utilizing risk models when attempting to connect prediction and investing decisions.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1315-1354"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145019329","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-02-08DOI: 10.1016/j.ijforecast.2025.01.003
Benedikt Alexander Schuler , Johann Peter Murmann , Marie Beisemann , Ville Satopää
Judgmental forecasting research on superforecasters has demonstrated that individuals differ in their foresight. However, the concept underlying this work focuses on accuracy and does not fully incorporate the time dimension of foresight. We reconceptualize foresight as the ability to predict future states of the world accurately, where accuracy becomes continuously more important over time. To operationalize foresight in forecasting tournaments, we propose various strictly proper scoring rules and compare them with existing scoring rules using a simulation study and real-world forecasting data consisting of 414,168 scores for 9694 forecasters on 498 questions from a four-year geopolitical forecasting tournament. The results suggest that the linear time-weighted Brier score should be the default operationalization of foresight and that probability training and teaming interventions as proposed by prior research may not improve foresight as we conceptualize it. We contribute to judgmental forecasting research by clarifying the concept, operationalization, and correlates of foresight.
{"title":"Individual foresight: Concept, operationalization, and correlates","authors":"Benedikt Alexander Schuler , Johann Peter Murmann , Marie Beisemann , Ville Satopää","doi":"10.1016/j.ijforecast.2025.01.003","DOIUrl":"10.1016/j.ijforecast.2025.01.003","url":null,"abstract":"<div><div>Judgmental forecasting research on superforecasters has demonstrated that individuals differ in their foresight. However, the concept underlying this work focuses on accuracy and does not fully incorporate the time dimension of foresight. We reconceptualize foresight as the ability to predict future states of the world accurately, where accuracy becomes continuously more important over time. To operationalize foresight in forecasting tournaments, we propose various strictly proper scoring rules and compare them with existing scoring rules using a simulation study and real-world forecasting data consisting of 414,168 scores for 9694 forecasters on 498 questions from a four-year geopolitical forecasting tournament. The results suggest that the linear time-weighted Brier score should be the default operationalization of foresight and that probability training and teaming interventions as proposed by prior research may not improve foresight as we conceptualize it. We contribute to judgmental forecasting research by clarifying the concept, operationalization, and correlates of foresight.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1521-1538"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020568","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-03-05DOI: 10.1016/j.ijforecast.2025.02.004
Rodrigo Alves
Football (also known as soccer or association football) is the most popular sport in the world. It is a blend of skill and luck, making it highly unpredictable. To address this unpredictability, there has been a surge in popularity over the past decade in employing machine learning techniques for forecasting football-related features. This trend aligns with the growing professionalism in football analytics. Despite this progress, the existing body of work remains in its early stages, lacking the depth required to capture the intricate nuances of the sport. In this study, we introduce a convolutional approach designed to predict the occurrence of the next event in a football match, such as a goal or a corner kick, relying solely on easy-to-access past events for predictions. Our methodology adopts an online approach, meaning predictions can be computed during a live match. To validate our approach, we conduct a comprehensive evaluation against five baseline models, utilizing data from various elite European football leagues. Additionally, an ablation study is performed to understand the underlying mechanisms of our method. Finally, we present practical applications and interpretable aspects of our proposed approach.
{"title":"SCORE: A convolutional approach for football event forecasting","authors":"Rodrigo Alves","doi":"10.1016/j.ijforecast.2025.02.004","DOIUrl":"10.1016/j.ijforecast.2025.02.004","url":null,"abstract":"<div><div>Football (also known as soccer or association football) is the most popular sport in the world. It is a blend of skill and luck, making it highly unpredictable. To address this unpredictability, there has been a surge in popularity over the past decade in employing machine learning techniques<span><span> for forecasting football-related features. This trend aligns with the growing professionalism in football analytics. Despite this progress, the existing body of work remains in its early stages, lacking the depth required to capture the intricate nuances of the sport. In this study, we introduce a convolutional approach designed to predict the occurrence of the next event in a football match, such as a goal or a corner kick, relying solely on easy-to-access past events for predictions. Our methodology adopts an online approach, meaning predictions can be computed during a live match. To validate our approach, we conduct a comprehensive evaluation against five </span>baseline models, utilizing data from various elite European football leagues. Additionally, an ablation study is performed to understand the underlying mechanisms of our method. Finally, we present practical applications and interpretable aspects of our proposed approach.</span></div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1636-1652"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145019322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The M6 competition evaluated investment performance over a period of one year, contributing to the efficient market hypothesis debate. This paper provides further insights into the outcomes of the competition by unraveling the effect that team engagement and performance consistency had on the final results. First, we identify three different types of engagement and investigate their relationship with portfolio efficiency, also making useful observations about the learning effect implied by a re-submission process. Then, we analyze the monthly performance of the teams and determine whether it aligned with their global performance or was affected significantly by extreme instances. Our results suggest that consistency is more important than engagement for making profitable investments. Nevertheless, we identify many cases where both regular portfolio updates and luck provided an advantage.
{"title":"Unraveling the effect of engagement and consistency in the results of the M6 forecasting competition","authors":"Anastasios Kaltsounis, Evangelos Theodorou, Evangelos Spiliotis, Vassilios Assimakopoulos","doi":"10.1016/j.ijforecast.2025.04.002","DOIUrl":"10.1016/j.ijforecast.2025.04.002","url":null,"abstract":"<div><div>The M6 competition evaluated investment performance over a period of one year, contributing to the efficient market hypothesis debate. This paper provides further insights into the outcomes of the competition by unraveling the effect that team engagement and performance consistency had on the final results. First, we identify three different types of engagement and investigate their relationship with portfolio efficiency, also making useful observations about the learning effect implied by a re-submission process. Then, we analyze the monthly performance of the teams and determine whether it aligned with their global performance or was affected significantly by extreme instances. Our results suggest that consistency is more important than engagement for making profitable investments. Nevertheless, we identify many cases where both regular portfolio updates and luck provided an advantage.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1404-1412"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-02-26DOI: 10.1016/j.ijforecast.2025.02.001
Luis Gruber, Gregor Kastner
Vector autoregressions (VARs) are widely applied when it comes to modeling and forecasting macroeconomic variables. In high dimensions, however, they are prone to overfitting. Bayesian methods—more concretely, shrinkage priors—have been shown to be successful at improving prediction performance. In the present paper, we introduce the semi-global framework, in which we replace the traditional global shrinkage parameter with group-specific shrinkage parameters. We show how this framework can be applied to various shrinkage priors, such as global–local priors and stochastic search variable selection priors. We demonstrate the virtues of the proposed framework in an extensive simulation study and in an empirical application forecasting data on the US economy. Further, we shed more light on the ongoing ‘illusion of sparsity’ debate, finding that forecasting performances under sparse/dense priors vary across evaluated economic variables and across time frames. Dynamic model averaging, however, can combine the merits of both worlds.
{"title":"Forecasting macroeconomic data with Bayesian VARs: Sparse or dense? It depends!","authors":"Luis Gruber, Gregor Kastner","doi":"10.1016/j.ijforecast.2025.02.001","DOIUrl":"10.1016/j.ijforecast.2025.02.001","url":null,"abstract":"<div><div>Vector autoregressions (VARs) are widely applied when it comes to modeling and forecasting macroeconomic variables. In high dimensions, however, they are prone to overfitting. Bayesian methods—more concretely, shrinkage priors—have been shown to be successful at improving prediction performance. In the present paper, we introduce the semi-global framework, in which we replace the traditional global shrinkage parameter with group-specific shrinkage parameters. We show how this framework can be applied to various shrinkage priors, such as global–local priors and stochastic search variable selection priors. We demonstrate the virtues of the proposed framework in an extensive simulation study and in an empirical application forecasting data on the US economy. Further, we shed more light on the ongoing ‘illusion of sparsity’ debate, finding that forecasting performances under sparse/dense priors vary across evaluated economic variables and across time frames. Dynamic model averaging, however, can combine the merits of both worlds.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1589-1619"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145019320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2024-07-12DOI: 10.1016/j.ijforecast.2024.06.007
Dan Weitzenfeld
The M6 Financial Forecasting Competition forecasting track required probabilistic forecasting of monthly returns for a universe of 100 assets. This paper describes a Bayesian dynamic factor model with heteroskedasticity that was used to win the year-long forecasting track. The model’s strengths include modularity, handling of missing data, and regularization through hierarchical distributions. Probability modeling and recent advances in probabilistic programming languages make defining such models and performing inference straightforward.
{"title":"Probabilistic forecasting of cross-sectional returns: A Bayesian dynamic factor model with heteroskedasticity","authors":"Dan Weitzenfeld","doi":"10.1016/j.ijforecast.2024.06.007","DOIUrl":"10.1016/j.ijforecast.2024.06.007","url":null,"abstract":"<div><div><span>The M6 Financial Forecasting Competition forecasting track required probabilistic forecasting of monthly returns for a universe of 100 assets. This paper describes a Bayesian dynamic factor model with </span>heteroskedasticity<span> that was used to win the year-long forecasting track. The model’s strengths include modularity, handling of missing data, and regularization through hierarchical distributions. Probability modeling and recent advances in probabilistic programming languages make defining such models and performing inference straightforward.</span></div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1477-1484"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141703433","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-01Epub Date: 2025-08-28DOI: 10.1016/j.ijforecast.2025.08.004
Colin Catlin
In contemporary forecasting, the challenges of navigating the intricacies of erratic human-induced patterns combine with the challenges of navigating the overwhelming number of methods and models available to manage these data. The M6 Competition, which emphasized repeated, real-time monthly forecasting of stock markets, featured many of these difficulties. Here, AutoTS, an open-source Python package designed specifically for probabilistic time series predictions, is evaluated within the context of this competition. AutoTS includes an extensive repertoire of models, augmented by robust data preprocessing utilities, and employs genetic algorithms to fine-tune model parameters, contingent upon user-delineated evaluation metrics. This study describes the deployment of AutoTS in the M6 Competition, which won the investment decision challenge, and outlines the model selection pipeline and the process of converting forecasts into decisions which produced this result. Although a single definitive model remains elusive, these findings underscore the potential value of methodologies that are dynamic and largely autonomous.
{"title":"Adaptive forecasting in dynamic markets: An evaluation of AutoTS within the M6 competition","authors":"Colin Catlin","doi":"10.1016/j.ijforecast.2025.08.004","DOIUrl":"10.1016/j.ijforecast.2025.08.004","url":null,"abstract":"<div><div>In contemporary forecasting, the challenges of navigating the intricacies of erratic human-induced patterns combine with the challenges of navigating the overwhelming number of methods and models available to manage these data. The M6 Competition, which emphasized repeated, real-time monthly forecasting of stock markets, featured many of these difficulties. Here, AutoTS, an open-source Python package designed specifically for probabilistic time series predictions, is evaluated within the context of this competition. AutoTS includes an extensive repertoire of models, augmented by robust data preprocessing utilities, and employs genetic algorithms to fine-tune model parameters, contingent upon user-delineated evaluation metrics. This study describes the deployment of AutoTS in the M6 Competition, which won the investment decision challenge, and outlines the model selection pipeline and the process of converting forecasts into decisions which produced this result. Although a single definitive model remains elusive, these findings underscore the potential value of methodologies that are dynamic and largely autonomous.</div></div>","PeriodicalId":14061,"journal":{"name":"International Journal of Forecasting","volume":"41 4","pages":"Pages 1485-1493"},"PeriodicalIF":7.1,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145020565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号