Predicting Round and Game Winners in CSGO

Win probabilities have become a staple on scoreboards in physical sports such as baseball and basketball. Esports, or competitive video games with sponsored teams and major audiences, typically lack this detailed statistical analysis, beyond bare-bones metrics and commentator intuition. However, the advantage of esports in their tendency to have a central record of every game event makes them ripe for statistical analysis through machine learning. Previous research has covered popular video game genres such as MOBAs, and has found success in predicting game winners most of the time [1]. Counterstrike: Global Offensive (CSGO) is an esport that is unique in its round and game-based nature, allowing researchers to examine how short and long-term decisions can interplay in competitive environments. We introduce a dataset of CSGO games To assess factors such as player purchasing decisions and individual scores, we introduce 3 round and game win probability models. Finally, we evaluate the performances of the models. We successfully predict winners in the majority of cases, better than the map average baseline win statistics.

Family background and classroom belonging among adolescent students in Finland

We study whether a family’s economic situation and parental educational level are associated with classroom belonging among students in comprehensive secondary, upper secondary general and upper secondary vocational education in Finland. We also study whether there are educational-level differences in this possible association. We use survey data from the Finnish School Health Promotion study from 2017 (N = 114,528). We conduct random effect linear probability models with schools as the second-level grouping variable. The results show that family’s low economic situation predicts a higher probability of lack of sense of classroom belonging in Finland, despite the country having one of the world’s most equal educational systems and comparably low economic inequality. Neither mother’s nor father’s educational level has any association. A family’s low economic situation seems to predict the lack of a sense of belonging most strongly in comprehensive secondary education and most weakly in upper secondary vocational education. Our results slightly support the proposed significance of context-specific hierarchies in determining the association between economic resources and sense of belonging. A family having a poor economic situation is not reflected in the sense of classroom belonging as strongly in schools where students have a low average economic situation compared to those where students have a high average economic situation. We suggest measures, in addition to alleviating economic inequalities, to support the sense of school belonging, especially for low-income students.

Evaluation of an integrative Bayesian peptide detection approach on a combinatorial peptide library

Detection of peptides lies at the core of bottom-up proteomics analyses. We examined a Bayesian approach to peptide detection, integrating match-based models (fragments, retention time, isotopic distribution, and precursor mass) and peptide prior probability models under a unified probabilistic framework. To assess the relevance of these models and their various combinations, we employed a complete- and a tail-complete search of a low-precursor-mass synthetic peptide library based on oncogenic KRAS peptides. The fragment match was by far the most informative match-based model, while the retention time match was the only remaining such model with an appreciable impact––increasing correct detections by around 8 %. A peptide prior probability model built from a reference proteome greatly improved the detection over a uniform prior, essentially transforming de novo sequencing into a reference-guided search. The knowledge of a correct sequence tag in advance to peptide-spectrum matching had only a moderate impact on peptide detection unless the tag was long and of high certainty. The approach also derived more precise error rates on the analyzed combinatorial peptide library than those estimated using PeptideProphet and Percolator, showing its potential applicability for the detection of homologous peptides. Although the approach requires further computational developments for routine data analysis, it illustrates the value of peptide prior probabilities and presents a Bayesian approach for their incorporation into peptide detection.

Food allergens: threshold levels and methodologies for risk management

Food allergies and allergen management are important problems of the public health and food industry. The idea of determining allergen concentrations in food ingredients and food products that are capable of causing severe allergic reactions is of great interest for regulatory bodies as well as consumer associations and the industry all over the world. In this connection, scientists proposed different approaches to determining the basis for assessment of severity of risks of food allergens for health of patients suffering from food allergy similar to methods of risk assessment for other hazards associated with food products (for example, chemical, microbiological). To assess risk of allergens, three different approaches were proposed: i) traditional risk assessment using the no observed adverse effect level (NOAEL)) and uncertainty factors; (ii) approach based on the benchmark dose (BMD)) and margin of exposure (MoE)); and (iii) probability models. These approaches can be used in risk management in food production and in the development of warning marking about the presence of allergens. The reliability of risk assessment will depend on a type, quality and quantity of data used for determining both population threshold levels (or threshold distributions) and an impact of an allergenic product/ingredient on a particular individual.

Improved approach to wave potential estimation using bivariate distributions

Probabilistic approaches are frequently used to describe irregular activity data to assist the design and development of devices. Unfortunately, useful estimations are not always feasible due to the large noise in the data modeled, as it occurs when estimating the sea waves potential for electricity generation. In this work we propose a simple methodology based on the use of joint probability models that allow discriminating extreme values, collected from measurements as pairs of independent points, while allowing the preservation of the essential statistics of the measurements. The outcome of the proposed methodology is an equivalent data series where large-amplitude fluctuations are suppressed and, therefore, can be used for design purposes. For the evaluation of the proposed method, we used year-long databases of hourly-collected measurements of the wave’s height and period, performed at maritime buoys located in the Gulf of Mexico. These measurements are used to obtain a fluctuations-reduced representation of the energy potential of the waves that can be useful, for instance, for the design of electric generators.

Regularization, Bayesian Inference, and Machine Learning Methods for Inverse Problems

Classical methods for inverse problems are mainly based on regularization theory, in particular those, that are based on optimization of a criterion with two parts: a data-model matching and a regularization term. Different choices for these two terms and a great number of optimization algorithms have been proposed. When these two terms are distance or divergence measures, they can have a Bayesian Maximum A Posteriori (MAP) interpretation where these two terms correspond to the likelihood and prior-probability models, respectively. The Bayesian approach gives more flexibility in choosing these terms and, in particular, the prior term via hierarchical models and hidden variables. However, the Bayesian computations can become very heavy computationally. The machine learning (ML) methods such as classification, clustering, segmentation, and regression, based on neural networks (NN) and particularly convolutional NN, deep NN, physics-informed neural networks, etc. can become helpful to obtain approximate practical solutions to inverse problems. In this tutorial article, particular examples of image denoising, image restoration, and computed-tomography (CT) image reconstruction will illustrate this cooperation between ML and inversion.

Negative Binomial Distribution to Explain the Domestic Fire Incidence in Nepal

Background: Fire disaster is one of the most destructive disasters. According to global dataset of Sendai Framework, domestic fire incidence was 9.9% up to 2019. In Nepal, 62% fire incidence was reported during 2017 and 2018. However, many studies have been conducted on fire incidence, few of them are based on domestic fire incidence. Objective: To find the descriptive statistics of fire occurrences and fire fatalities, and to identify the probability distributions that best fit the data of fire occurrences observed in three ecological regions as well as overall in Nepal. Material and Methods: The data of fire incidences from May 2011 to April 2021 were retrieved from Nepal Disaster Risk Reduction Portal, Government of Nepal. At first, a statistical software "Mathwave EasyFit" of 30 days trial version was used to identify the candidate probability models. Further, the best probability model was determined after testing the goodness of fit of the candidate models by using graphical tools-histogram and theoretical densities, empirical and theoretical CDFs, Q-Q plot and P-P plot; and mathematical tools-maximum likelihood, Akaike Information Criteria and Bayesian Information Criteria by using the package “fitdistrplus” of software R version 4.1.1. Results: On an average, 135 fire incidences per month were occurred in Nepal. However, the Terai faced the highest monthly fire incidences compared to the Hill and the Mountain, it has less fatality per 100 fire incidence followed by the Hill and the Mountain. Descriptive statistics reveals that fire occurrences are moderate during November to February and high in March and April. The fire incidences were reported high during spring and winter and low during summer and autumn season which reveals that fire incidence might be related with the precipitation and temperature. The sample data was run in "Mathwave EasyFit" software which suggested Poisson, geometric and negative binomial distribution as candidate probability models. The goodness of fit of these models were further tested by graphical as well as mathematical tools where negative binomial distribution was found to be best among the candidate models for the data set. Conclusion: Incidence of fire disasters varies by ecological regions as well as by seasons. It is low in the Mountain region and during Monsoon/rainy season. Negative binomial distribution fits the best to monthly data of fire incidence in Nepal.

Local Opt-Out Provisions in a State Gambling Expansion: An Analysis of Pennsylvania Municipalities

A 2017 gambling expansion bill in Pennsylvania included a provision that gave municipalities the option to ban a new casino from opening within their borders. This paper examined how different factors influenced local decisions on whether to allow casino gambling. Multilevel linear probability models indicated that municipalities were influenced by economic characteristics, as evidenced by a higher likelihood of allowing casinos in communities with lower levels of household income. Results also suggested that municipalities were influenced by variables related to tax competition and the percentage of residents who were black. The findings of this study identify factors that may influence municipal governments when given the authority to opt out of a state gambling expansion capable of generating a new source of local tax revenue.