Int J Appl Sports Sci > Volume 33(2); 2021 > Article
Park, Chang, Ahn, Kim, and Lee: Comparative analysis of win and loss factors in women's handball using international competition records

ABSTRACT

This study identifies factors affecting match results from major international competitions in women's handball in the last four years. The 12 countries that participated for the 2020 Tokyo Olympics were included in the analysis, and a total of 281 matches from 4 major international competitions were analyzed. To identify factors affecting winning and losing, independent sample t-test and logistic regression analysis were conducted on the variables present in the official records. The findings present several factors that have positive and negative effects on match results. In the analysis of differences in win and loss factors, 6m goals success rate, 9m success rate, FB goals and shooting, AS, BS, and ST had positive effects on winning. Logistic regression analysis had 84.5% accuracy. 6m and Wing goal, 9m success rate, FB shooting, GK Wing save rate, and GK 9m save rate increased the probability of winning.

Introduction

The development of technology has enabled the measurement of athletes’ movements and the analysis of movement data. Technology has also been widely used in the sports industry to improve performance in combination with sports science (Fujii, 2021; Kim, 2012). Recently, technology has been developed for the analysis of game records data, allowing for the quantitative analysis of various events and phenomena observed during games. In addition, game contents and progress can be analyzed in real-time, providing information for the real-time adaptation of tactics and strategies (Kim et al., 2008, Fernandez et al., 2006).
Handball is a representative ball sports game that requires the analysis of the opponent's strategy and tactical information. To collect information on the opponent, game records and videos are analyzed (Luteberget, 2018; Kim, 2012). Since the 1996 Barcelona Olympics, all events that occur during matches for all participating teams in international handball competitions have been recorded by professional record keepers from the International Handball Federation (IHF). All results are available on the Federation’s website immediately after each game. These records are used as basic data for individual movement and scientific training and are an essential factor for establishing skills and tactics. Analysis methods are continuously being further developed (Taborsky, 2011; Bilge, 2012), and accurate and detailed analyses of opposing teams are more important than ever. Such analyses are considered to be an essential factor for coaches’ effective tactics display (Jung, 2006).
The IHF is continuously modifying game rules to enable fast-pace and exciting matches. The pace of transition between offense and defense is greater than that in the past, such as Quick start that leads to an immediate attack after conceding goals. This has increased the demand for high-intensity exercise. Since 2016, empty goal strategy with extra field players participating in attacks instead of goalkeepers also shows the diversity of attack strategies (IHF, 2016, 2019). Sevim and Bilge (2007) and Pokrajac (2010) reported that the new rules on QuickStart have allowed swift attacks in possession of the ball and led to more dynamic and diverse strategies for top-tier teams.
In previous studies conducted in Korea and other countries, Kim (2012) and Kim et al. (2013) found that the success rate of 6m and 9m shots as well as defensive factors such as blocks and steals affect match results. Kim et al. (2011) and Hong and Park (2016) developed an objective model to evaluate players' goals per position. Srhoj et al. (2001) reported that the movement of back position players and goals from swift attack and breakthrough have decisive effects on the final results of matches and that the number of shots from specific locations do not affect the match outcome. Similarly, Bilge (2012) analyzed the results of the World Championships and European Championships and reported that fast swift attack and efficient movement of pivot and back position players affect team standings. In addition, Pfeiffer and Perl (2006) analyzed the tactical structure using an artificial neural network analysis technique and created and applied an optimized attack pattern. However, these studies analyzed one to two competitions before the revision of the game rules. Thus, their findings cannot be generalized, and new studies are needed to analyze the recent international trend of women's handball and evaluate their tactical characteristics.
In this study, we aimed to provide basic data for the establishment of customized tactics and strategies and the development of training programs by assessing changes in the main factors that determine the outcome of handball games using data from recent international competitions in women's handball.

Methods

Analysis target

To analyze global trends in women's handball, a total of 281 games from the 12 countries that participated in the 2020 Tokyo Olympics were analyzed. IHF official data for the 2017 IHF World Women's Handball Championship (66 games), the 2018 European Championship (54 games), the 2019 HF World Women's Handball Championship (96 games), and the 2021 Tokyo Olympics (65 games) were collected. As a result of the game, 169 wins and 112 losses were classified in a total of 281 games and used for analysis. The characteristics of the data from each competition are shown in Table 1.
Table 1.

Data from each competition

Nation 2017 WC 2018 EC 2019 WC 2020 OL Result
1 Netherlands 7 8 10 4 22 W 7 L
2 Norway 7 7 10 8 25 W 7 L
3 Russia 6 8 10 7 23 W 8 L
4 Montenegro 5 6 9 6 15 W 11L
5 Brazil 4 0 6 4 6 W 8 L
6 Sweden 7 6 8 6 17 W 10 L
7 Spain 4 6 9 5 12 W 12 L
8 Angola 5 0 7 4 5 W 11 L
9 Japan 4 0 8 5 6 W 11 L
10 France 7 7 6 6 21 W 5 L
11 Korea 5 0 6 5 7 W 9 L
12 Hungary 5 6 7 5 10 W 13 L
Total 66 54 96 65

WC: World women's handball championship, EC: European women's handball championship

OL: Olympic, W: win, L: Loss

Analysis variables

A total of 43 variables including 18 shooting variables, 2 offense variables, 2 defense variables, 3 penalty variables, and 18 goalkeeping variables were analyzed.
Shooting variables of Goal, number of shots and success rates included shots from 6m, wing, 9m, and 7m positions. FastBreaks (FB) indicates swift attacks from counterattacks, and Break Through s(BT) indicates shots after breakthroughs. Offense variables included assists (AS) and turnover (TO), which indicates giving away the possession of the ball to the opponent. Defense variables included steals (ST) and blocked shots (BS). Penalty variables were yellow card (YC), 2-minute suspension (2min), and red card (RC). Goalkeeping (GK) variables were number of saves (SV) per shooting location, number of shots allowed (SH), and the save rate compared to the total number of shots (%).

Data analysis

To identify factors that affect match outcomes and present differences between winning and losing matches, the collected data were analyzed per country using Excel 2016 (Microsoft, USA).
A t-test was conducted to assess the differences in the variables between winning and losing matches, and logistic regression analysis was performed to analyze the factors that determine wins and losses. In logistic regression analysis, β is the logistic regression coefficient of statistically selected independent variables, and SE is the standard error considering the number of samples. Wald value is calculated by dividing β by SE to verify the significance of the logistic regression coefficient and allows the verification of the χ2 distribution. Exp(B) was odds ratio. Exp(B) equal to 1, greater than 1, and less than 1 indicated invalid, positive, and negative effects, respectively (Kim et al., 2008). A p value of >.05 was considered statistically significant, and all data were analyzed using SPSS 23.0 (IBM, Chicago, IL, USA). Feedforward selection that allows the program to automatically select and analyze statistically significant variables was conducted for the logistic regression analysis.

Results

Verification of differences in win and loss factors

Among the 18 shooting variables, a total of 7 variables showed significant differences: 6m goal (t= 3.522, p= .000), 6m success rate (t= 3.198, p=. 002), 9m shooting (t= -2.730, p= .007), 9m success rate (t= 4.721, p= .000), 7m shooting (t= -2.323, p= .021), FB goal (t= 9.224, p= .000), and FB shooting (t= 9.282, p= .000) (Table 2).
Table 2.

Results of analysis of difference between winning and losing factors related to shooting

 Variables Result Mean SD t p
6m Goal Win 7.54 3.55 3.522 .000
Loss 6.10 3.08
Shooting Win 11.11 5.07 1.563 .119
Loss 10.14 5.12
Success rate Win 70.17 17.30 3.198 .002
Loss 63.16 18.97
Wing Goal Win 5.56 3.07 1.903 .058
Loss 4.86 2.99
Shooting Win 8.83 4.16 .024 .981
Loss 8.82 4.76
Success rate Win 62.69 22.13 -.018 .986
Loss 62.82 91.99
9m Goal Win 5.49 3.05 1.731 .084
Loss 4.87 2.75
Shooting Win 12.20 5.52 -2.730 .007
Loss 14.06 5.71
Success rate Win 46.05 17.91 4.721 .000
Loss 35.73 17.97
7m Goal Win 3.30 1.94 -1.299 .195
Lose 3.61 2.00
Shooting Win 4.21 2.34 -2.323 .021
Loss 4.88 2.42
Success rate Win 77.54 24.33 .622 .534
Loss 75.73 23.13
FB Goal Win 5.13 3.65 9.224 .000
Loss 2.13 1.73
Shooting Win 6.56 4.35 9.282 .000
Loss 2.93 2.15
Success rate Win 75.44 24.30 1.913 .057
Loss 68.34 33.92
BT Goal Win 3.33 2.52 1.088 .278
Loss 3.01 2.29
Shooting Win 4.12 2.97 -.143 .887
Loss 4.17 2.93
Success rate Win 74.86 30.93 1.800 .073
Loss 67.99 31.75
Among the seven variables related to offense, defense, and penalties, AS (t= 7.264, p= .000), TO (t= -4.770, p= .000), BS (t= 5.501, p= .000), and ST (t= 4.886, p= .000) showed significant differences (Table 3).
Table 3.

Results of analysis of difference between winning and losing factors related to offense, defense, and penalties

 Variables Result Mean SD t p
AS Win 16.24 5.43 7.264 .000
Loss 12.15 3.96
TO Win 11.43 3.34 -4.770 .000
Loss 13.45 3.67
BS Win 2.82 2.25 5.501 .000
Loss 1.58 1.52
ST Win 3.95 2.11 4.886 .000
Loss 2.76 1.80
YC Win 1.53 1.08 -.158 .875
Loss 1.55 1.11
RC Win 0.08 0.27 -.869 .385
Loss 0.11 0.31
2min Win 3.62 1.86 1.436 .152
Loss 3.33 1.52
Table 4.

Results of analysis of difference between winning and losing factors related to GK

Variables   Result Mean SD t p
6m Save Win 2.92 2.10 2.610 .010
Loss 2.29 1.85
6m Shooting Win 8.92 4.32 -1.791 .074
Loss 9.91 4.82
6m Save Rates Win 30.99 17.28 3.921 .000
Loss 23.06 15.52
Wing Save Win 2.56 1.85 3.836 .000
Loss 1.76 1.50
Wing Shooting Win 6.70 3.73 -.424 .672
Loss 6.88 3.37
Wing Save Rates Win 40.17 23.93 5.233 .000
Loss 25.63 20.88
9m Save Win 4.91 2.71 4.852 .000
Loss 3.56 1.93
9m Shooting Win 9.91 4.40 1.005 .316
Loss 9.38 4.22
9m Save Rates Win 49.55 19.82 4.449 .000
Loss 38.91 19.35
7m Save Win 0.76 0.92 2.374 .018
Loss 0.54 0.68
7m Shooting Win 3.97 2.16 -.172 .864
Loss 4.02 2.42
7m Save Rates Win 17.30 21.07 1.437 .152
Loss 13.73 19.29
FB Save Win 0.52 0.76 -2.322 .021
Loss 0.75 0.89
FB Shooting Win 2.46 1.79 -7.038 .000
Loss 4.63 2.91
FB Save Rates Win 18.27 28.10 .579 .563
Loss 16.54 21.76
BT Save Win 0.66 0.91 1.248 .213
Loss 0.53 0.77
BT Shooting Win 3.23 2.67 -2.177 .030
Loss 3.97 2.98
BT Save Rates Win 16.57 24.12 2.061 .040
Loss 11.35 18.20
Table 5.

Logistic regression analysis results for winning or losing factors

Variables β SE Wald df p Exp(B)
6m Goal .336 .073 21.177 1 .000 1.399
Wing Goal .562 .130 18.605 1 .000 1.754
Wing Shooting -.274 .085 10.478 1 .001 0.760
9m Success rate .046 .011 16.068 1 .000 1.047
FB Shooting .485 .087 31.118 1 .000 1.624
GK 6m Shooting -.212 .049 18.551 1 .000 0.809
GK Wing Save rate .044 .010 20.963 1 .000 1.045
GK 9m Save rate .036 .011 10.867 1 .001 1.036
GK FB Save -.404 .092 19.443 1 .000 0.668

β: logistic regression coefficient, S.E.: standard error, Wald: X2 distribution verification statistics, p: p-value, Exp(B) = 1: invalid, Exp(B) > 1 positive effect, 0 < Exp(B) < 1: negative effect

Table 6.

Logistic regression analysis results for the factors of winning or losing each competition from 2017 to 2021

Variables β SE Wald df p Exp(B)
2017 FB Goal 1.257 0.386 10.609 1 0.001 3.516
GK Wing Save 1.295 0.530 5.980 1 0.014 3.652
GK FB Shooting -0.719 0.297 5.860 1 0.015 0.487
X2 =59.051, df=3, p= .000
2018 9m Shooting 0.443 0.173 6.568 1 0.010 1.557
AS 0.580 0.205 7.980 1 0.005 1.785
ST 1.223 0.576 4.517 1 0.034 3.398
GK Wing Save Rates 0.097 0.041 5.631 1 0.018 1.102
GK FB Shooting -1.121 0.384 8.514 1 0.004 0.326
X2 =47.976, df=4, p= .000
2019 6m Shooting 0.274 0.123 4.980 1 0.026 1.316
9m Success Rate 0.170 0.051 11.003 1 0.001 1.186
FB Goal 0.867 0.333 6.784 1 0.009 2.379
AS 0.486 0.196 6.139 1 0.013 1.626
TO -0.560 0.215 6.751 1 0.009 0.571
GK 6m Shooting -0.441 0.175 6.342 1 0.012 0.644
GK Wing Save Rates 0.098 0.032 9.305 1 0.002 1.103
GK 9m Save Rates 0.046 0.021 4.615 1 0.032 1.047
GK 7m Save 2.331 0.953 5.983 1 0.014 10.289
X2 =84.286, df=9, p= .000
2020 6m Goal 0.515 0.173 8.800 1 0.003 1.673
FB Shooting 0.576 0.199 8.370 1 0.004 1.779
AS 0.303 0.111 7.423 1 0.006 1.354
GK BT Save Rates 0.054 0.023 5.389 1 0.020 1.055
X2 =42.393, df=4, p= .000

β: logistic regression coefficient, S.E.: standard error, Wald: X2 distribution verification statistics, p: p-value, Exp(B) = 1: invalid, Exp(B) > 1 positive effect, 0 < Exp(B) < 1: negative effect

Among the 18 variables related to goalkeeping, 11 variables showed significant differences: 6m save (t= 2.610, p= .010), 6m save rate (t= 3.921, p= .000), Wing save (t= 3.836, p= .000), Wing save rate (t= 5.233, p= .000), 9m save (t= 4.852, p= .000), 9m save rate (t= 4.449, p= .000), 7m save (t= 2.374, p= .018), FB save (t= -2.322, p= .021), FB shooting (t= -7.038, p= .000), BT shooting (t= -2.177, p= .030), and BT save rate (t= 2.061, p= .040).

Logistic regression analysis of win and loss factors

The logistic regression analysis of independent variables predicting wins and losses had statistical significance and an accuracy of χ2=197.441, p<.000. The logistic regression model correctly predicted 78% of wins and 88.7% of losses with an overall accuracy of 84.5%.
Among the 43 independent variables, 6m goal (β= 0.336, Wald= 21.177, p= .000), Wing goal (β= 0.562, Wald= 18.605, p= .000), Wing shooting (β= -0.274, Wald= 10.478, p= .001), 9m success rate (β= 0.046, Wald= 16.068, p= .000), FB shooting (β= 0.485, Wald= 31.118, p= .000), GK 6m shooting (β= -0.212, Wald= 18.551, p= .000), GK Wing save rate (β= 0.044, Wald= 20.963, p= .000), GK 9m save rate (β= 0.036, Wald= 10.867, p= .001), and GK FB save (β= -0.404, Wald= 19.443, p= .000) affected wins and losses.
The changes in important factors determining wins and losses in women's handball international competitions were assessed per year. Among the 43 variables, 3 variables in 2017 (FB goal, GK Wing save, GK FB shooting), 5 variables in 2018 (9m shooting, AS, ST, GK Wing save rates, GK FB shooting), 9 variables in 2019 (6m shooting, 9m success rate, FB goal, AS, TO, GK 6m shooting, GK Wing save rate, GK 9m save rate, GK 7m save), and 4 variables in 2020 (6m goal, FB shooting, AS, GK BT save rate) affected wins and losses.

Discussion

In this study, we observed that the factors determining wins and losses in women's handball matches diversified after 2017. Kim(2012) reported that the 7m and Wing success rate had a low influence on the match results, and Kim (2021) stated that the 6m and 9m success rate had significant effects on the match results. However, in our analysis of international competitions from multiple years rather than single competitions, Wing's goal and FB shooting had the greatest effects on the match results. This may be attributed to the Quick Start system, which has led to faster transitions between offense and defense, the empty goal rule with seven players on field, and diverse tactics using the space in the wing position for penalties. Players in the Wing position travel the longest distance in games, move at the fastest pace during swift attack, and have become an essential part of recent handball team strategies.
Empty goal strategy that adds an extra field player instead of a goalkeeper is used during matches by approximately 88% of all participating teams. In the 2019 IHF Women's Handball World Championship, 11.3% of total offense was executed by the extra player replacing the goalkeeper with an average of 5.3 goals per game. The extra players also accounted for 9.9% of total goals [IHF Education Center].(2021,Oct 31). https://ihfeducation.ihf.info
Kim et al. (2013) reported that a goalkeeper save rate greater than 35.29% and shooting success rate greater than 56.40% were associated with a 91.11% probability of winning. In addition, weighted defense goal’s conceded balance index for each position was greater than 12.6% and was associated with a 100% probability of winning. In agreement with these findings, factors related to goalkeeping such as the number of saves and save rate affected wins and losses of matches. In particular, increased FB saves had negative effects on the match results (Exp(B)=0.668). As goalkeepers face one-on-one situations in most contexts except for 9m shooting, many aspects of a goalkeeper's record are closely related to the team's defense. Thus, the enhancement of teamwork in defense such as backcourt transitions that are not reflected in game records would have greater impacts on the match results.
6m and 9m goals (success rate) were also important factors affecting the match results. This indicates that basic offense formation to penetrate the opponent's defense is a basic requirement for winning matches. Consistent with our findings, previous studies also reported that winning teams had a balanced offense (Rogulj, 2000; Ferrari et al., 2014).
In another study that analyzed factors affecting wins and losses in the Men's European Championship, AS was found to have significant effects on match results (Ferrari et al., 2020). However, although AS was found to be significantly different between the won and lost matches of women's handball in this study (t= 7.264, p= .000), the regression analysis did not show significant differences. This may be due to the distinct characteristics of women's matches, in which offense patterns using FB and BT are mainly used in crowded spaces instead of AS.
Our findings suggest that back and pivot position players who mainly stay in 6m and 9m areas that are important factors for the most basic attack type (offense), GK defense capacity, increased frequency of FB for fast-paced matches, and the role of Wing players have become important factors in recent international competitions for women's handball. In addition, transition into defense and teamwork were factors with significant effects on match results.

Conclusion

In this study, we aimed to analyze the match records of major international competitions for women's handball in the last four years and identified factors affecting match results to provide basic data for the establishment of tactics, strategies, and training programs. The following results were observed.
In the analysis of differences in win and loss factors, 6m goals success rate, 9m success rate, FB goals and shooting, AS, BS, and ST had positive effects on winning. In contrast, 9m and 7m shooting and TO had negative effects. Among the factors related to goalkeeping, 6m wing, and 9m saves and save rate, 7m saves, and BT save rate had positive effects on the match results, while FB saves and shooting and BT shooting had negative effects. Increased attempts of FB during offense regardless of success rate had positive influences on winning and negative effects on match results for goalkeepers regardless of an increased number of saves.
Logistic regression analysis had 84.5% accuracy. 6m and Wing goal, 9m success rate, FB shooting, GK Wing save rate, and GK 9m save rate increased the probability of winning. However, Wing shooting, GK 6m shooting, and GK FB save lowered the probability of winning. Wing players required high success rate and fast transition into defense to prevent FB for goalkeepers for an increased probability of winning matches.

Acknowledgements

This work was supported by the Pukyong National University Research Fund in 2020 (CD20201547)

References

1. Bilge, M. (2012). Game analysis of Olympic, World and European Championships in men’s handball. Journal of Human Kinetics, 35(1), 109-118. DOI: 10.2478/v10078-012-0084-7.
2. Fernandez, J., Mendez-Villanueva, A., & Pluim, B. M. (2006). Intensity of tennis match play. British journal of sports medicine, 40(5), 387-391. DOI: 10.1136/bjsm.2005.023168.
3. Ferrari, W., Dias, G., Sousa, T., Sarmento, H., & Vaz, V. (2020). Comparative analysis of the offensive effectiveness in winner and losing handball teams. Frontiers in Psychology, 11, 2566. DOI: 10.3389/fpsyg.2020.547110.
4. Ferrari, W., Vaz, V., and Valente-dos-Santos, J. (2014). Offensive process analysis in handball: identification of game actions that differentiate winning from losing teams. Am. J. Sports Sci. 2, 92-96. DOI: 10.11648/j.ajss.20140204.14.
5. Fujii, K. (2021). Data-driven Analysis for Understanding Team Sports Behaviors. Journal of Robotics and Mechatronics, 33(3), 505-514. DOI: 10.20965/jrm.2021.p0505.
6. Hong, J. H., & Park. J. H. (2016). Analysis Athletic Performance Assessment Factors and Importance Based on Handball Players' Position. The Korean Society of Sports Science, 25(4), 1443-1454.
7. International Handball Federation (2016). Rules of the Game.
8. International Handball Federation (2019). Guidelines and Interpretations of the IHF Rules of the Game.
9. IHF Education Center(2021). https://ihfeducation.ihf.info.
10. Jung, H. K. (2006). Analysis of Games for Performance Evaluation in Men's Handball. Korea Coaching Development Center, 8(1), 125-132.
11. Kim, H. (2012). Estimating the Determinants of Victory and Defeat through Analyzing Records of Korea Handball Game. Korea Institute of Sport Science, 23(2), 244-253. DOI: 10.24985/kjss.2012.23.2.244.
12. Kim, H., Kang, S. J., Park, J. H., & Kim, H. J. (2008). The Factor of Victory and Defeat through Analyzing the Data of the Pro-basketball. Korean Society For Measurement And Evaluation In Physical Education And Sports Science, 10(1), 1-12. DOI: 10.21797/ksme.2008.10.1.001.
13. Kim, H., Kim, H. J., & Park, J. H. (2011). Development of Model to Evaluate Handball Shooting Ability: Weight Elicitation of Shooting Positions. Korean Society For Measurement And Evaluation In Physical Education And Sports Science, 13(3), 77-87. DOI: 10.21797/ksme.2011.13.3.007.
14. Kim, H. C. (2021). A Study of Influencing Factors on World Handball Win-Loss using the Decision Tree Analysis. Journal of Digital Convergence. 19(5) 461-468.
15. Kim, H. J., Park, J. R., Park, J. H., & Cho, E. H. (2013). Evaluation of Handball Performance Based on Quantitative Index. The Korean Journal of Measurement and Evaluation in Physical Education and Sport Science, 15(1), 1-12. DOI: 10.21797/ksme.2013.15.1.001.
16. Luteberget, L. (2018). Physical demands in elite female team handball: Analyses of high intensity events in match and training data via inertial measurement units. Dissertation from the Norwegian School of Sport Sciences.
17. Pokrajac, B. (2010). Analysis, discussion, comparison, tendencies in modern handball. EHF Web Periodical 2010.
18. Pfeiffer, M., & Perl, J. (2006). Analysis of tactical structures in team handball by means of artificial neural networks. International Journal of Computer Science in Sport, 5(1), 4-14.
19. Sevim, Y., & Bilge M. (2007). The comparison of the last Olympic, World and European Men Handball Championships and the current developments in World Handball. Res Yearbook, 13(1), 70-76.
20. Srhoj, V., Rogulj, N., & Katić, R. (2001). Influence of the attack end conduction on match result in handball. Collegium antropologicum, 25(2), 611-617.
21. Taborsky, F. (2011). Competitive loading in top team handball, EHF Web Periodical.
22. Rogulj, N. (2000). Differences in situation-related indicators of the handball game in relation to the achieved competitive results of teams at 1999 world championship in Egypt. Kinesiology, 32, 63-74.
TOOLS
METRICS Graph View
  • 0 Crossref
  •  0 Scopus
  • 395 View
  • 13 Download
Related articles


ABOUT
BROWSE ARTICLES
EDITORIAL POLICY
FOR CONTRIBUTORS
Editorial Office
Korea Institute of Sport Science, 727 Hwarang-lo, Nowon-gu, Seoul 01794, Korea
Tel: +82-2-970-9570    Fax: +82-2-970-9651    E-mail: publ@kspo.or.kr                

Copyright © 2022 by Korea Institute of Sport Science.

Developed in M2PI

Close layer
prev next