A Study on the Winning and Losing Factors of Para Ice Hockey Using Data Mining-Based Decision Tree Analysis

Abstract

This study aims to explore the winning and losing factors of Para Ice Hockey (PIH) games using a data mining-based decision tree analysis targeting the PIH games in international competitions organized by World Para Ice Hockey (WPIH). To achieve the study purpose and data collection, 66 games among the nations participating in the four international competitions for the last two years organized by WPIH, were selected as study subjects. The 3432 game records provided by WPIH were collected as study variables. The results of this study are as follows: First, the winning teams recorded 5.79 goals, 24.05 total shots on goal (SOG), 57.79% face-off winning percentage, 7.62 total saves (SVS), 0.24 total power play goals (PPGs), and 0.39 penalty-killing goals (PKGs) per game were recorded. The losing teams recorded 0.89 goals, 8.52 SOG, 42.21% face-off winning percentage, 18.26 SVS, 0.82 PPG, and 0.05 PKG and showed significant differences. When looking at game records by period, significant differences were revealed in the goals, SOG, and PPG except in the third period, and total shorthanded goals (SHGs) and SVS except in the second period. The winning teams showed the highest goals and SOG in the first, second, and third periods, while the losing teams showed the reverse order. Second, (1) according to the exploration result of winning and losing factors, excluding total goal-related variables, the probability of winning was 82.8% if the goal was 1 or more in the first period. The critical winning and losing factors were revealed as a goal in the first period and SOG in the second period in that order. (2) According to the exploration result of the winning and losing factors, excluding each period’s goal-related variables, the probability of winning was 81.4% if the SOG was 5 or more in the first period. As the essential winning and losing factors, SOG in the first period, SOG in the second period, and time on power play (TPP) in the third period were revealed. (3) The exploration result of winning and losing factors, excluding goals and shots-related variables, the probability of winning was 70.9% if the total face-off winning percentage was 46.23% or more. As the essential winning and losing factors, the total face-off winning percentage and SVS in the second period are shown in order.

1. Introduction

Para ice hockey (PIH) is a representative game of the Winter Paralympic Games in which lower limb-disabled athletes participate. The same rules of general ice hockey are applied, and six players, including a goalkeeper, comprise one team. Riding a double-bladed sled, players use two sticks with a pick at one end of the stick, propelling the sled on ice in the PIH [1,2].

Significant changes have been happening in the PIH games in international athletic performance trends [3,4,5]. In line with such trends, nations fostering PIH must closely analyze nations with excellent competitiveness for strategies to enhance their own countries PIH competitiveness and to expand PIH games. Moreover, those nations need information about the PIH games’ athletic performance requirement for training, tactics, and athlete fostering.

Above all, it is vital to use game records in sports to check the details related to athletic performance and winning and losing [6]. The most crucial information in sports determines winning and losing. Although predicting winning and losing factors is not easy, many experts in the sports field are continuously performing studies to explore the factors determining the dynamics of winning and losing [7].

In the disabled people’s sports area, the scientific sports approach is carried out in diverse fields based on the London 2012 Paralympic Games. An analytic study using game records is emphasized to offer information that improves athletic performance [8,9,10,11,12]. A game record analysis on high-level international competitions helps analyze athletic performance and relevant information as data related to nations with excellent athletic performance [10]; therefore, its utilization is high. However, in the field of sports, there is a lack of research on para ice hockey, and it is difficult to find research on game records from an analytical point of view.

To strengthen international competitiveness in international competitions and to develop the sport of para ice hockey, it is necessary to conduct an objective analysis that can confirm the trend of victory and defeat factors using data from recent major international competitions. As para ice hockey is a team sport, various game-related variables such as shots, passes, face-offs, penalties, and tactical situations occur. So the analysis data on the game contents can be used as major information for improving performance.

Consequently, an objective analysis by which the factors of winning and losing PIH games can be ascertained as essential to foster excellent players of the PIH and achieve good winning records. In 2018, WPIH started systematically providing data on game records in international competitions [13]. As information that can be used in an authentic setting is required, a study that can offer information on the factors affecting winning and losing should be conducted based on game records in international competitions.

In previous studies conducted in the field of sports, the discriminant analysis, logistic regression analysis, artificial neural network analysis, and decision tree analysis have explored variables significantly affecting winning and losing and athletic performance based on game records in the sports field [14,15,16,17].

The decision tree analysis is one of the most widely used and practical methods for inductive inference over supervised data. Based on various attributes a decision tree represents a procedure that classifies the categorical data. The decision tree analysis is a statistical method predicting the group with optimal relevance as a small group or classifying them by making a decision tree chart. The decision tree analysis can indicate relevance and patterns between variables as the combination of predictable rules [18,19].

For constructing decision trees, no parameter setting or domain knowledge is required. Thus, decision trees are sufficient as well as appropriate for exploratory knowledge discovery and their representation of acquired knowledge in tree form is intuitive and easy to understand. Through a tree structure model and result offering, the analysis method can visualize decision rules, and an interpretation becomes easy. Moreover, as a non-parametric model, assumptions such as normality, equal variance, and linearity are not required, and interaction effects functioning as dependent variables can be explored which can be an advantage.

Due to these advantages, decision trees have recently been used to explore the winning and losing factors of games using decision tree analysis in various sports [20,21,22].

Therefore, this study attempts to explore the winning and losing factors affecting the para ice hockey game using decision tree analysis. The results of this study are meaningful in providing information necessary to organize the training and tactics of the PIH teams. It is also able to contribute to improving the performance of the PIH event.

2. Methods

2.1. Data Collection Process

This study aimed to explore the following: First, difference verification by major game factors depending on winning and losing was conducted. Second, game factors affecting the determination and classification of discerning winning and losing were explored by applying the data mining-based decision tree analysis.

For data collection, 66 nations participating in the four international competitions for the last two years, organized by WPIH, were selected as study subjects. The 3432 game records provided by WPIH (https://www.paralympic.org, accessed on 5 November 2022.) were collected and used as study variables.

Table 1. Research subjects and variables.

Subject	Event	Teams
Ostrava 2021 World Para Ice Hockey World Championships	20	USA	CAN
		RUS	KOR
		NOR	ITA
		CZE	SVK
Ostersund 2021 World Para Ice Hockey Championships	15	CHN	FIN
		GER	JPN
		POL	SWE
Berlin 2021 World Para Ice Hockey Paralympic Winter Games Qualification Tournament	15	GER	ITA
		JPN	NOR
		SVK	SWE
Beijing 2022 Paralympic Winter Games	16	USA	CAN
		KOR	CZE
		ITA	CHN
		SVK	RUS

shows the study subjects and variables used in this study.

2.2. Data Analysis Process

The statistical significance level for data processing and analysis was set to 0.05, and the IBM SPSS 24.0 ver. was used. First, descriptive statistics of data were carried out. Each game record’s mean and standard deviation (SD) was calculated for descriptive statistics. To verify differences in game records depending on winning and losing groups in the PIH games, the Mann–Whitney U test, a non-parametric statistical technique, was performed. This was applied because the game records of some study variables did not meet the basic assumptions of parametric statistics.

Second, a decision tree analysis was applied to explore game records affecting winning and losing. A classification and regression tree (CART) was applied for the algorithm to apply the decision tree analysis. CART is a decision tree method that determines and classifies the relationship between dependent and independent variables. The CART algorithm is practical if the dependent variables act as category data. The result of CART is a classification tree or a regression tree, depending on the type of independent variable [19]. If the explanation and prediction power is the highest in the relationship between dependent and independent (prediction) variables, the decision tree is classified as optimal separation and child nodes are selected if estimated errors are the lowest. The CART algorithm shows the highest utilization in the decision tree analysis. Figure 1 is the process structure of the CART algorithm.

For the precise application method of the CART algorithm, this study tried to produce optimal separation standards by setting improvement levels and decreasing variance to 0.001. For a simple model of decision-making, a three-step depth of the tree, five-step parent node, and three-step child node were applied.

The standard error, risk estimate, and classification accuracy index were calculated to ascertain the adequacy of the decision model. To rank the main game factors classifying a winning group and a losing group, the normalized importance (NI) index was calculated. The NI index indicates the relative importance of variables affecting the relationship between target and prediction variables among the total input variables in the decision tree analysis. The prediction variable chiefly affecting outcomes was set to 100, and the NI index indicates relative influence compared with the standard [19].

In applying the decision tree analysis, this study selected three clear goals. First, winning and losing prize-determining factors were explored, except the total goals generated in each game. Second, the winning and losing factors were explored by removing total goals and goals in each period. Third, an analysis was performed by excluding total goals, period goals, and shot goals. These are the known variables directly affecting a game’s winning and losing. Therefore, those variables were excluded from prediction variables to draw detailed results in analyzing game factors that influence winning and losing in PIH games. Finally, records connected to direct winning and losing by step were excluded, and this study tried to explore winning and losing, along with prize-winning determinants based on game detail variables as

Table 2. Research subjects and variables.

Variable	Variable Description
Goal	Goals
SVS	Saves
SOG	Shots on goal
PN	Penalty number
PIM	Penalties in seconds
TPP	Time on power play
PPG	Power play goals
SHG	Shorthanded goals
FO%	Total Face-offs won as percentage

.

3. Results

This study was carried out to explore the factors affecting, determining, and classifying winning and losing in PIH games through a game record analysis of the four international competitions for the last two years organized by WPIH. To achieve the study purpose, this study performed the following: first, verification of differences by game variables depending on winning and losing; second, exploration of winning and losing factors depending on game variables through the data mining-based decision tree analysis. The study results are as follows.

3.1. Descriptive Statistics of PIH Game Records

The first study result is descriptive statistics of PIH game records, and the result is shown in

Table 3. Descriptive statistics of competitions record.

Period	Variable	Mean	SD	Period	Variable	Mean	SD
Total	Total Goal	3.34	3.85	1 period	1p Goal	1.18	1.80
	Total SOG	16.28	12.49		1p SOG	5.52	5.24
	Total FO%	50.00	14.54		1p TPP	85.04	86.24
	Total TPP	285.33	176.03		1p PIM	2.47	3.92
	Total PIM	7.60	6.65		1p PN	1.09	1.01
	Total PPG	0.53	0.80		1p PPG	0.16	0.39
	Total SHG	0.22	0.68		1p SHG	0.11	0.38
	Total SVS	12.94	9.45		1p SVS	4.23	3.98
2 period	2p Goal	1.05	1.41	3 period	3p Goal	1.03	1.48
	2p SOG	5.63	4.72		3p SOG	5.82	10.10
	2p TPP	109.37	108.92		3p TPP	91.02	102.15
	2p PIM	2.73	3.63		3p PIM	2.37	3.69
	2p PN	1.20	1.11		3p PN	1.01	1.14
	2p PPG	0.18	0.54		3p PPG	0.17	0.42
	2p SHG	0.04	0.23		3p SHG	0.10	0.49
	2p SVS	4.58	3.76		3p SVS	4.06	3.33

. An average of 3.34 goals were scored in one game, and the mean goal shots were 16.28. A mean of 12.94 saves was recorded, and the face-off winning percentage was 50.0% on average. Regarding game records per period, 1.18 goals, 1.05 goals, and 1.03 goals were recorded in the first, second, and third periods, respectively, and relatively more goals were recorded in the first period. A total of 5.52, 5.63, and 5.82 shots were recorded in the first, second, and third periods, respectively; consequently, the highest shots on goal were recorded in the third period. Regarding saves, 4.23, 4.58, and 4.06 saves were recorded in the first, second, and third periods, respectively. With all these points, the shots on goal and saves showed the highest value in the third period.

3.2. Differences in Verification of Game Records Depending on Winning and Losing

The second study result is a difference in verification of game records depending on winning and losing, and

Table 4. Differences verification of game records depending on winning and losing.

Variable	Match Win Teams		Match Lose Teams		Mann–Whitney U	Z	Sig
	Mean	SD	Mean	SD
Total Goal	5.79	3.96	0.89	1.45	290.50	−8.72	0.000
Total SVS	7.62	5.49	18.26	9.62	673.50	−6.85	0.000
Total SOG	24.05	12.30	8.52	6.39	494.00	−7.67	0.000
Total PIM	7.76	7.33	7.44	5.94	2170.50	−0.03	0.972
Total TPP	271.98	164.04	298.67	187.55	1991.50	−0.85	0.395
Total PPG	0.82	0.93	0.24	0.50	1368.00	−4.27	0.000
Total SHG	0.39	0.89	0.05	0.27	1779.50	−3.20	0.001
Total FO%	57.79	12.31	42.21	12.30	784.50	−6.34	0.000
1p Goal	2.15	2.08	0.21	0.54	643.50	−7.56	0.000
1p SOG	8.41	5.61	2.62	2.62	711.00	−6.71	0.000
1p PIM	2.48	4.79	2.45	2.85	1927.00	−1.21	0.227
1p TPP	91.56	87.72	78.53	84.90	1981.00	−0.92	0.357
1p PPG	0.27	0.48	0.05	0.21	1714.50	−3.39	0.001
1p SHG	0.17	0.45	0.05	0.27	1949.50	−2.17	0.030
1p PN	0.91	0.82	1.27	1.14	1827.00	−1.69	0.091
1p SVS	2.41	2.42	6.05	4.41	1027.50	−5.27	0.000
2p Goal	1.82	1.56	0.27	0.60	730.50	−7.09	0.000
2p SOG	8.35	4.72	2.91	2.74	654.00	−6.96	0.000
2p PIM	2.89	4.33	2.58	2.80	2120.00	−0.28	0.783
2p TPP	97.93	107.32	120.80	110.12	1955.00	−1.04	0.299
2p PPG	0.32	0.71	0.05	0.21	1807.50	−2.90	0.004
2p SHG	0.06	0.30	0.02	0.12	2111.50	−1.02	0.308
2p PN	1.23	1.09	1.17	1.13	2092.50	−0.41	0.684
2p SVS	2.67	2.38	6.48	3.92	877.50	−5.95	0.000
3p Goal	1.65	1.76	0.41	0.74	1055.50	−5.50	0.000
3p SOG	7.18	4.41	4.45	13.50	861.00	−6.02	0.000
3p PIM	2.59	4.12	2.15	3.22	2003.00	−0.84	0.399
3p TPP	84.77	94.09	97.27	109.99	2098.50	−0.37	0.708
3p PPG	0.21	0.45	0.14	0.39	2015.50	−1.17	0.244
3p SHG	0.18	0.68	0.02	0.12	1946.50	−2.41	0.016
3p PN	1.09	1.06	0.92	1.22	1882.50	−1.42	0.154
3p SVS	2.52	2.08	5.61	3.63	978.50	−5.49	0.000

shows that. Significant differences were shown in total goals, total saves (SVS), total shots on goal (SOG), total power play goals (PPG), total shorthanded goals (SHG), and total face-off % (FO winning percentage) in the total game records. When looking at the records displaying significant differences, winning teams recorded 5.79 goals and 24.05 SOG on average in one game. However, losing teams recorded an average of 0.89 goals and 8.52 SOG in one game. As for the face-off winning percentage, the winning teams recorded 57.79%, but the losing teams recorded 42.21%. The winning teams showed 7.62 SVS on average, and the losing teams recorded 18.26 SVS in one game. The winning teams displayed 0.82 goals, and the losing teams displayed 0.24 goals on average in the power play situation (PPS), where the same team players were. The goal showed a significant difference in the penalty-killing situation in which the same team players were lacking; specifically, the winning teams scored 0.39 goals, while the losing teams scored 0.05 goals.

Regarding game records per period, significant differences in SHG and SVS were revealed, excluding goals, along with SOG in the third period and PPG in the second period. The winning teams displayed 2.15, 1.82, and 1.65 goals in the first, second, and third periods, respectively, but the losing team showed 0.21, 0.27, and 0.41 goals in the first, second, and third periods, respectively. The winning teams scored the highest goals in the first, second, and third periods, respectively, but the losing teams scored in the reverse order. Regarding SOG, the winning teams registered 8.41, 8.35, and 7.18 in the first, second, and third periods, respectively. However, the losing teams showed 2.62, 2.91, and 4.45 shots on goal. Concerning SVS, the winning teams recorded 2.41, 2.67, and 2.52 SVS in the first, second, and third periods, respectively; the losing teams recorded 6.05, 6.48, and 5.61 SVS in the first, second, and third periods, respectively. Overall, the winning teams intensively recorded SOG and goals in the first period, while the losing teams recorded SOG and goals in the third period.

3.3. Decision Tree Analysis According to Winning and Losing

3.3.1. Winning and Losing Factors Exploration Result, Excluding Total Goals-Related Variables

The third study’s result explores winning and losing factors using game records. Concerning the first study result of the exploration of winning and losing factors, this study explored the factors affecting winning and losing, excluding total goal-related variables occurring in the games (total goals, total PPG, and total SHG). The analysis result is shown in Figure 2.

When looking at the result, the first factor classifying winning and losing in PIH games was goals in the first period. If the goal was 1 or more in the first period, the probability of winning was 82.8%, and if the goal was 1 or less in the first period, the probability of losing was 80.9%. If the goal was 1 or more in the first period and 3 or more of SOG in the second period, the probability of winning increased to 89.5%. When the goal was 1 or less in the first period and 2 or more in the second period, the probability of winning was 83.3%. Meanwhile, if the goal was 1 or less in the first period and 2 or less in the second period, the probability of losing was 87.1%. The classification accuracy of the result was 87.1%. As a result of the importance analysis among the winning and losing factors, excluding goal-related variables, scoring goals in the first period was the most crucial factor determining winning and losing, followed by SOG in the second period.

3.3.2. Winning and Losing Factors Exploration Result, Excluding Goal-Related Variables in Each Period

The winning and losing factor analysis result, excluding goal-related variables in each period (goal in the first, second, and third periods, PPG in the first, second, and third periods, and SHG in the first, second, and third periods) is shown in Figure 3. The probability of winning was 81.4% if SOG was 5 or more in the first period. Meanwhile, the probability of winning if SOG in the first period was 5 or less. The result’s classification accuracy was 87.1%, and the standard error was 0.029. According to the analysis of winning and losing factors excluding goal-related variables in the total period, maintaining SOG above a certain level in the first period was a factor that increased the probability of winning, followed by SOG in the second period.

3.3.3. Winning and Losing Factors Exploration Result, Excluding Goal and Shot-Related Variables

Figure 4 shows the analysis results of the winning and losing factors, excluding goal and shot-related variables. When looking at the results, if the entire face-off winning percentage was 46.23% or more, the probability of winning was 70.9%. Meanwhile, if the face-off winning percentage was 46.23% or less, the probability of winning was 18.9%, The result’s classification accuracy was 79.5%, and the standard error was 0.035. According to the vital analysis of winning and losing factors, excluding goal and shot-related variables, the total face-off winning percentage was the most critical factor determining winning and losing, followed by maintaining a certain level or less of SVS in the second period.

4. Discussions and Conclusions

This study explored PIH games’ winning and losing factors using a data mining-based decision tree analysis targeting the PIH games in international competitions organized by WPIH.

The 66 games of the nations participating in four international competitions for the last two years were selected as study subjects, and game records were collected to achieve the study purpose. The descriptive statistics, Mann–Whitney U test, and data mining-based decision tree analysis were performed for data processing. This study analyzed the following upon the decision tree analysis: (1) winning and losing factors, excluding total goals occurring in each game; (2) winning and losing factors, excluding each period’s goal-related factors; and (3) winning and losing factors, excluding total goals and shot-related factors. The statistical significance level was set to 1, and the IBM SPSS 24.0 version was used. The total decision tree according to para ice hockey’s winning and losing factor analysis result is shown in Figure 5.

The conclusions drawn through this study are as follows: first, the result of differences in the verification of game records depending on winning and losing, total goals, total saves (SVS), total shots on goal (SOG), total power play goals (PPG), total shorthanded goals (SHG), and total FO winning percentage (FO%) showed significant differences among the total game records. Regarding winning teams, 5.79 goals, 24.05 SOG, 57.79% face-off winning percentage, 7.62 SVS, 0.24 PPG, and 0.39 penalty-killing goals (PKGs) on average per game were recorded. Concerning the losing teams, 0.89 goals, 8.52 SOG, 42.21% face-off winning percentage, 18.26 SVS, 0.82 PPG, and 0.05 PKG were recorded and showed significant differences. When looking at game records by period, significant differences were revealed in the goals, SOG, and PPG except in the third period, and SHG and SVS except in the second period. The winning teams showed the highest goals and SOG in the first, second, and third periods, respectively, and the losing teams showed the reverse order.

Regarding the goal in the penalty-killing situation in which the same team players lacked, the winning and losing teams recorded 0.39 and 0.05, respectively, and significant differences were shown. This results from the winning teams’ average scoring being high, with goal-scoring power in unfavorable situations being high, not to mention favorable situations.

The winning teams recorded more goals and SOG in the order of the first, second, and third periods, but the losing teams showed the reverse order. From the result, the winning teams scored the most SOG and goals in the first period. The losing teams recorded the most SOG and goals in the third period. When looking at total periods, the SOG and SVS, excluding goals, showed higher values in the third period. Through this, it was judged that the probability of winning became higher if shots and scoring were attempted intensively in the early part of a game, regarding game operation tactics. As for the face-off winning percentage, the winning teams showed 57.79%, but the losing teams showed 42.21%. As the mean face-off winning percentage was 50.0%, it was judged that the probability of winning can be elevated if the face-off winning percentage was secured over a certain level. Concerning SVS, the winning teams recorded 7.62 SVS, and the losing teams showed 18.26 SVS on average per game. Because the mean SVS per game was 12.94, it was judged that the probability of winning can be elevated by lowering the saving rate and reducing opportunities to shoot.

As a result of this analysis, there is a tendency to lead the pace of an advantageous game only when goals are scored in the early stages of the game. One of the reasons for this result is that in the case of general ice hockey, the game time is 20 min per period, whereas, in para ice hockey, the game time is 15 min, which is 5 min less than general ice hockey [1]. Therefore, it is judged that while the characteristics of sports such as the fast game operation speed and high physical consumption of general ice hockey are similar, the difference of 5 min in the game time can have a significant effect on determining the win or loss in the early part of the game.

In addition, due to the nature of the para ice hockey event, all complex movements such as propulsion, shooting, dribbling, and passing only by one stick must be played using the upper body while riding on a sled. As a result, the physical burden on the players inevitably increases towards the second half of the para ice hockey game [23,24,25]. Therefore, scoring in the early stages of a game may have an impact on a team’s victory. The reason why the defeated team scored high in the second half may be that the winning team’s offense did not decline, but rather that it strengthened its organizational power over defense. Moreover, this game pattern is seen as a factor that can increase the win rate considering the physical limit in the head-to-head win rate. Moreover, the fact that the winning team’s SVS is low and the losing team’s high means that the winning team has many offensive chances, so having many scoring opportunities in the early stages of the game can be an important factor in the flow of the game.

The results of this study are similar to the win-and-lose factor analyses of previous Handball and water polo sports studies [20,21,22,26].

Second, this study analyzed factors by excluding the winning and losing-related in the decision tree analysis to explore game records affecting winning and losing. (1) As a result of the winning and losing factor analysis, excluding total goals, the probabilities of winning were 82.8% and 80.9%, respectively, in case the goal was 0.5 or more in the first period. The results of this study are similar to the win-and-lose factor analyses of previous soccer studies [27].

Above all, the higher SOG in the 2nd period than in the 1st period seems be influenced by strategic strategies by identifying patterns of the opposing team’s game operation in the early stages of the game and preparing for them. As mentioned above, SOG at the beginning of the game is a factor that determines the flow of the game. In particular, as in the results excluding variables related to each period goal, the increase in the probability of winning when the SOG is 4.5 or higher can be attributed to the psychological factor of the players’ superior management of the game flow.

As a result of the analysis, excluding total goal-related variables, goals in the first period and SOG in the second period were the factors for winning. Based on the analysis result, the probability of winning can be enhanced if the goal is scored, and an attack is attempted in the first period through aggressive tactical operation at the beginning of the first and second periods.

(2) As a result of the winning and losing factor analysis, excluding each period’s goal-related variables, if SOG was 4.5 or more in the first period, the probability of winning was 81.4%. The results of this study are similar to the win-and-lose factor analyses of previous volleyball studies [28]. Above all, having a lot of shots on goal (SOG) means that they can be linked to scores, which can lead the team to victory.

However, when it was 4.5 or less, the probability of losing was 75.3%. As a result of the analysis, excluding goal-related variables in total periods, the SOG (≤5) in the first period, SOG (≤4) in the second period, and TPP (≤101) in the third period were essential factors for winning. Based on the analysis result, it was revealed that the probability of winning became higher if shot attempts were more from the beginning of a game to elevate scoring probability. Shot attempts should be made beyond a certain level. If players aim to shoot accurately towards the goal post so they can be recognized as SOG, the probability of winning is judged to go up. In the results of a study by Hyrinen et al. (2011), a study of match analysis in para ice hockey, the number of shots was generally higher for the losing team, but the scoring effect was lower for the losing team. This supports the findings of this study by reporting that the winning team’s good defense is explained by game management that forces the losing team to shoot from a worse position and rush to score at the end of the game [29].

(3) For the winning and losing factor analysis result, excluding goals and shot-related variables, the probability of winning was 70.9% when the face-off winning percentage was 46.23% or more. However, if the face-off winning percentage was 46.23% or less, the probability of losing was 81.1%. When the total face-off winning percentage was 46.23% or more, and SVS in the second period was 3.5 or less, the probability of winning was 82.7%. This can be seen in the form of team play, which acquires the right to attack through face-off and connects it directly to the goal. In particular, a face-off in the opponent’s camp can have a chance to connect with the goal by making a planned play. When the face-off winning percentage was 62.02% or more, although SVS was 3.5 or more in the second period, the probability of winning was 100%. This is seen as being connected to the goal through the team’s planned play.

According to the analysis result, excluding goal and shot-related variables, the total face-off winning percentage was the most critical factor determining winning and losing, followed by SVS in the second period.

It was confirmed that the probability of winning went up if the face-off winning percentage was higher. If the face-off winning percentage was lower, the probability of losing became higher. If players endeavored to elevate the face-off winning percentage in every moment in each period, it is judged that the probability of winning can be increased. This conclusion is consistent with the study results of factors contributing to winning in ice hockey and supports this study. In addition, the previous para ice hockey game analysis study reported that the winning team occupied more pucks [29].

The factors directly affecting winning and losing are excluded by step in this study, and an effort was made to ascertain the factors affecting winning and losing. Synthesizing all these factors, goals and SOG in the first period were drawn as important factors to increase the probability of winning in the PIH games. Goals should be secured at the beginning of a game, and it was analyzed to improve the probability of winning by increasing SOG to the fullest from the beginning of a game. From a game operation aspect, it was analyzed that the probability of winning decreased as attack-related records increased in the latter part of a game. Goals and SOG are the factors related to direct attack power. The factor with the most decisive influence among game factors was the face-off winning percentage. The probability of winning is considered to be elevated if the success rate in the face-off is enhanced in case scoring is not made and when tactics to dominate from the beginning of a game are operated. Although it is impossible to compare these results with previous studies directly, most of them were similar to the winning determinants of team sports. PIH is a team sport, and various game variables occur, including shots, passes, penalties, face-offs, and diverse tactics. Therefore, an analysis of game records can be used as vital information to improve thematic performance. Competition in the PIH is fierce, and few nations participate in the PIH games. Analysis, continuous and systematic data collection, and accumulation centered on recent major competitions can be the key.

The study results provided objective information, including objective game factors, through which some factors can determine winning and losing of PIH games and what probability can affect winning and losing and at what level of the game these factors can be grasped. This study is meaningful because crucial information can contribute to the PIH game’s athletic performance improvement, and expertise reinforcement is offered. Studies on PIH games are currently insufficient compared with studies on other games.

Consequently, careful judgment is required, as limitations exist in interpretation through a comparison of previous studies with the results of this study. Previously, several analyses of ice hockey-related studies have also been conducted in the sports field. However, most studies have aimed to analyze the injuries and game patterns of ice hockey studies [23,24,25,30,31,32]. Although the data applied in this study collected the latest competition result of PIH, the data mining-based analysis method was applied due to the limited number of data. In future follow-up studies, various analysis methods need to be applied. If new analysis methods such as deep neural network analysis are applied in subsequent studies, higher accuracy can identify variables and levels that affect winning and losing. Moreover, now is the time for invigoration-relevant studies for developing PIH, and base expansion is urgent. As understanding PIH games’ characteristics is enhanced, analytical studies that can provide more delicate information are required. Attempts at various types of analysis are necessary, and a diversified approach related to athletic performance, including factors determining winning a prize in major international competitions, should be conducted in addition to winning and losing factors.