An Evaluation Research on Dynamic Hit Stop Using Eye Gaze

Rena Tomizawa

and Tomokazu Ishikawa

1,2 a

Toyo University, 1-7-11 Akabanedai Kita-ku, Tokyo, Japan

Prometech CG Research, 3-34-3 Hongo bunkyo-ku, Tokyo, Japan

Keywords:

Hit Stop, Eye Gaze, Game Design.

Abstract:

The purpose of this study was to verify whether the response changes when hit stop, one of the elements

of GameFeel, is changed according to gaze information. First, several participants play the action game by

changing the hit stop duration and answer a questionnaire. Based on the results of this questionnaire, the

boundary between pleasant and unpleasant hit stop duration is determined based on the idea of discriminant

analysis method. Next, within a comfortable hit stop duration range, we examined whether the response is

different when the hit stop duration is changed according to the gazing duration. As a result, designing hit stop

duration that corresponds to staring duration is important to improve GameFeel.

1 INTRODUCTION

In recent years, the topic of indie games and game

development has often been discussed at the research

level. At the same time, attention is being paid to

the mechanics of the game. There has been a lot of

talk about the “GameFeel” element as a mechanism

in the game. GameFeel is the “sense of feeling” that

players receive during gameplay. Games are multi-

sensory experiences, but the narrative content, mu-

sic, art, and many other aspects of the game inﬂu-

ence the feel of the game. GameFeel places more em-

phasis on the role played by interactivity. This paper

will focus on the design of player interaction with the

game, with reference to the survey by Pichlmair and

Johansen (Pichlmair and Johansen, 2022).

In explaining GameFeel, it is necessary to de-

scribe “juice”. Juice is the excessive amount of feed-

back in relation to user input to enhance interactivity.

The purpose of juice is to make people feel that the

players’ actions have meaning and that game players

can predict the outcome. A game that is good texture,

properly staged and lively is sometimes called a juicy

game.

A research on GameFeel concerns how players’

minds and bodies experience emotions when playing

games. How to design for the emotional aspects of

the play experience should be studied not only in the

ﬁeld of games, but also in relation to design theory,

https://orcid.org/0000-0002-9176-1336

psychology, ergonomics, philosophy, and many other

ﬁelds. It is believed that elucidating the elements of

GameFeel will help us to understand what variables

are involved in enhancing the immersive experience

of a game, and will broaden the scope of expression

in game development. Pichlmair and Johansen classi-

ﬁed the components of GameFeel into the following

ﬁve categories; “movement and actions”, “event sig-

niﬁcation”, “time manipulation”, “persistence” and

“scene framing”in Figure 1 (Pichlmair and Johansen,

2022). These elements are further subdivided into 31

items. When designers intentionally elicit emotion,

“hit stop” is often used from among these elements.

The purpose of this study was to verify whether

the response changes when hit stop, one of the ele-

ments of GameFeel, is changed according to gaze in-

formation. Hit stop is a type of visual feedback in

which the animation displays a pause or slow-motion

effect at the moment of impact (attack, being hit,

landing depending on the falling altitude). In previ-

ous researches related to hit stop, it has been discov-

ered that the pseudo-shock sensation is increased by

vibrating the remote control simultaneously with hit

stop (Hachisu et al., 2011) or by changing the dura-

tion of hit stop according to body velocity (Ban and

Ujitoko, 2021).

In this study, an experiment is conducted to test

the hypothesis that if the pseudo-shock sensation can

be increased by combining hit stop in addition to tac-

tile sensation or body velocity, then the combination

between visual information and hit stop may provide

Tomizawa, R. and Ishikawa, T.

An Evaluation Research on Dynamic Hit Stop Using Eye Gaze.

DOI: 10.5220/0012461400003660

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2024) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 151-158

ISBN: 978-989-758-679-8; ISSN: 2184-4321

151

Figure 1: Components of GameFeel.

a change in GameFeel.

• To propose and validate a methodology for de-

termining the boundaries of pleasantness and un-

pleasantness of hit stops

• To conﬁrm that GameFeel is improved by chang-

ing the hit stop duration according to the staring

duration

Since action games are frequently implemented with

the hit-stop direction, which is effective for users, this

study implements and veriﬁes the proposed method

on an action game.

2 RELATED WORKS

Brown and Cairns interviewed game players to deﬁne

immersion based on their experiences and discussed

the quality of immersion (Brown and Cairns, 2004).

Using Grounded Theory, they ﬁrmly categorized im-

mersion into three levels: engagement, engrossment,

and total immersion. This division suggests new lines

of demarcation for investigating immersion and mov-

ing into software domains other than games.

Pichlmair and Johansen classiﬁed GameFeel with

reference to over 200 games (Pichlmair and Johansen,

2022). As a result, three distinct domains of the in-

tended player experience were derived: physicality,

ampliﬁcation, and support. In this paper, it is also

noted that another study must be conducted to deter-

mine when elements of GameFeel are called ”good

game feel” for game players. Therefore, we examine

effective hit stop direction methods when new physi-

cal information is inputted.

Ban and Ujitoko proposed and evaluated the incor-

poration of hit stop effects and vibratory tactile sensa-

tions into VR (virtual reality), which pause the action

at the moment of impact or display slow motion an-

imation on VR tennis, a VR sport (Ban and Ujitoko,

2021). They evaluated the effect of hit stop with and

without vibration and the effect of hit stop in duration

by using the magnitude estimation method to estimate

the number of seconds of hit stop that is comfortable.

They also stated in their paper that they need to con-

ﬁrm whether hit stops are effective in other VR expe-

riences.

Lin et al. analyzed the elements of juicy impact

feedback in action games and found that hit stops,

sound coherence, and camera control have a signiﬁ-

cant impact on the player’s sense of hitting (Lin et al.,

2022). Within this paper, they presented 19 feature

frameworks, and tested them in actual games. It was

suggested that the player’s impact feeling may be

compromised if any one of the three functions, hit

stop, sound coherence, and camera control, is not de-

signed speciﬁcally for the player.

Hachisu et al. investigated the use of pseudo-

haptic feedback effects using vibration and proposed

two new methods for enhancing pseudo-haptic feed-

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

152

0.0s

0.2s

0.4s

0.6s

Figure 2: Difference of animation by hit stop duration: 1, 11, 21, and 31st frames at 24 fps.

back in virtual object exploration (Hachisu et al.,

2011). In this paper, two methods are proposed to ad-

just the cursor change on the screen: the ﬁrst method

uses stripe patterns to enhance the pseudo-haptic tex-

ture, using vibrations to enhance the pseudo-haptic

texture, thereby adjusting the amount of cursor move-

ment; the second method uses visual vibrations to ad-

just the cursor change by simulating the texture of the

virtual object.

In recent years, applications that incorporate eye

gaze have also been well studied (Skaramagkas et al.,

2023). Eye tracking is gaining ground in the re-

search community, but it is not yet a common ap-

proach for detecting emotional and cognitive states or

for incorporating it as an element in games. Charoen-

pit and Ohkura designed and implemented a proto-

type to record learners’ eye movements and inves-

tigate the relationship between two emotions, inter-

est and boredom, and evaluated the experimental re-

sults (Charoenpit and Ohkura, 2015). From this ex-

periment, they found that gaze ﬁxation time and num-

ber of gazes were negatively correlated with boredom

content. If interest and boredom can be detected by

eye gaze, we thought we would evaluate the contribu-

tion to the fun of the game by performing a dynamic

hit stop as an output of eye gaze and behavior in game.

3 PROPOSED METHOD

We create an original action game in which hit stops

are presented in a variable manner depending on the

gazing point and gazing time, and evaluate the effects

of dynamic hit stops based on eye gaze. This study is

conducted in two phases: to determine the borderline

between pleasantness and unpleasantness of hit stops,

and to conﬁrm the effectiveness of the dynamic hit

stops designed based on this borderline. This section

describes the methodology for each phase.

3.1 Phase 1 : Borderline Between

Pleasantness and Unpleasantness of

Hit Stops

Several players are asked to play the game by chang-

ing the hit stop duration in the range of 0.0s to 0.7s

and to answer a questionnaire. An animation of the

game created and with the hit stops changed is shown

in Figure 2. In our implementation, hit-stop duration

refers to the amount of time that an enemy charac-

ter’s vanishing animation is prolonged when a player

attacks an enemy character (Figure 3). From the

perspective of the game context, we experimented

with a total of three scenes, preparing two patterns of

weapons (sword and bow) and a mode in which the

player could freely switch between these weapons.

Figure 4 shows a scene in which a bow is used. So,

participants play up to 24 patterns (with 8 levels of hit

stop duration for each of the 3 scenes). As shown in

Figure 5, we ask the participants to play a reference

task with a hit-stop time of 0 s at the beginning of

each scene.

For the post-play questionnaire, ﬁve items were

selected from the indices of Ban and Ujitoko’s

An Evaluation Research on Dynamic Hit Stop Using Eye Gaze

153

Figure 3: Diagram of hit stop effect. We call T

hitstop

hit

stop duration. The playback speed during T

hitstop

is s times

normal speed, and we set s = 0.01.

Figure 4: A scene where a player is attacking an enemy

using a bow.

Figure 5: Phase 1 task procedure.

study (Ban and Ujitoko, 2021) and the GEQ (Game

Engagement Questionnaire) items (Brockmyer et al.,

2009): “sense of impact”, “sense of presence”, “en-

joyment”, “initiative”, and “stressfulness”. For each

item, the magnitude estimation method is employed

by setting the condition without hit stops as 100 and

asking the respondents to respond with a numerical

value ranging from 0 to 200. From the question-

naire items, assuming that “stressfulness” is an ele-

ment of unpleasantness, items that conﬂict with un-

pleasantness are extracted by calculating correlation

coefﬁcients. Based on the rating values of the pleas-

ant/unpleasant items in the questionnaire, the distri-

bution of which hit stop duration the participants felt

were the best/worst is determined. To obtain the dis-

tribution of unpleasantness, each participant votes for

the hit stop duration at which the participants an-

swered the highest “stressfulness”.

For game design of the next phase, it is necessary

to determine the borderline between pleasant and un-

pleasant. We considered obtaining this borderline as

a discriminant analysis method for the two classes of

pleasantness and unpleasantness. Assuming that each

distribution follows a normal distribution, we set the

boundary between pleasant and unpleasant as the hit-

stop time at which the Mahalanobis distance (Maha-

lanobis, 2018) from each distribution is equal. If the

Mahalanobis distance from the pleasant/unpleasant

distribution at the boundary x

is D

, D

, respec-

tively, we ﬁnd x

such that D

= D

and x

< x

− x

, (1)

− x

, (2)

where x

and x

are the mean, s

and s

are the stan-

dard deviation.

For the next phase, we also obtain the gazing du-

ration for the enemy character for each player during

this experiment. Based on this measurement data, the

average and maximum gazing duration (y and y

max

) of

the player toward the enemy character are obtained.

3.2 Phase 2 : Design of Dynamic Hit

Stop Based on Gazing Duration

One of the novel game effects in this research is to

change the hit stop duration according to the gazing

duration. Since it is likely to be judged as unpleas-

ant if hit stop duration crosses the borderline obtained

by the method in Section 3.1, the hit stop duration

should be varied up to this boundary line. First, we

prepare Function (1) as the reference task. Function

(1) is a constant function characterized by a ﬁxed hit-

stop time of 0.39 seconds, representing the average

duration of a comfortable hit stop. Since there are sev-

eral possible functions that map the gazing duration y

to the hit stop duration x, we prepare four functions

for experiment (see APPENDIX for more information

on trial and error during design). To these patterns of

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

154

Figure 6: Functions of gazing duration and hit stop duration

used in Phase 2. Function (1) is a constant function with

the average value of the pleasant hit stop duration. Func-

tion (2) is a linear function that passes through the origin

and the point (

, y). Function (3) is clipped from function

(2) in the range of x

−



− x



≤ x ≤ x

(outside this

range is a constant value function). Function (4) is a linear

function that passes through the origin and the point (x

max

). Function (5) is a linear function that passes through

two points (x

, y) and (x

, y

max

). Function (6) cannot be

represented graphically because the hit stop duration is ran-

domly determined relative to the gazing duration.

Figure 7: Phase 2 task procedure.

Figure 8: Implementation of a game user interface that

presents the gazing point and gazing duration.

functional change, we add one with a constant hit stop

duration and one with a randomly changing hit stop

duration as shown in Figure 6.

A total of six patterns are played by the experi-

mental collaborators, who are asked to complete the

same questionnaire as in the Section 3.1. As shown

in Figure 7, we ask the participants to play a reference

task at the start of the experiment that uses Function

(1).

We implement an user interface that allows the ex-

perimenter to know the gazing point and gazing dura-

tion while playing the game. The implemented game-

play screen is shown in Figure 8. Gauges near each

enemy character increase with the amount of time

spent gazing at them.

4 EXPERIMENTS AND RESULTS

VIVE Pro Eye is used to measure the player’s line of

sight and reﬂect it in the game in real time. We de-

velop a game for experiment on Unity. FEEL (More-

Mountains, 2015) is used as an asset, and hit stops

are introduced to the action game. We use a PC

with CPU: 11th Gen Intel® Core™ i9-11900 CPU @

2.50GHz, RAM: 16GB and GPU : NVIDIA GeForce

RTX 3060 Ti 8GB for our experiments.

As shown in Figure 9, participants are equipped

with VR headsets and enter a room reproduced in the

VR space to play the game. Inside this VR room, a

television is placed, and participants play the game

through this television. This setup allows participants

to feel a gaming experience in VR that closely resem-

bled playing a game in the real world. The experi-

mental picture is shown in Figure 10.

Figure 9: Schematic diagram of the VR space that partici-

pants will experience during experiment.

Figure 10: A photograph during the experiment. Partici-

pants in the experiment wear VIVE Pro Eye and play games

to answer the questions. A questionnaire response form was

also created within the VR space.

An Evaluation Research on Dynamic Hit Stop Using Eye Gaze

155

4.1 Results of Phase 1

In the Phase 1 experiment, men and women in their

teens and twenties participated, and data were col-

lected from 16 participants for Scenes 1 and 2 and

15 participants for Scene 3. First, we calculated cor-

relation coefﬁcients for the ﬁve items of the question-

naire that were answered, varying the hit stop dura-

tion for each scene. The results are shown in Fig-

ure 11. In all scenes, “enjoyment” and “stressful-

ness” were found to have a strong negative correla-

tion. Therefore, we employed “enjoyment” as pleas-

ant and “stressfulness” as unpleasant, and calculated

their distributions.

Based on the questionnaires, two distributions

were created by voting one vote for each hit stop du-

ration for which the experimental participant gave the

highest score for “enjoyment” or “stressfulness”. If

there were n tied scores, 1/n voted for each hit stop

duration. The histogram created according to this pro-

cedure is shown in Figure 12. Overall, the longer the

hit stop duration, the more uncomfortable it tended to

be.

Since no normality was observed in the histogram

of unpleasantness in Scene 2, the Mahalanobis dis-

tance was used to obtain the boundary line between

pleasant and unpleasant for Scene 1 and Scene 3.

The averages of pleasant and unpleasant for Scene

1 x

, x

were 0.37 and 0.56 seconds, respectively,

and the variances s

, s

were 0.03704 and 0.05163.

The means x

and x

were 0.39 and 0.57 seconds, re-

spectively, and the variances s

, s

were 0.03782 and

0.04214 of Scene 3. From these results, the border-

lines between pleasant and unpleasant hit stop dura-

tion x

obtained were 0.537 and 0.55 seconds, respec-

tively.

4.2 Results of Phase 2 and Discussion

We examined whether the response differs when the

hit stopping duration is changed according to the gaz-

ing duration within the range of comfortable hit stop

duration determined in Phase 1. Scene 3, which has a

longer range of comfortable hit stop duration, was se-

lected as the experimental scene for Phase 2. The re-

sults of Phase 2 questionnaire are shown in Figure 13.

Since the scores of the reference when the hit stop

duration is constant are 100, it was found that the re-

search purpose, whether the sense of impact improves

according to the gazing duration, can be achieved by

methods other than function (4). In addition, as for

enjoyment, games with dynamically changing hit stop

duration was rated higher than those with ﬁxed hit

stop duration. Since function (5) had the narrowest

(a) Scene 1 (using only sword)

(b) Scene 2 (using only bow)

Figure 11: Correlation matrix of the rating values of the

survey items in each scene. The axis labels, A to E, cor-

respond to “sense of impact”, “sense of presence”, “enjoy-

ment”, “initiative”, and “stressfulness” in order.

value range among the functions (2) to (6) designed

in this study, it can be seen from the enjoyment and

initiative items that the hit stop duration was compa-

rable to the constant condition.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

156

Scene 1 Scene 2 Scene 3

Figure 12: Histograms of pleasantness (top) and unpleasantness (bottom) in each scene.

The following is a discussion of the impact sen-

sation, which was the objective of this study. Func-

tions (2) and (4) are considered to have less impact

sensation than the other functions because the hit stop

duration becomes 0s when the gazing duration is 0s.

We consider that function (3) differs from function

(2) in that it includes a bottom value, which results

in a stronger sense of impact. Since the respondents

felt a stronger sense of impact with function (3) than

with random (6), we conclude that the hit stop effect,

which varies with the gazing duration, is highly effec-

tive in increasing the sense of impact.

5 CONCLUSIONS AND FUTURE

WORK

In this study, the boundary between pleasant and un-

pleasant hit stop duration was determined by discrim-

inant analysis method based on the questionnaire of

the experimental participants. As result of this ex-

periment, we found that the context in the game (in

this case, the type of weapon) caused differences in

the time perceived as unpleasant. Next, we conducted

an experiment in which we varied the hit stop dura-

tion according to the gazing duration up to the hit stop

duration that felt comfortable. It was concluded that

changing the hit stop duration according to the gazing

duration can enhance the impact sensation.

As a future work, the distribution and bounds of

pleasant and unpleasant hit stop duration for different

types of weapons in action games need to be studied

in more detail. Our study suggests that humans might

react nervously to animations of consequences, espe-

cially when the action is aimed well, as in the case

of bows. In this experiment, we counted up the gaz-

ing duration only when the gazing point was inside

the enemy character, but we plan to study how the re-

sults change when this calculation is applied to the

area around the gazing point.

ACKNOWLEDGEMENTS

This work was supported by Toyo University Top Pri-

ority Research Program.

REFERENCES

Ban, Y. and Ujitoko, Y. (2021). Hit-stop in VR: Combina-

tion of pseudo-haptics and vibration enhances impact

sensation. In 2021 IEEE World Haptics Conference

(WHC), pages 991–996.

Brockmyer, J. H., Fox, C. M., Curtiss, K. A., McBroom, E.,

Burkhart, K. M., and Pidruzny, J. N. (2009). The de-

velopment of the game engagement questionnaire: A

measure of engagement in video game-playing. Jour-

nal of Experimental Social Psychology, 45(4):624–

634.

Brown, E. and Cairns, P. (2004). A grounded investigation

of game immersion. In CHI ’04 Extended Abstracts on

Human Factors in Computing Systems, CHI EA ’04,

page 1297–1300, New York, NY, USA. Association

for Computing Machinery.

Charoenpit, S. and Ohkura, M. (2015). Exploring emotion

in an e-learning system using eye tracking. Interna-

tional Journal of Affective Engineering, 14(4):309–

316.

Hachisu, T., Cirio, G., Marchal, M., L

ecuyer, A., and Kaji-

moto, H. (2011). Pseudo-haptic feedback augmented

An Evaluation Research on Dynamic Hit Stop Using Eye Gaze

157

(a) “sense of impact”

(b) “sense of presence”

(d) “initiative”

(e) “stressfulness”

Figure 13: Score distribution of Phase 2 questionnaire re-

sults. 100 is the reference score when the hit stop duration

is ﬁxed.

with visual and tactile vibrations. In 2011 IEEE In-

ternational Symposium on VR Innovation, pages 327–

328.

Lin, Z., Duan, H., Wen, Z. A., and Cai, W. (2022). What

features inﬂuence impact feel? A study of impact

feedback in action games.

Mahalanobis, P. C. (2018). On the generalized distance in

statistics. Sankhy

a: The Indian Journal of Statistics,

Series A (2008-), 80:S1–S7.

MoreMountains (2015). FEEL. https://feel.moremountains.

com/ (2023.11.17. accessed).

Pichlmair, M. and Johansen, M. (2022). Designing game

feel: A survey. IEEE Transactions on Games,

14(2):138–152.

Skaramagkas, V., Giannakakis, G., Ktistakis, E., Manousos,

D., Karatzanis, I., Tachos, N. S., Tripoliti, E., Marias,

K., Fotiadis, D. I., and Tsiknakis, M. (2023). Review

of eye tracking metrics involved in emotional and cog-

nitive processes. IEEE Reviews in Biomedical Engi-

neering, 16:260–277.

APPENDIX

In designing the functions of gazing duration and hit

stop duration, we considered using cumulative distri-

bution function (CDF) with a pleasant normal dis-

tribution. We expected to draw an S-shaped curve

that could correspond to a pleasant hit stop duration

around the mean gazing duration. However, when we

plotted the CDF of the normal distribution of pleas-

antness obtained from the questionnaire, we found

that a linear approximation was not problematic ( Fig-

ure 14). Therefore, in Phase 2, all functions were de-

signed as linear functions. Function (3) was designed

to linearly approximate the graph of this CDF.

Figure 14: CDF in the range of x

−



− x



≤ x ≤ x

with pleasure as the normal distribution function.

GRAPP 2024 - 19th International Conference on Computer Graphics Theory and Applications

158