scispace - formally typeset
Search or ask a question
Posted ContentDOI

A Case Study of First Person Aiming at Low Latency for Esports.

TL;DR: In this paper, the authors provide a case study with data demonstrating the importance of local system latency improvement, even at latency values below 20 ms, for aiming in first person shooter (FPS) games.
Abstract: Lower computer system input-to-output latency substantially reduces many task completion times. In fact, literature shows that reduction in targeting task completion time from decreased latency often exceeds the decrease in latency alone. However, for aiming in first person shooter (FPS) games, some prior work has demonstrated diminishing returns below 40 ms of local input-to-output computer system latency. In this paper, we review this prior art and provide an additional case study with data demonstrating the importance of local system latency improvement, even at latency values below 20 ms. Though other factors may determine victory in a particular esports challenge, ensuring balanced local computer latency among competitors is essential to fair competition.

Summary (3 min read)

1 INTRODUCTION

  • In the world of esports every possible advantage is sought out in order to beat the competition.
  • Market trends and online message boards suggest that competitive gamers seek the highest refresh rates and lowest latency, constantly pursuing every minor advantage they might provide.
  • The authors investigate the impact of local system latency on FPS aiming performance, with a particular focus on the low end of latency that most interests competitive FPS gamers.
  • This is the author’s version of the work.

1.1 Local System Latency

  • In computing systems, there are many types of latency that contribute to the total time it takes for a user’s actions to produce output from the system.
  • A commonly discussed latency in esports and online games is the network latency.
  • Network latency is important because it describes how long it takes for local actions to be received at the server, and subsequently, how long it then takes to deliver those actions to each of the connected players.
  • The authors measured system latency as the time for a mouse click to cause a pixel change in the display, and report the average of many measurements in each condition.
  • Note that while the authors report the average for their latency values, the actual measured latencies occur over a range surrounding that average, and furthermore, other latencies (such as mouse sensor input, or audio output) were not measured and are assumed to be different by some relatively fixed offset from the measured values.

2 BACKGROUND

  • A number of prior studies consider pointing and aiming tasks in the context of a computer system’s added latency.
  • It has been observed that, when the latency differences inherent to frame rate changes are controlled for, latency reduction offers a more significant aiming task performance benefit than frame rate [7].
  • The authors collected data from Ivkovic [3], Cattan [1], Jota [4], Teather [9], andMacKenzie [6] in an effort to visualize how latency may change aiming task completion time, similar to how target size and distance are known to affect task.

2.5 bits except for the Jota data, which has an index of difficulty of 1.58 bits.

  • Performance via a combined metric known as "index of difficulty" (ID) [2].
  • These surveys are interesting in that they cover a diverse set of interfacing modalities.
  • Ivkovic [3] also looks at mouse-based interaction, but for non-pointer FPS aiming tasks, similar to the focus of their study.
  • In Figure 2, you can see results selected from these studies with a similar ID to make it more reasonable to compare them on the same plot.
  • More study around these low latency levels would certainly augment this limited understanding.

3 COMPARING 12 AND 20 MS

  • Published research, including that found in Figure 2, has often struggled to study latency below 20ms.
  • In developing this study, the authors found that the lowest system latency they were able to achieve consistently on their test systems was 12 ms.
  • Through further testing injecting additional latency, the authors found that by adding 8 ms of latency on average, the measured latency distributions only slight overlap (see Fig. 3).

As expected, the two peaks are mostly separated, though there is some overlap in the middle.

  • Latency condition for their experiment, and 20 ms as a condition sufficiently low, yet high enough to show distinct results from the 12 ms condition.
  • In order to get the minimum average latency to 12 ms, the authors selected appropriate hardware and software tools.
  • Subjects begin the task by centering their aim direction indicator, or reticle, on a reference target then press shift on the keyboard to destroy this target.
  • Following a missed shot a weapon cooldown penalty was assessed, not allowing the user to fire for another 0.5 seconds.
  • The median measured completion time for this task across all 3200 trials.

1.5-2.2s is likely caused by the 0.5s weapon cooldown and may loosely represent the second shot clustering.

  • These medians are shownwith the corresponding standard error metric in Figure 5.
  • Consistent with prior art the difference in median task completion time (182 ms) far exceeds the reduction of latency (8 ms).
  • Note that though there is a significant difference in mean and median between the two distributions, the mode values do not differ as substantially.
  • This is common for such heavy-tailed task completion time distributions wherein the tail weight (or shape) tends to account for much of the difference in average task completion time.
  • FPS aiming requires detecting a change or target, determining where the target is, planning for perceived target motion, moving the fingers, hand, arm and wrist to position the aiming reticle over the target, and finally clicking the mouse button.

4 INDIVIDUAL RESULTS

  • The remaining five users saw no statistically significant difference in this experiment.
  • Furthermore, it is worth noting that though the authors find statistical significance for the claim of benefits from minor latency reduction in this study, high sample count is necessary for reaching significance for this claim.
  • This is due to the significantly reduced sample size when considering user count (8) to trial count (3200 or 400 per user).
  • These variations could be found across all demographics, or they could be correlated with skill level, and additional work is needed to identify likely causes of these differences.
  • If tournament organizers endeavor may be tempted to artificially increase latency to balance across users, which would be similar to adding handicaps in traditional sports, and is not generally recommended.

5 CONCLUSION

  • Based on previous publications and the data the authors present here, they conclude that reducing the latency of a computer system is beneficial to FPS aiming performance, even at low latency levels.
  • One option for reducing the impacts of varied latency on tournament outcomes would be establishing more rigorous latency standards for hardware, enforcing that all competitors are made aware of their system latency and how different hardware components impact that latency.
  • Alternatively identical systems could be provided by tournament hosts, guaranteeing a level playing field between competitors, at the expense of allowing competitors to choose their own hardware.
  • The authors recognize that players and teams may have reasons to make decisions that trade latency for some other benefit.
  • A player may greatly prefer a particular mouse grip for example, and a loss of latency may be seen as less important than a mouse supporting this grip.

Did you find this useful? Give us your feedback

Figures (7)

Content maybe subject to copyright    Report

A Case Study of First Person Aiming at Low Latency for Esports
Josef Spjut
jspjut@nvidia.com
NVIDIA
USA
Ben Boudaoud
bboudaoud@nvidia.com
NVIDIA
USA
Joohwan Kim
sckim@nvidia.com
NVIDIA
USA
ABSTRACT
Lower computer system input-to-output latency substantially re-
duces many task completion times. In fact, literature shows that
reduction in targeting task completion time from decreased latency
often exceeds the decrease in latency alone. However, for aiming
in rst person shooter (FPS) games, some prior work has demon-
strated diminishing returns below 40 ms of local input-to-output
computer system latency. In this paper, we review this prior art
and provide an additional case study with data demonstrating the
importance of local system latency improvement, even at latency
values below 20 ms. Though other factors may determine victory
in a particular esports challenge, ensuring balanced local computer
latency among competitors is essential to fair competition.
CCS CONCEPTS
Human-centered computing Pointing devices
;
User mod-
els; User studies; Applied computing Computer games.
KEYWORDS
pointing devices, mouse, rst person targeting, rst person games
ACM Reference Format:
Josef Spjut, Ben Boudaoud, and Joohwan Kim. 2021. A Case Study of First
Person Aiming at Low Latency for Esports. In EHPHCI 2021: ACM CHI
Workshop on Esports and High Performance Human Computer Interaction,
May 08–13, 2021, Yokohama, Japan. ACM, New York, NY, USA, 4 pages.
https://doi.org/10.31219/osf.io/nu9p3
1 INTRODUCTION
In the world of esports every possible advantage is sought out
in order to beat the competition. Competitive gamers play games
to win, and get most of their enjoyment from doing well. First
Person Shooter (FPS) games like Counter-Strike, Overwatch and
Valorant allow players to compete at many skill levels in online
ladders to see who will come out on top. Market trends and online
message boards suggest that competitive gamers seek the highest
refresh rates and lowest latency, constantly pursuing every minor
advantage they might provide. In this study, we investigate the
impact of local system latency on FPS aiming performance, with
a particular focus on the low end of latency that most interests
competitive FPS gamers.
EHPHCI 2021, May 08–13, 2021, Yokohama, Japan
© 2021 Association for Computing Machinery.
This is the author’s version of the work. It is posted here for your personal use. Not
for redistribution. The denitive Version of Record was published in EHPHCI 2021:
ACM CHI Workshop on Esports and High Performance Human Computer Interaction,
May 08–13, 2021, Yokohama, Japan, https://doi.org/10.31219/osf.io/nu9p3.
Local
System
Display
Input Device
Network Gateway
Game Server
Local Latency
Network Latency
Network
Figure 1: Diagram outlining the dierence between local la-
tency, the focus of this work, and network latency, the more
commonly referred to term historically. Note that many in-
game events are handled at the local latency, without game
server intervention.
1.1 Local System Latency
In computing systems, there are many types of latency that con-
tribute to the total time it takes for a user’s actions to produce
output from the system. A commonly discussed latency in esports
and online games is the network latency. Network latency is im-
portant because it describes how long it takes for local actions to
be received at the server, and subsequently, how long it then takes
to deliver those actions to each of the connected players. However,
in this work, we focus on the (local) system latency, that is the
time from a user’s input until the result of that input is delivered
by a computer. In this study, we measured system latency as the
time for a mouse click to cause a pixel change in the display, and
report the average of many measurements in each condition. Note
that while we report the average for our latency values, the actual
measured latencies occur over a range surrounding that average,
and furthermore, other latencies (such as mouse sensor input, or
audio output) were not measured and are assumed to be dierent
by some relatively xed oset from the measured values.
2 BACKGROUND
A number of prior studies consider pointing and aiming tasks in the
context of a computer system’s added latency. It has been observed
that, when the latency dierences inherent to frame rate changes
are controlled for, latency reduction oers a more signicant aiming
task performance benet than frame rate [
7
]. We collected data from
Ivkovic [
3
], Cattan [
1
], Jota [
4
], Teather [
9
], and MacKenzie [
6
] in an
eort to visualize how latency may change aiming task completion
time, similar to how target size and distance are known to aect task

EHPHCI 2021, May 08–13, 2021, Yokohama, Japan Spjut et al.
0 50 100 150 200
System Latency (ms)
0.6
0.7
0.8
0.9
1.0
1.1
Task Completion Time (s)
Ivkovic 2015
Cattan 2015
Jota 2013
Teather 2009
MacKenzie 1993
Figure 2: The eect of local system latency on completion
time for simple pointing tasks from previous publications.
Trial data was selected for index of diculty between 2 and
2.5 bits except for the Jota data, which has an index of di-
culty of 1.58 bits.
performance via a combined metric known as "index of diculty"
(ID) [2]. These surveys are interesting in that they cover a diverse
set of interfacing modalities. Jota [
4
] and Cattan [
1
] look at touch-
based, dragging interfaces over dierent ranges of latency. Teather
[
9
] and MacKenzie [
6
] both look at pointer-based mouse interfaces,
but MacKenzie considers traditional 2D UI pointing tasks, while
Teather focuses on completing 3D visualized spatial tasks. Ivkovic
[
3
] also looks at mouse-based interaction, but for non-pointer FPS
aiming tasks, similar to the focus of our study. We include FPS
aiming alongside other aiming tasks since Looser [
5
] shows that
FPS aiming follows the same Fitts’ law relationship as other pointing
tasks. In Figure 2, you can see results selected from these studies
with a similar ID to make it more reasonable to compare them on
the same plot.
The apparent trend in this related work is that task comple-
tion time increases (gets worse) as latency increases. In all studies
shown, this trend shows continued improvements all the way to-
ward 0 latency. The Ivkovic [
3
] study is interesting, as it is the
only one using a FPS aiming task similar to our work, and shows
diminishing returns from latency reduction below about 50 ms.
However, this work’s lowest-latency data point represents a change
in the V-SYNC setting, which may be a confounding factor. More
study around these low latency levels would certainly augment this
limited understanding.
3 COMPARING 12 AND 20 MS
Published research, including that found in Figure 2, has often
struggled to study latency below 20 ms. For this reason, we endeavor
to study exclusively 20 ms and below in this work. In developing
this study, we found that the lowest system latency we were able
to achieve consistently on our test systems was 12 ms. Through
further testing injecting additional latency, we found that by adding
8 ms of latency on average, the measured latency distributions only
slight overlap (see Fig. 3). Thus we selected 12 ms as the lowest
0 4 8 12 16 20 24 28
C2P Latency (ms)
0
100
200
300
400
500
600
700
Count
Figure 3: Click to photon distribution from a selection of
clicks in our study across both 12 ms and 20 ms conditions.
As expected, the two peaks are mostly separated, though
there is some overlap in the middle.
latency condition for our experiment, and 20 ms as a condition
suciently low, yet high enough to show distinct results from the
12 ms condition.
In order to get the minimum average latency to 12 ms, we se-
lected appropriate hardware and software tools. G-SYNC monitors
with 240 Hz refresh rate and fast pixel response time were a key
component of this low latency system. We also use Logitech G203
wired mice, which were measured to consistently have relatively
low click-to-USB packet latency. The computer hardware included
Intel Core i7-9700K CPUs and NVIDIA RTX 2080 Ti GPUs. Finally,
we selected the First Person Science (FPSci) [
8
] platform we de-
signed previously, giving us full control of the inner loop of the
application and allowing us to optimize this loop for minimal input-
to-visual latency. FPSci was created to support customized rst
person aiming experiments such as the one we designed for this
work. This work uses FPSci release v20.07.01
1
.
1
Experiment conguration at https://github.com/jspjutNV/latencyExpEHPHCI21.
Figure 4: Annotated in-app user view just after the destruc-
tion of a target showing the dummy target (in white).

A Case Study of First Person Aiming at Low Latency for Esports EHPHCI 2021, May 08–13, 2021, Yokohama, Japan
12 ms 20 ms
System Latency Condition
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
Task completion time (s)
Figure 5: Median aiming task completion time and standard
error metric for 3200 trials at 12 ms and 20 ms of system
latency.
Our experiment consisted of a simple repeated FPS aiming task
evaluated based on completion time. Subjects begin the task by cen-
tering their aim direction indicator, or reticle, on a (white) reference
target (as shown in Figure 4) then press shift on the keyboard to de-
stroy this target. Once the reference target is destroyed, a test target
appears at a randomized position within a world space bounding
box, moving along a linear path with a constant velocity. Periodi-
cally (once every 0.8-1.5 seconds) this test target changes direction
and velocity. If the test target leaves its world space bounding box
its velocity is reected about the bounds producing a "bounce" be-
havior and keeping the target in a desired range of distance and
visual angle. Subjects attempt to align their reticle with the target
and click, destroying the target. If the subject clicks without the
reticle over the target a miss is registered and the trial continues.
Following a missed shot a weapon cooldown penalty was assessed,
not allowing the user to re for another 0.5 seconds. This penalty
was included to mimic various game weapon cooldown periods for
similar styles of weapons, and has the eect of amplifying missed
shots in terms of task completion time. Comparing a 0.5 second
cooldown to no cooldown is outside the scope of this work.
We distributed our experiment to 8 users, each of which com-
pleted the aiming task 400 times at 12 ms latency and another 400
times at 20 ms latency (with 8 ms of added delay). These 800 trials
were spread across 8 dierent sessions after a 20 trial warm up
session was completed. During the warm up session each player
could customize and select their preferred mouse sensitivity setting.
Each of the 100 trial sessions, including the warm up, were broken
up into blocks of 10 trials. The players were given brief breaks of at
least 1 second after each block, and allowed a longer break between
sessions if they desired.
We choose to analyze our results primarily using median task
completion time, as the median is often more representative of a
typical performance of a subject and robust to individual extreme
values than mean. The underlying completion time distributions,
with mean and median values plotted is provided in Figure 6. The
median measured completion time for this task across all 3200 trials
0 1 2 3 4 5
Completion Time (s)
0
50
100
150
200
250
Count
12ms
20ms
12 ms mean
12 ms median
20 ms mean
20 ms median
Figure 6: A histogram showing the completion time distri-
bution for the 12 and 20 ms latency conditions. Means are
indicated by the dashed vertical lines and medians by the al-
ternating dotted/dashed lines. Note that the plateau between
1.5-2.2s is likely caused by the 0.5s weapon cooldown and
may loosely represent the second shot clustering.
is lower for 12 ms of latency (1.348 s) than for 20 ms of latency
(1.530 s). These medians are shown with the corresponding standard
error metric in Figure 5. Consistent with prior art the dierence in
median task completion time (182 ms) far exceeds the reduction of
latency (8 ms). A Wilcoxon signed rank test (over all subject data)
shows that the medians are signicantly dierent (p-value < 0.005).
We use Cohen’s D to examine eect size for the dierence in mean
completion time between the 12 and 20 ms latency conditions as
well. The overall Cohen’s D value for this dierence of means is
0.319 indicating a small-to-medium eect.
We present the distribution of task completion time across all
subjects by latency condition in Figure 6. Note that though there is
a signicant dierence in mean and median between the two distri-
butions, the mode values do not dier as substantially. In addition,
the dierence between mean and median for the distributions at
12 and 20 ms remains fairly consistent. This is common for such
heavy-tailed task completion time distributions wherein the tail
weight (or shape) tends to account for much of the dierence in
average task completion time.
A keen observer will notice that the task completion times in
our experiment are higher than the results for the simple pointing
tasks from the prior research publications. This is because the
task we asked our users to perform was more dicult. FPS aiming
requires detecting a change or target, determining where the target
is, planning for perceived target motion, moving the ngers, hand,
arm and wrist to position the aiming reticle over the target, and
nally clicking the mouse button. This sequence of actions can
be completed quite rapidly for those who are experienced and
plan well, but when the target has complicated shapes and motion
characteristics, players may need to continually update their aim to
click on the target. The process of updating the aim while a target
is in motion may be a part of the underlying reason for the increase
in task completion time beyond the dierence in system latency.

EHPHCI 2021, May 08–13, 2021, Yokohama, Japan Spjut et al.
0
1
2
3
User 1 Median User 2 Median User 3 Median User 4 Median
12 ms 20 ms
0
1
2
3
User 5 Median
12 ms 20 ms
User 6 Median
12 ms 20 ms
User 7 Median
12 ms 20 ms
User 8 Median
Figure 7: Per-user median aiming task completion time and
standard error metric for 400 trials each at 12 ms and 20 ms
of system latency. Users whose results did not reach statisti-
cal signicance are plotted in the lighter shade.
4 INDIVIDUAL RESULTS
While the overall data is interesting on its own, it is also interest-
ing to consider the results of individual users. Figure 7 plots these
individual results. Three users saw statistically signicant improve-
ments at 12 ms of local latency compared to 20 ms. However, the
remaining ve users saw no statistically signicant dierence in
this experiment. This suggests that there may be an individual sen-
sitivity to impacts of latency. This sensitivity may be a result of any
number of factors including skill level, ambient light conditions,
previous practice or experience at low latency, and other factors.
Further study is needed to investigate which, if any, of the many
possible causes for this individual variation is to blame.
Furthermore, it is worth noting that though we nd statistical
signicance for the claim of benets from minor latency reduction
in this study, high sample count is necessary for reaching signif-
icance for this claim. An alternative analysis approach wherein
per-user medians are compared between latency conditions found
no statistically signicant dierence between the means of these
sets of per-user median task completion times. This is due to the
signicantly reduced sample size when considering user count (8)
to trial count (3200 or 400 per user).
We are reluctant to draw too many strong conclusions about
individual variation given the small pool of users, and the lack of
additional user classication data presented here. These variations
could be found across all demographics, or they could be correlated
with skill level, and additional work is needed to identify likely
causes of these dierences. If tournament organizers endeavor may
be tempted to articially increase latency to balance across users,
which would be similar to adding handicaps in traditional sports,
and is not generally recommended.
5 CONCLUSION
Based on previous publications and the data we present here, we
conclude that reducing the latency of a computer system is bene-
cial to FPS aiming performance, even at low latency levels. Certainly
based on response time alone, we believe that the most competitive
of players should try to have the lowest latency possible. However,
beyond this minimal advantage, latency does seem to oer benets
to users even in the sub-20 ms range, with our subjects demonstrat-
ing 182 ms of median task time improvement at an 8 ms change in
latency.
While in many cases latency will not be the determining fac-
tor in which player "wins" an interaction, having competitors use
computer systems with the same latency behavior removes this
important variable from distorting the outcome of a match. One
option for reducing the impacts of varied latency on tournament
outcomes would be establishing more rigorous latency standards
for hardware, enforcing that all competitors are made aware of
their system latency and how dierent hardware components im-
pact that latency. Alternatively identical systems could be provided
by tournament hosts, guaranteeing a level playing eld between
competitors, at the expense of allowing competitors to choose their
own (peripheral) hardware. By better leveling the playing eld of
system latency in esports we hope to strengthen its competitive
integrity. We strongly encourage esports players, teams, and tour-
nament organizers to monitor and control latency both in training
and competition.
We recognize that players and teams may have reasons to make
decisions that trade latency for some other benet. A player may
greatly prefer a particular mouse grip for example, and a loss of
latency may be seen as less important than a mouse supporting this
grip. A team may sign a sponsorship deal that includes hardware
restrictions. In these situations, we believe the decision makers
should be aware of the aiming performance cost of increased latency
and include it in their cost-benet analysis.
REFERENCES
[1]
Elie Cattan, Amélie Rochet-Capellan, Pascal Perrier, and François Bérard. 2015. Re-
ducing Latency with a Continuous Prediction: Eects on Users’ Performance
in Direct-Touch Target Acquisitions. In Proceedings of the 2015 International
Conference on Interactive Tabletops and Surfaces (Madeira, Portugal) (ITS ’15).
Association for Computing Machinery, New York, NY, USA, 205–214. https:
//doi.org/10.1145/2817721.2817736
[2]
Paul M Fitts and James R Peterson. 1964. Information capacity of discrete motor
responses. Journal of experimental psychology 67, 2 (1964), 103.
[3]
Zenja Ivkovic, Ian Stavness, Carl Gutwin, and Steven Sutclie. 2015. Quantifying
and Mitigating the Negative Eects of Local Latencies on Aiming in 3D Shooter
Games. Association for Computing Machinery, New York, NY, USA, 135–144.
https://doi.org/10.1145/2702123.2702432
[4]
Ricardo Jota, Albert Ng, Paul Dietz, and Daniel Wigdor. 2013. How Fast is Fast
Enough? A Study of the Eects of Latency in Direct-Touch Pointing Tasks. Association
for Computing Machinery, New York, NY, USA, 2291–2300. https://doi.org/10.
1145/2470654.2481317
[5]
Julian Looser, Andy Cockburn, and Joshua Savage. 2005. On the validity of using
First-Person Shooters for Fitts’ law studies. People and Computers XIX 2 (2005),
33–36.
[6]
I. Scott MacKenzie and Colin Ware. 1993. Lag as a Determinant of Human Per-
formance in Interactive Systems. In Proceedings of the INTERACT ’93 and CHI ’93
Conference on Human Factors in Computing Systems (Amsterdam, The Netherlands)
(CHI ’93). Association for Computing Machinery, New York, NY, USA, 488–493.
https://doi.org/10.1145/169059.169431
[7]
Josef Spjut, Ben Boudaoud, Kamran Binaee, Jonghyun Kim, Alexander Majercik,
Morgan McGuire, David Luebke, and Joohwan Kim. 2019. Latency of 30 ms
Benets First Person Targeting Tasks More Than Refresh Rate Above 60 Hz. In
SIGGRAPH Asia 2019 Technical Briefs. ACM, Brisbane, Australia, 110–113.
[8]
Josef Spjut, Ben Boudaoud, Kamran Binaee, Alexander Majercik, Morgan McGuire,
and Joohwan Kim. 2019. FirstPersonScience: Quantifying Psychophysics for First
Person Shooter Tasks. In UCI Esports Conference. UCI, Irvine, CA, 7.
[9]
R. J. Teather, A. Pavlovych, W. Stuerzlinger, and I. S. MacKenzie. 2009. Eects
of tracking technology, latency, and spatial jitter on object movement. In 2009
IEEE Symposium on 3D User Interfaces. IEEE, Lafayette, LA, USA, 43–50. https:
//doi.org/10.1109/3DUI.2009.4811204
Citations
More filters
Journal ArticleDOI
06 Oct 2021
TL;DR: In this paper, a 3D video game was developed and coupled with the prediction of an ANN to compensate for latency in a live first-person action game and the results showed that players achieved significantly higher scores, substantially more hits per shot and associate the game significantly stronger with a positive affect when supported by the ANN.
Abstract: Cloud gaming services and remote play offer a wide range of advantages but can inherent a considerable delay between input and action also known as latency. Previous work indicates that deep learning algorithms such as artificial neural networks (ANN) are able to compensate for latency. As high latency in video games significantly reduces player performance and game experience, this work investigates if latency can be compensated using ANNs within a live first-person action game. We developed a 3D video game and coupled it with the prediction of an ANN. We trained our network on data of 24 participants who played the game in a first study. We evaluated our system in a second user study with 96 participants. To simulate latency in cloud game streaming services, we added 180 ms latency to the game by buffering user inputs. In the study we predicted latency values of 60 ms, 120 ms and 180 ms. Our results show that players achieve significantly higher scores, substantially more hits per shot and associate the game significantly stronger with a positive affect when supported by our ANN. This work illustrates that high latency systems, such as game streaming services, benefit from utilizing a predictive system.

7 citations

Proceedings ArticleDOI
27 Jul 2022
TL;DR: FPSci, a tool for controlled user studies in FPS gaming, achieves a level of granularity of control not offered by other solutions by allowing finer grained parametric control of the environment together with frame-wise logging of player state and performance metrics.
Abstract: First-person shooters (FPS) games are dominant in the competitive gaming and esports community. However, relatively few tools are available for experimenters interested in studying mechanics of these games in a controlled, repeatable environment. While other researchers have made progress with one-off applications as well as custom content and mods for existing games, we are not aware of a general purpose application for empirically studying a broad set of user interactions in the FPS context. For the past few years our team has developed, maintained, and deployed First Person Science (FPSci), a tool for controlled user studies in FPS gaming. FPSci experimenters configure their desired base environment, as well as conditions and user preferences using a simplified JSON-esque set of input configurations, and results are stored in an SQLite database. By allowing finer grained parametric control of the environment together with frame-wise logging of player state and performance metrics, we achieve a level of granularity of control not offered by other solutions. FPSci is available as an open source project 1 under a CC BY-NC-SA 4.0 license.

4 citations

Proceedings ArticleDOI
05 Aug 2021
TL;DR: In this article, the authors provide an interactive demonstration, playable in a web browser, that shows how much latency limits aiming performance, and how late warp can help, and demonstrate the usefulness of late warp as a potential solution to FPS latency.
Abstract: Latency can make all the difference in competitive online games. Late warp is a class of techniques used in VR that can reduce latency in FPS games as well. Prior work has demonstrated these techniques can recover most of the player performance lost to computer or network latency. Inspired by work demonstrating the usefulness of late warp as a potential solution to FPS latency, we provide an interactive demonstration, playable in a web browser, that shows how much latency limits aiming performance, and how late warp can help.

2 citations

References
More filters
Proceedings ArticleDOI
01 May 1993
TL;DR: A model according to which lag should have a multiplicative effect on Fitts' index of difficulty is proposed, which accounts for 94% of the variance and is better than alternative models which propose only an additive effect for lag.
Abstract: The sources of lag (the delay between input action and output response) and its effects on human performance are discussed. We measured the effects in a study of target acquisition using the classic Fitts' law paradigm with the addition of four lag conditions. At the highest lag tested (225 ms), movement times and error rates increased by 64% and 214% respectively, compared to the zero lag condition. We propose a model according to which lag should have a multiplicative effect on Fitts' index of difficulty. The model accounts for 94% of the variance and is better than alternative models which propose only an additive effect for lag. The implications for the design of virtual reality systems are discussed.

442 citations

Proceedings ArticleDOI
14 Mar 2009
TL;DR: It is indicated that latency has a much stronger effect on human performance than low amounts of spatial jitter, and large, uncharacterized jitter “spikes” significantly impact 3D performance.
Abstract: We investigate the effects of input device latency and spatial jitter on 2D pointing tasks and 3D object movement tasks. First, we characterize jitter and latency in a 3D tracking device and an optical mouse used as a baseline comparison. We then present an experiment based on ISO 9241-9, which measures performance characteristics of pointing devices. We artificially introduce latency and jitter to the mouse and compared the results to the 3D tracker. Results indicate that latency has a much stronger effect on human performance than low amounts of spatial jitter. In a second study, we use a subset of conditions from the first to test latency and jitter on 3D object movement. The results indicate that large, uncharacterized jitter “spikes” significantly impact 3D performance.

156 citations

Proceedings ArticleDOI
15 Nov 2015
TL;DR: This study studies the use of a continuous prediction of the touch location as an alternative to the hardware only approach to reduce the latency gap and reveals that the prediction length is strongly constrained by the nature of target acquisition tasks, but that the approach can be successfully applied to counteract a large part of the negative effect of latency on users' performances.
Abstract: Latency in direct-touch systems creates a spatial gap between the finger and the digital object when dragging. This breaks the illusion of presence, and has a negative effect on users' performances in common tasks such as target acquisitions. Latency can be reduced with faster hardware, but reaching imperceptible levels of latency with a hardware-only approach is a difficult challenge and an energy inefficient solution. We studied the use of a continuous prediction of the touch location as an alternative to the hardware only approach to reduce the latency gap. We implemented a low latency touch surface and experimented with a constant speed linear prediction with various system latencies in the range [25ms-75ms]. We ran a user experiment to objectively assess the benefits of the prediction on users' performances in target acquisition tasks. Our study reveals that the prediction length is strongly constrained by the nature of target acquisition tasks, but that the approach can be successfully applied to counteract a large part of the negative effect of latency on users' performances.

30 citations

Proceedings ArticleDOI
17 Nov 2019
TL;DR: It is shown that reduced latency has a clear benefit in task completion time while increased refresh rate has relatively minor effects on performance when the inherent latency reduction present at high refresh rates is removed.
Abstract: In competitive sports, human performance makes the difference between who wins and loses. In some competitive video games (esports), response time is an essential factor of human performance. When the athlete’s equipment (computer, input and output device) responds with lower latency, it provides a measurable advantage. In this study, we isolate latency and refresh rate by artificially increasing latency when operating at high refresh rates. Eight skilled esports athletes then perform gaming-inspired first person targeting tasks under varying conditions of refresh rate and latency, completing the tasks as quickly as possible. We show that reduced latency has a clear benefit in task completion time while increased refresh rate has relatively minor effects on performance when the inherent latency reduction present at high refresh rates is removed. Additionally, for certain tracking tasks, there is a small, but marginally significant effect from high refresh rates alone.

24 citations

Frequently Asked Questions (1)
Q1. What contributions have the authors mentioned in the paper "A case study of first person aiming at low latency for esports" ?

In this paper, the authors review this prior art and provide an additional case study with data demonstrating the importance of local system latency improvement, even at latency values below 20 ms.