You are on page 1of 3

INFOMSCIP A2G15

Rayne Blair Frank Johan van Dijk Jan Willem de Ruig


5237548, 9 hours 0847216, 10 hours 6369502, 9 hours
Pavlos Sakellaridis
0881198, 10 hours
September 2021

1 Puzzle game description squares depending on the level.


The game has difficulty based on the size of the
Car Escape is a puzzle game developed by Puzzle grid, the number of obstructing cars relative to open
Saga Studio(Puzzle Saga Studio, 2018). It is avail- spaces, the number of vertical cars versus horizon-
able on the Google Play store for download on An- tal cars, the number of small cars compared to big
droid devices. The game has positive reviews on the cars, and the position of the goal relative to the red
Google Play app store, at roughly 4.6 stars out of 5 car. Some levels require the player to make counter-
with 13,520 reviews at time of writing. intuitive moves in order to reach the goal. An exam-
The game functions similarly to a traditional sliding ple of this is having to move the red car away from
the goal to free up space.
Figure 1: Example of an easy Car Escape level. Left
image indicates the level’s starting state. The right
indicates the state immediately before being won.
2 Puzzle game features that in-
fluence difficulty
Game features are a necessity of any game and are
able to define the difficulty of each level. When fea-
tures are added to a level they can increase or de-
crease the difficulty.
As described in (Kreveld et al., 2015) game features
can be categorized in three categories. In the game
Car Escape we were able to find features that fit two
puzzle. The goal of the game is to remove the red car of these three categories.
from the grid through the goal by sliding it, and other
cars, across the board. Cars, including the red car,
can only move on the vector that they are on; hori-
zontal cars move horizontally and vertical cars move 2.1 Initial Features
vertically. Obstructing cars can be either 2-squares
long (small cars) or 3-squares long (big cars). The Initial features are features that are clear from the
red car is always a small car. The grid is 5x5 or 6x6 moment the level starts.

1
2.1.1 Level Size always being evident to the player at initialization
(that there is a clear path for the red car to be moved
The size of the level, measured by the number of pos-
off of the grid) it was determined there is no solution
sible positions on the grid. We expect how bigger the
feature.
level size is the more difficult a level becomes.

2.1.2 The number of cars 2.3 Dynamic Features


It is expected that when the number of cars is in- Dynamic features are features that relate to the pro-
creased the level becomes more difficult. It has to be cess of solving a level.
noted that the number of cars possible in one level
has a maximum which deepens on the level size and
the ratio between the small cars and big cars. 2.3.1 Moves

The minimum number of moves required to solve the


2.1.3 The percentage of big cars related to level. A low number of moves will result in an easy
the number of total cars level while a high number of moves is expected to be
more difficult.
Big cars take up more squares on the playing field and
because of that have less free movement on the grid.
This makes them more difficult to maneuver around. 2.3.2 Counter intuitive moves
It is expected when the percentage of free spaces on
the grid relative to obstructed spaces is closer to 0 The number of counter intuitive moves. The moves
percent that the level becomes more difficult. are moves needed to solve the puzzle but are counter
intuitive to the player. This might be a move where
Figure 2: Example of a level with a 6x6 grid and big the red car needs to be moved a spot farther from
cars. the goal, or a move where an obstacle car needs to be
moved towards a spot that is in the way of the red car
and the goal. It is expected that when the number
of these moves is increased the difficulty increases.

3 Non-linear options for a diffi-


culty function
Non-linear features are features that, when their ef-
fect on the difficulty of a level is assessed, the rela-
tionship is represented on a graph as a curve instead
of a line or line segment. What this means for the
purposes of difficulty calculations is simple; increas-
ing or decreasing this specific feature would cause an
abnormality in the curvature of the difficulty of the
game. Determining if any feature would behave that
way requires extensive research data from both sim-
2.2 Solution Features
ulations of in-game levels and play testers. Because
These features are features of the level in its solved of that reason we can only give our guesses about the
configuration. Due to the nature of the solution state non-linearity of any given game feature.

2
3.1 Level Size 4 Acknowledgements and Fur-
Because this is the most basic of the features, it is also ther research
the most difficult to assess in a vacuum. Everything
We have presented a set of features, as well as a sam-
from the number of cars to the amount of moves a
ple function to approximate the difficulty of the game,
player can make is bound by this feature. Further-
based on the methodology proposed by Kreveld et
more, because of the game increasing this feature af-
al(Kreveld et al., 2015). However, this function in its
ter only the first levels of play, it is not (at least by
current state is based purely on theoretical knowledge
the game’s interpretation of difficulty curve) treated
about the game, and can only be used as a general
as a factor for difficulty increase.
indication of what features might weigh heavier than
others.
In order to verify these indications and use the pro-
3.2 Number of cars
posed features to construct an accurate function,
This is also a fundamental feature of the game, so more research and extensive testing would need to
we can argue that the non-linear approach for this be done. In this paper there is a distinct lack of em-
specific aspect of the game is also tricky. One obser- pirical data and user studies, such as what Kreveld
vation that can be made is that as we increase the et al used to verify their expectations and to create
number of cars, we obstruct the ability of the player the function. It is thus proposed that in order to ac-
to predict the future moves that can be made. This tualise a difficulty function for the game Car Escape,
happens mainly because each move ”locks” more po- difficulty data should be gathered in a similar man-
tential moves when the grid is overpopulated with ner.
cars. Additionally, while for the purposes of this paper we
adhered to the game rules as described in section one,
there exist similar games for which the rules may only
3.3 Moves differ slight amounts. While for these games, the
core features influencing difficulty may be the same
The number of moves required to solve the level is the as for Car Escape, nuanced differences may have un-
only feature that we can highly derive linearity from. expected effects on game-play and difficulty. If a dif-
It is most probable that with the correct ”weight” for ficulty function for this ’family’ of games is desired,
the difficulty function that this can be very closely rather than solely a difficulty function for this per-
and correctly measured. However, we suspect that mutation of the game, these nuances should be taken
very high number of moves do not change the dif- into account, and it is not to be assumed the function
ficulty as much as one might think, so it could be proposed here would satisfy all games of this family.
expressed for example as a log(n) where n is the min-
imum number of moves required to solve the level.
References
3.4 Counter intuitive moves Kreveld, M. V., Löffler, M., and Mutser, P. (2015).
Automated puzzle difficulty estimation. 2015
This is also a very complex feature to correctly as- IEEE Conference on Computational Intelligence
sess. However, because of the non-standard corre- and Games (CIG), pages 415–422.
lation that this feature has on difficulty mainly be-
Puzzle Saga Studio (2018). Car Escape.
cause the ”weight” of this feature can vary from
https://play.google.com/store/apps/
player to player we cannot make any predictions on
details?id=com.puzzlegames.parking.
how it could be affecting the difficulty function non-
unblock.puzzle&hl=en&gl=US.
linearly.

You might also like