International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 5 (2018) pp.

2814-2816

Predictive Analysis of Sports Data using Google Prediction API

Ujwal UJ1, Dr Antony PJ2 and Sachin DN3

Abstract based on specific rules set. Until the batsmen gets out, he can
keep batting till the end of innings. The bowling team gets a
Prediction system is very essential to curb the curiosity of
wicket when each batsmen is out. An innings gets over when
anything. Many sports prediction systems are in great demand
either 10 wickets are down or when the specified number of
and data analysis plays a great role in prediction. Previous
overs is reached. The runs scored by the first team has to be
efforts in sports data analysis have resulted in prediction of
chased by the second team in order to win. If the team which
sports such as football, prediction of next shot location in
bats first defends their total, they win.
tennis, performance of athletes in Olympics, slam dunk shots
frequency in Basket Ball and many more. Cricket prediction is Prediction system works on the principles of machine learning.
comparatively difficult as there are many factors that can There are two types of machine learning namely supervised
influence the result or outcome of the cricket match. Earlier machine learning and unsupervised machine learning. In
basic prediction systems for cricket match consider only the supervised machine learning we must train the machine by
venue and disregard the factors like weather, stadium size, providing huge data sets and the outcomes. In this paper one
captaincy etc. The factors like venue of the match, pitch, such prediction methods is introduced which is used to make
weather conditions first batting or fielding all play a vital role predictions of the outcome of a cricket match using Google
in predicting the winner of the match. Suitable models are Prediction API.
necessary to predict and data mining makes it possible to
Google API is a black box prediction technique [2]. It is a form
extract required information from the data files. This paper
of supervised learning and hence it is required to provide huge
presents the usage of Google Prediction API to analyse the data
data and train the models. Google Prediction APIs make use of
of previous cricket matches and predict outcome of a given
Regression Algorithms when numerical predictions have to be
cricket match.
made. And Classifiers when the target output can assume only
Keywords: Google Prediction API, Linear Regression, Naive a limited set of values, either numbers or strings, based on the
Bayes application content. This API can only account for relatable
data. If the attributes are not related to one other, then a correct
probability curve will not be drawn. By providing a CSV file
INTRODUCTION of previous cricket matches and using appropriate queries to
extract the required data and train the model, predictions can be
The most popular sports in the world after Football is Cricket.
made.
Cricket is a very popular game in India, Pakistan, Australia,
England, South Africa and many more countries and it has
billions of fans across the globe. Especially in India, many fans
RELATED WORK
say "Cricket is my religion". The game is played between two
teams, each consisting of 11 players (15 when extra players are "Winning and Score Predicting (WASP)" [3] is a work done by
included) .The main part of the game takes place in the pitch of Scott Broker and Seamus Hogan at University of Canterbury as
22 yards length. It is played in 3 different formats, viz. One Day a part of their PhD research project. It predicts the additional
International (ODI), Twenty Overs International (T20) and runs that can be scored with the remaining wickets and balls in
Test. A bowler can bowl 6 successive legal deliveries and such the first innings. In the second innings it estimates the winning
6 deliveries are called as an "over". Initially depending on the probability.
toss one team shall bat first and the other team shall bowl. That
Tejinder Singh., proposed a model that has two methods to
will be the first innings. In the second innings, the team which
predict. First method predicts the score of first innings using
batted first shall bowl and the other team will bat. One Day is
Linear Regression Classifier and the second method uses Naive
played for 50 overs, each innings. Test is played for 5 days and
Bayes Classifier to predict the outcome of the match in second
for the first four days a minimum of 90 overs is to be played
innings. [4]
and on the last day a minimum of 75 overs [1] is to be played.
T20 is called limited overs match as only 20 overs per innings. Swartz et al. made use of Bayesian Latent variable model and
The pitch will be of 22 yards and the inner circle will be of 30 Markov Chain Monte Carlo [5] method to simulate ball by ball
yards. outcome. Kaluarachchi and Varde implemented association
rules and Naive Bayes classifier to analyze the factors that
contribute to a team's win. [6]
The batsmen have to score runs by hitting the ball bowled at
him by the bowlers. The batsman gets out on certain conditions