1
00:00:12,960 --> 00:00:17,279
hi everyone so i am internal rosh from

2
00:00:15,839 --> 00:00:19,840
the school of fundamental sciences

3
00:00:17,279 --> 00:00:21,439
massive university so today i am going

4
00:00:19,840 --> 00:00:24,160
to talk about

5
00:00:21,439 --> 00:00:26,480
football uh soccer data analysis uh

6
00:00:24,160 --> 00:00:27,840
pedagogy introduction this is my twitter

7
00:00:26,480 --> 00:00:30,480
handle

8
00:00:27,840 --> 00:00:32,960
and this is the link to my website

9
00:00:30,480 --> 00:00:35,200
thanks spike on a you for giving me this

10
00:00:32,960 --> 00:00:38,239
opportunity to speak at this conference

11
00:00:35,200 --> 00:00:40,960
so like i'm really glad that uh i'm

12
00:00:38,239 --> 00:00:43,360
present here with you all so this is my

13
00:00:40,960 --> 00:00:44,879
second year speaking at this renowned

14
00:00:43,360 --> 00:00:47,039
conference and

15
00:00:44,879 --> 00:00:48,879
welcome everyone so i'm going to

16
00:00:47,039 --> 00:00:51,039
talk about how to get open access even

17
00:00:48,879 --> 00:00:52,960
data from status bomb using the package

18
00:00:51,039 --> 00:00:54,719
called stats bom pi in python

19
00:00:52,960 --> 00:00:56,719
then how to draw a soccer pitch using

20
00:00:54,719 --> 00:00:58,399
mpl soccer package

21
00:00:56,719 --> 00:01:00,719
how to visualize a pass network for a

22
00:00:58,399 --> 00:01:03,760
particular team in a particular match

23
00:01:00,719 --> 00:01:06,960
and how to use uh network x module to

24
00:01:03,760 --> 00:01:08,479
analyze the pass network to draw pass

25
00:01:06,960 --> 00:01:09,760
maps along uh

26
00:01:08,479 --> 00:01:12,080
their corresponding and their

27
00:01:09,760 --> 00:01:13,119
corresponding heat maps

28
00:01:12,080 --> 00:01:14,880
of a

29
00:01:13,119 --> 00:01:16,320
of players from a particular team in a

30
00:01:14,880 --> 00:01:18,320
particular match

31
00:01:16,320 --> 00:01:19,840
and finally we will talk about how to

32
00:01:18,320 --> 00:01:21,680
implement computational geometric

33
00:01:19,840 --> 00:01:23,360
concepts like convex hulls voronoi

34
00:01:21,680 --> 00:01:25,520
diagrams and delanite triangulations

35
00:01:23,360 --> 00:01:27,520
using a python package called scipy dot

36
00:01:25,520 --> 00:01:29,520
special like any other package we use

37
00:01:27,520 --> 00:01:31,119
peep to install status bomb pipe

38
00:01:29,520 --> 00:01:33,360
and then uh

39
00:01:31,119 --> 00:01:35,520
the open data from status pike and stats

40
00:01:33,360 --> 00:01:37,600
bomb can be accessed and like uh please

41
00:01:35,520 --> 00:01:38,720
go through the terms and conditions uh

42
00:01:37,600 --> 00:01:40,560
section

43
00:01:38,720 --> 00:01:42,799
stated at their github documentation

44
00:01:40,560 --> 00:01:45,119
we'll go step by step uh to understand

45
00:01:42,799 --> 00:01:47,759
how to extract the relevant data

46
00:01:45,119 --> 00:01:49,920
and uh using the stats bomb by package

47
00:01:47,759 --> 00:01:51,280
so we also need to import the package

48
00:01:49,920 --> 00:01:52,799
and next we get access to the

49
00:01:51,280 --> 00:01:54,560
competitions data set what the

50
00:01:52,799 --> 00:01:56,799
competition's data set looks uh like

51
00:01:54,560 --> 00:01:58,240
comp data set looks like so we have a

52
00:01:56,799 --> 00:02:00,240
competition id

53
00:01:58,240 --> 00:02:02,719
and a season id so both these together

54
00:02:00,240 --> 00:02:06,960
act as uh like

55
00:02:02,719 --> 00:02:09,280
a unique id for extracting a particular

56
00:02:06,960 --> 00:02:11,200
row we also have the country name here

57
00:02:09,280 --> 00:02:12,640
actually it's the continent on uh from

58
00:02:11,200 --> 00:02:14,000
where the

59
00:02:12,640 --> 00:02:16,239
league belongs

60
00:02:14,000 --> 00:02:18,800
uh so here it's europe uh the

61
00:02:16,239 --> 00:02:21,599
competition name is uh champions league

62
00:02:18,800 --> 00:02:24,480
uh so and the gender is male so

63
00:02:21,599 --> 00:02:26,800
unfortunately the first 15 rows here

64
00:02:24,480 --> 00:02:28,720
that i am showing is uh

65
00:02:26,800 --> 00:02:30,319
of gender male but if you print the

66
00:02:28,720 --> 00:02:33,360
whole uh

67
00:02:30,319 --> 00:02:36,080
data set you will see there are

68
00:02:33,360 --> 00:02:40,000
like rows where genders are females too

69
00:02:36,080 --> 00:02:41,920
season name is uh 2018-19 or like 17-18

70
00:02:40,000 --> 00:02:44,800
and so on you can extract the column

71
00:02:41,920 --> 00:02:46,800
names of comp and this is what ah the

72
00:02:44,800 --> 00:02:49,280
column names are compressionary season

73
00:02:46,800 --> 00:02:51,200
id and so on ah now let us make sense of

74
00:02:49,280 --> 00:02:53,120
a particular row from the comp data set

75
00:02:51,200 --> 00:02:54,800
so for example we if we look into the

76
00:02:53,120 --> 00:02:56,800
row where the competition at is 16 and

77
00:02:54,800 --> 00:02:58,400
the season id is one we notice that the

78
00:02:56,800 --> 00:02:59,760
country name is europe the competition

79
00:02:58,400 --> 00:03:03,599
name is champions league and the season

80
00:02:59,760 --> 00:03:04,720
name is 2017-18 and so so on so suppose

81
00:03:03,599 --> 00:03:06,720
we are satisfied with the above

82
00:03:04,720 --> 00:03:09,599
information and we want to analyze a

83
00:03:06,720 --> 00:03:11,840
game from 1718's champions league season

84
00:03:09,599 --> 00:03:13,040
so we use both this competition id and

85
00:03:11,840 --> 00:03:13,920
season ids

86
00:03:13,040 --> 00:03:15,920
and

87
00:03:13,920 --> 00:03:17,920
this is what we write uh to extract that

88
00:03:15,920 --> 00:03:19,519
particular row so math equals to sb dot

89
00:03:17,920 --> 00:03:22,239
matches competition id equals to 16 and

90
00:03:19,519 --> 00:03:24,080
season 8 equals to 1 so this gives

91
00:03:22,239 --> 00:03:26,879
us the particular match from where we

92
00:03:24,080 --> 00:03:30,640
will extract the even data

93
00:03:26,879 --> 00:03:32,560
so this is uh the matte data set and we

94
00:03:30,640 --> 00:03:34,480
see it also has a match id which is

95
00:03:32,560 --> 00:03:37,599
again the unique id for this particular

96
00:03:34,480 --> 00:03:38,640
data set mat but here we have only one

97
00:03:37,599 --> 00:03:40,879
row

98
00:03:38,640 --> 00:03:43,599
for this data set and the match date

99
00:03:40,879 --> 00:03:44,400
when the match took place uh the kickoff

100
00:03:43,599 --> 00:03:46,799
time

101
00:03:44,400 --> 00:03:49,400
uh as we know it's an european champions

102
00:03:46,799 --> 00:03:51,760
league uh game so it was on

103
00:03:49,400 --> 00:03:53,519
2017-18 the home team was real madrid

104
00:03:51,760 --> 00:03:55,360
and the other team was liverpool and the

105
00:03:53,519 --> 00:03:58,400
home team score was three and their team

106
00:03:55,360 --> 00:04:00,560
score was one now we use this match id

107
00:03:58,400 --> 00:04:01,840
uh to extract the events and store it in

108
00:04:00,560 --> 00:04:04,159
events data set

109
00:04:01,840 --> 00:04:05,519
okay ah so the events data set uh

110
00:04:04,159 --> 00:04:07,200
fetching as the events data for the

111
00:04:05,519 --> 00:04:09,519
particular match and that's what it

112
00:04:07,200 --> 00:04:11,439
looks like it has a

113
00:04:09,519 --> 00:04:13,040
large number of rows

114
00:04:11,439 --> 00:04:15,120
and columns so we see that we are able

115
00:04:13,040 --> 00:04:18,160
to get access to all the events from uh

116
00:04:15,120 --> 00:04:20,400
the real madrid versus the purple match

117
00:04:18,160 --> 00:04:23,520
and like we can also print the column

118
00:04:20,400 --> 00:04:27,360
names using events.columns uh now we

119
00:04:23,520 --> 00:04:31,520
will see how we can draw a

120
00:04:27,360 --> 00:04:31,520
football pitch using mpl soccer

121
00:04:31,680 --> 00:04:36,560
and this package was developed by uh

122
00:04:34,639 --> 00:04:38,560
and developed and maintained by animal

123
00:04:36,560 --> 00:04:40,000
rogapal and andrew rollinson please go

124
00:04:38,560 --> 00:04:41,520
and check them out

125
00:04:40,000 --> 00:04:43,440
and like

126
00:04:41,520 --> 00:04:44,639
like any other package we pp installed

127
00:04:43,440 --> 00:04:46,880
it first

128
00:04:44,639 --> 00:04:48,639
and then we uh

129
00:04:46,880 --> 00:04:49,600
we also need to import the matplotlib

130
00:04:48,639 --> 00:04:53,040
and the

131
00:04:49,600 --> 00:04:54,840
matplotlib and the pitch classes uh so

132
00:04:53,040 --> 00:04:57,120
this is what we do here import macbook

133
00:04:54,840 --> 00:05:00,000
live.pipe.spl and uh from mpl

134
00:04:57,120 --> 00:05:03,440
soccer.page import

135
00:05:00,000 --> 00:05:06,160
page but here p is capacity that uh with

136
00:05:03,440 --> 00:05:08,080
the function pitch with again p capital

137
00:05:06,160 --> 00:05:10,400
uh we set the page pitch color to be

138
00:05:08,080 --> 00:05:12,800
grass the line color to be white

139
00:05:10,400 --> 00:05:14,240
stripe to be true and so on okay and

140
00:05:12,800 --> 00:05:16,160
these are different parameters that you

141
00:05:14,240 --> 00:05:18,560
can enter and if you draw it you will

142
00:05:16,160 --> 00:05:21,199
end up with this page so this is the

143
00:05:18,560 --> 00:05:23,600
default one uh the stats bomb uh type

144
00:05:21,199 --> 00:05:26,960
page uh which is provide provided by the

145
00:05:23,600 --> 00:05:30,160
stats boom by package where uh the x and

146
00:05:26,960 --> 00:05:33,360
like x axis ranges from zero to 120 and

147
00:05:30,160 --> 00:05:35,440
the y axis ranges from 80 to zero

148
00:05:33,360 --> 00:05:37,120
and but there are different uh types of

149
00:05:35,440 --> 00:05:39,120
features that you can draw with uh just

150
00:05:37,120 --> 00:05:40,960
as one by package uh except the default

151
00:05:39,120 --> 00:05:42,479
one like apart from the default one you

152
00:05:40,960 --> 00:05:44,320
can also set the color to be black and

153
00:05:42,479 --> 00:05:46,400
this is what you will end up with then

154
00:05:44,320 --> 00:05:48,960
the default one stats bomb by their

155
00:05:46,400 --> 00:05:51,360
stats bomb there are different uh

156
00:05:48,960 --> 00:05:54,320
other pictures like opta track cap skill

157
00:05:51,360 --> 00:05:55,520
corner y scout metric sports ufr custom

158
00:05:54,320 --> 00:05:56,800
and so on

159
00:05:55,520 --> 00:05:58,639
okay

160
00:05:56,800 --> 00:06:01,280
like this can be set using pitch type

161
00:05:58,639 --> 00:06:04,160
argument inside the pitch function

162
00:06:01,280 --> 00:06:07,520
here i have set the pitch type to be ufo

163
00:06:04,160 --> 00:06:10,000
and here we see that x ranges from 0 to

164
00:06:07,520 --> 00:06:12,160
slightly above 100 and y ranges from 0

165
00:06:10,000 --> 00:06:13,520
to 70. these are the most basic concepts

166
00:06:12,160 --> 00:06:15,759
covering the topic of drawing and

167
00:06:13,520 --> 00:06:19,039
visualizing a football page using mpl

168
00:06:15,759 --> 00:06:20,639
soccer and now we will use uh

169
00:06:19,039 --> 00:06:23,680
the concepts from complex network

170
00:06:20,639 --> 00:06:25,840
analysis to draw a pass network uh for a

171
00:06:23,680 --> 00:06:27,440
particular team from a particular match

172
00:06:25,840 --> 00:06:29,759
and then we will analyze that network

173
00:06:27,440 --> 00:06:31,440
using networkx package for that we need

174
00:06:29,759 --> 00:06:33,120
to keep install network x we import

175
00:06:31,440 --> 00:06:35,680
network access nx

176
00:06:33,120 --> 00:06:38,560
we also keep install seaborn

177
00:06:35,680 --> 00:06:40,160
and import it at sns if you go back and

178
00:06:38,560 --> 00:06:41,680
look into the events data set you will

179
00:06:40,160 --> 00:06:43,280
notice that there is a column named

180
00:06:41,680 --> 00:06:45,039
tactics that provides us with the team

181
00:06:43,280 --> 00:06:46,319
lineups formations player ids and

182
00:06:45,039 --> 00:06:49,759
they're just a number from both the

183
00:06:46,319 --> 00:06:52,560
teams okay so the corresponding uh

184
00:06:49,759 --> 00:06:54,240
row values for a column type gives us an

185
00:06:52,560 --> 00:06:56,319
idea about whether it was the starting

186
00:06:54,240 --> 00:06:58,000
element for formation or was a tactical

187
00:06:56,319 --> 00:06:59,520
shift so let us generate a completely

188
00:06:58,000 --> 00:07:01,440
new data set only focusing on the

189
00:06:59,520 --> 00:07:03,120
tactics and the type columns

190
00:07:01,440 --> 00:07:05,440
and we'll filter the data in such a way

191
00:07:03,120 --> 00:07:07,919
that the tactics column has no rows set

192
00:07:05,440 --> 00:07:09,599
to name okay so that's what we do here

193
00:07:07,919 --> 00:07:11,759
so this is what it looks like here we

194
00:07:09,599 --> 00:07:14,639
see

195
00:07:11,759 --> 00:07:15,599
the entry uh of the tactics column

196
00:07:14,639 --> 00:07:16,479
uh

197
00:07:15,599 --> 00:07:19,840
is a

198
00:07:16,479 --> 00:07:22,319
dictionary but the key formation

199
00:07:19,840 --> 00:07:25,520
uh gives the formation of the players

200
00:07:22,319 --> 00:07:25,520
the key lineup gives the

201
00:07:25,759 --> 00:07:29,759
player ids names and jersey numbers

202
00:07:28,400 --> 00:07:31,199
for both the team let's focus only on

203
00:07:29,759 --> 00:07:34,240
the tactics for the starting eleven

204
00:07:31,199 --> 00:07:36,720
setup from both the teams and uh like uh

205
00:07:34,240 --> 00:07:38,240
this is given so this i got this idea

206
00:07:36,720 --> 00:07:41,919
from uh

207
00:07:38,240 --> 00:07:44,960
like from the mpl soccer

208
00:07:41,919 --> 00:07:47,759
package uh documentation uh website and

209
00:07:44,960 --> 00:07:49,599
also from the uh youtube video by mac

210
00:07:47,759 --> 00:07:51,840
johns and like uh you should go and

211
00:07:49,599 --> 00:07:54,080
check both these resources out

212
00:07:51,840 --> 00:07:55,520
okay so we will build the analysis uh

213
00:07:54,080 --> 00:07:57,039
build and analyze the past network

214
00:07:55,520 --> 00:07:58,960
generated from among the starting 11

215
00:07:57,039 --> 00:08:00,560
players from either of the tips

216
00:07:58,960 --> 00:08:03,360
okay if you look into the first two rows

217
00:08:00,560 --> 00:08:05,680
of the type column intact we see that

218
00:08:03,360 --> 00:08:08,080
they are set as starting 11

219
00:08:05,680 --> 00:08:10,240
one for each team so let us separately

220
00:08:08,080 --> 00:08:12,160
face the data for the teams so that's

221
00:08:10,240 --> 00:08:14,080
what we do here so it's as i said it was

222
00:08:12,160 --> 00:08:16,960
a the tactics column is made up of a

223
00:08:14,080 --> 00:08:18,960
python dictionary object and for now we

224
00:08:16,960 --> 00:08:19,919
are only interested in the key lineup

225
00:08:18,960 --> 00:08:22,319
okay

226
00:08:19,919 --> 00:08:23,440
so we just extract the information from

227
00:08:22,319 --> 00:08:25,840
the lineup

228
00:08:23,440 --> 00:08:27,039
key we then convert that into a data

229
00:08:25,840 --> 00:08:30,479
frame using

230
00:08:27,039 --> 00:08:32,080
from dict from dict function of the

231
00:08:30,479 --> 00:08:34,719
pandas data frame

232
00:08:32,080 --> 00:08:36,080
we end up with this okay so we say that

233
00:08:34,719 --> 00:08:37,599
the

234
00:08:36,080 --> 00:08:39,039
for the

235
00:08:37,599 --> 00:08:42,000
column player

236
00:08:39,039 --> 00:08:44,640
suppose the id is 5597 and his name was

237
00:08:42,000 --> 00:08:46,720
keller nevers and his position is a

238
00:08:44,640 --> 00:08:48,560
goalkeeper and

239
00:08:46,720 --> 00:08:50,000
his id is one and the jersey number is

240
00:08:48,560 --> 00:08:51,519
also given so we are basically

241
00:08:50,000 --> 00:08:53,680
interested in the players names and

242
00:08:51,519 --> 00:08:55,440
their corresponding jersey numbers so we

243
00:08:53,680 --> 00:08:57,279
will use a simple for loop and store the

244
00:08:55,440 --> 00:08:59,600
information in separate dictionaries for

245
00:08:57,279 --> 00:09:00,959
both the teams okay so that's what we do

246
00:08:59,600 --> 00:09:02,240
here okay so now we have collected the

247
00:09:00,959 --> 00:09:04,160
names and the jeffy numbers of the

248
00:09:02,240 --> 00:09:05,839
players the starting eleven players

249
00:09:04,160 --> 00:09:08,080
uh from both the team from the events

250
00:09:05,839 --> 00:09:10,240
data set will extract out the relevant

251
00:09:08,080 --> 00:09:12,000
columns for our past network analysis

252
00:09:10,240 --> 00:09:15,360
and these are the relevant columns

253
00:09:12,000 --> 00:09:17,760
minute second team type location

254
00:09:15,360 --> 00:09:19,760
person location pass outcome and player

255
00:09:17,760 --> 00:09:22,480
so this is the first 10 rows of the

256
00:09:19,760 --> 00:09:23,680
events pn data frame so minute second

257
00:09:22,480 --> 00:09:25,519
team

258
00:09:23,680 --> 00:09:27,120
type location password location pass

259
00:09:25,519 --> 00:09:28,720
outcome player next step is to filter

260
00:09:27,120 --> 00:09:30,800
out the data set by teams and store them

261
00:09:28,720 --> 00:09:33,200
as new data sets okay one for liverpool

262
00:09:30,800 --> 00:09:34,720
and one for real madrid

263
00:09:33,200 --> 00:09:36,560
that's what we do here as we are only

264
00:09:34,720 --> 00:09:38,560
interested in the past uh network

265
00:09:36,560 --> 00:09:40,720
generation we will filter the data sets

266
00:09:38,560 --> 00:09:43,200
by keeping only those rows where type is

267
00:09:40,720 --> 00:09:45,120
said to be pass and will dis discard all

268
00:09:43,200 --> 00:09:47,200
the other types again view the first 10

269
00:09:45,120 --> 00:09:51,040
rows of the filter data sets so we see

270
00:09:47,200 --> 00:09:53,120
that for real madrid uh here the team is

271
00:09:51,040 --> 00:09:55,279
only real madrid and like it was the

272
00:09:53,120 --> 00:09:56,959
type is only pass

273
00:09:55,279 --> 00:09:59,020
location

274
00:09:56,959 --> 00:10:00,560
here if you notice the location is a

275
00:09:59,020 --> 00:10:02,880
[Music]

276
00:10:00,560 --> 00:10:04,880
like it's a list of two numbers which

277
00:10:02,880 --> 00:10:07,200
gives the x and y coordinates and same

278
00:10:04,880 --> 00:10:09,600
for pass and location okay so later we

279
00:10:07,200 --> 00:10:12,000
will uh like separate this location

280
00:10:09,600 --> 00:10:13,440
column into location x and location y uh

281
00:10:12,000 --> 00:10:14,959
for the

282
00:10:13,440 --> 00:10:16,640
one column for x coordinates and the

283
00:10:14,959 --> 00:10:18,560
other column for y coordinates and same

284
00:10:16,640 --> 00:10:19,839
for pass and location let us now uh very

285
00:10:18,560 --> 00:10:22,480
carefully observe the datasets so

286
00:10:19,839 --> 00:10:24,240
suppose from eventspn uh real dataset we

287
00:10:22,480 --> 00:10:26,640
are focusing on the second and the third

288
00:10:24,240 --> 00:10:29,040
row command which makes the pass at

289
00:10:26,640 --> 00:10:30,720
around zeroth minute and tenth second

290
00:10:29,040 --> 00:10:33,040
second row and danny carvajal receives

291
00:10:30,720 --> 00:10:35,920
the pass at around 0th minute and 11

292
00:10:33,040 --> 00:10:38,000
seconds so in both the data sets we now

293
00:10:35,920 --> 00:10:39,920
need to add two extra columns named as

294
00:10:38,000 --> 00:10:42,160
pass maker and pass receiver but the

295
00:10:39,920 --> 00:10:43,600
pass maker column would be similar to

296
00:10:42,160 --> 00:10:45,680
player column and the pass receiver

297
00:10:43,600 --> 00:10:48,160
column would be the player column whose

298
00:10:45,680 --> 00:10:49,920
index would be shifted by one place in

299
00:10:48,160 --> 00:10:51,839
the negative direction what i learned uh

300
00:10:49,920 --> 00:10:54,240
from the

301
00:10:51,839 --> 00:10:56,640
like mpl soccer's uh documentation and

302
00:10:54,240 --> 00:10:58,800
the youtube video by mike johns so go

303
00:10:56,640 --> 00:11:00,480
take again go check them out

304
00:10:58,800 --> 00:11:02,640
so this can be achieved by the shift

305
00:11:00,480 --> 00:11:04,640
function provided by pandas so we will

306
00:11:02,640 --> 00:11:06,880
perform the operation on both the uh

307
00:11:04,640 --> 00:11:08,000
both events p and real and even spin uh

308
00:11:06,880 --> 00:11:09,040
leave

309
00:11:08,000 --> 00:11:10,720
okay

310
00:11:09,040 --> 00:11:13,440
so that's what we do here so this is the

311
00:11:10,720 --> 00:11:16,079
modified data set um so

312
00:11:13,440 --> 00:11:17,920
we see that uh we have added

313
00:11:16,079 --> 00:11:19,680
the pass maker uh column and the past

314
00:11:17,920 --> 00:11:20,959
receiver column uh now there might be

315
00:11:19,680 --> 00:11:23,839
passes which are which were not

316
00:11:20,959 --> 00:11:25,839
successful okay so one trick like one

317
00:11:23,839 --> 00:11:28,480
thing to notice is that in stats bomb

318
00:11:25,839 --> 00:11:30,160
data passes whose pass outcome are set

319
00:11:28,480 --> 00:11:32,240
as name are actually the successful

320
00:11:30,160 --> 00:11:34,320
passes okay so this is very important

321
00:11:32,240 --> 00:11:36,560
thing to keep in mind we'll again filter

322
00:11:34,320 --> 00:11:38,720
the data sets by successful process

323
00:11:36,560 --> 00:11:39,760
uh so these are the modified data sets

324
00:11:38,720 --> 00:11:41,200
again

325
00:11:39,760 --> 00:11:43,279
so it seems we have been able to

326
00:11:41,200 --> 00:11:44,880
logically clean and modify the data sets

327
00:11:43,279 --> 00:11:46,480
now we are only focused on building the

328
00:11:44,880 --> 00:11:47,920
past network among the players who are

329
00:11:46,480 --> 00:11:49,440
in the starting level

330
00:11:47,920 --> 00:11:51,120
okay

331
00:11:49,440 --> 00:11:53,200
so for that we need the time when the

332
00:11:51,120 --> 00:11:55,360
first substitution took place there were

333
00:11:53,200 --> 00:11:57,040
three substitution for real madrid and

334
00:11:55,360 --> 00:11:59,760
the first substitution took place at

335
00:11:57,040 --> 00:12:01,519
36th minute at 17 second

336
00:11:59,760 --> 00:12:02,639
okay and same for liverpool there were

337
00:12:01,519 --> 00:12:05,040
two substitutions and the first

338
00:12:02,639 --> 00:12:07,200
substitution took place at 29 2013

339
00:12:05,040 --> 00:12:09,600
second

340
00:12:07,200 --> 00:12:12,240
so that's what we do here with this

341
00:12:09,600 --> 00:12:13,760
sample code we like try to find out the

342
00:12:12,240 --> 00:12:17,120
minute and second

343
00:12:13,760 --> 00:12:18,560
so we use the minute

344
00:12:17,120 --> 00:12:21,279
value to

345
00:12:18,560 --> 00:12:23,680
like discard all those uh rows which

346
00:12:21,279 --> 00:12:25,680
appeared after that

347
00:12:23,680 --> 00:12:27,600
minute okay

348
00:12:25,680 --> 00:12:30,720
all those events that occurred after

349
00:12:27,600 --> 00:12:32,720
that uh time now from the data sets we

350
00:12:30,720 --> 00:12:34,800
split the location and the parcel

351
00:12:32,720 --> 00:12:36,880
location into two columns as i said one

352
00:12:34,800 --> 00:12:38,880
for uh x coordinates and another for y

353
00:12:36,880 --> 00:12:41,120
coordinates so inspired by the way given

354
00:12:38,880 --> 00:12:43,839
here uh if you look click in this link

355
00:12:41,120 --> 00:12:45,839
you will see uh in the apple circle uh

356
00:12:43,839 --> 00:12:47,519
documentation page we will take the

357
00:12:45,839 --> 00:12:49,600
average locations of the starting eleven

358
00:12:47,519 --> 00:12:52,000
players uh on the field for for a

359
00:12:49,600 --> 00:12:53,920
unified construction of the past network

360
00:12:52,000 --> 00:12:55,200
so for that we will use the aggregate

361
00:12:53,920 --> 00:12:57,120
function

362
00:12:55,200 --> 00:12:58,720
and what it does is that uh we'll count

363
00:12:57,120 --> 00:12:59,600
the number of passes created by this

364
00:12:58,720 --> 00:13:02,240
player

365
00:12:59,600 --> 00:13:04,560
okay so

366
00:13:02,240 --> 00:13:06,639
here f lock real gives the average

367
00:13:04,560 --> 00:13:08,800
location of the player of a particular

368
00:13:06,639 --> 00:13:10,560
pillar from a particular team in that

369
00:13:08,800 --> 00:13:13,279
match as i said using the aggregate

370
00:13:10,560 --> 00:13:15,279
function gives the mean of

371
00:13:13,279 --> 00:13:17,360
the x and y coordinates of for a

372
00:13:15,279 --> 00:13:19,040
particular player and for case scenario

373
00:13:17,360 --> 00:13:21,600
we see that the mean is

374
00:13:19,040 --> 00:13:22,959
eight four five in the x coordinate and

375
00:13:21,600 --> 00:13:26,079
thirty one point eight three six in the

376
00:13:22,959 --> 00:13:27,600
y coordinate and he completed uh eleven

377
00:13:26,079 --> 00:13:29,120
passes once we sort out the starting

378
00:13:27,600 --> 00:13:30,560
eleven players average locations in a

379
00:13:29,120 --> 00:13:33,120
game we'll try to figure out the number

380
00:13:30,560 --> 00:13:34,959
of times a particular pass maker pass

381
00:13:33,120 --> 00:13:36,399
the ball to a particular pass receiver

382
00:13:34,959 --> 00:13:38,399
because just to keep the direction of

383
00:13:36,399 --> 00:13:40,480
pass in mind that is a pass from player

384
00:13:38,399 --> 00:13:42,480
a to player b is not equal identical to

385
00:13:40,480 --> 00:13:44,959
the pass from player b to player a so it

386
00:13:42,480 --> 00:13:46,399
would give a directed ah network so

387
00:13:44,959 --> 00:13:48,079
we'll use group by and the count

388
00:13:46,399 --> 00:13:50,399
function to count the number of rows

389
00:13:48,079 --> 00:13:52,880
where a unique player a pass the ball to

390
00:13:50,399 --> 00:13:54,880
another unique player b

391
00:13:52,880 --> 00:13:57,360
so here we use the group by function and

392
00:13:54,880 --> 00:13:59,440
here we just reset the index accordingly

393
00:13:57,360 --> 00:14:01,120
and we see that uh here the pass maker

394
00:13:59,440 --> 00:14:03,760
is kasamiro and the first receiver is

395
00:14:01,120 --> 00:14:06,079
carvajal and kasamiro passed only once

396
00:14:03,760 --> 00:14:08,800
in the entire game to uh danny carvajal

397
00:14:06,079 --> 00:14:10,800
he passed six times to tony cruz so we

398
00:14:08,800 --> 00:14:12,880
see that six plus seven eight nine ten

399
00:14:10,800 --> 00:14:15,040
eleven so yeah there there were eleven

400
00:14:12,880 --> 00:14:16,959
passes if you remember for uh december

401
00:14:15,040 --> 00:14:18,959
that will merge the data sets have

402
00:14:16,959 --> 00:14:21,199
looked real and pass real and let us

403
00:14:18,959 --> 00:14:23,839
identify the left and right data frames

404
00:14:21,199 --> 00:14:25,839
for uh performing the merge okay here

405
00:14:23,839 --> 00:14:28,000
avlog real is the left data frame and

406
00:14:25,839 --> 00:14:31,440
the pass real is the right

407
00:14:28,000 --> 00:14:33,360
so we just merge them and uh

408
00:14:31,440 --> 00:14:35,279
this is what the new data set looks like

409
00:14:33,360 --> 00:14:37,760
so pass maker pass receiver the number

410
00:14:35,279 --> 00:14:39,040
of passes uh given by the pass maker to

411
00:14:37,760 --> 00:14:41,680
pass receiver

412
00:14:39,040 --> 00:14:43,760
pass makers average x location

413
00:14:41,680 --> 00:14:44,800
x coordinate pass makers average y

414
00:14:43,760 --> 00:14:46,560
coordinate

415
00:14:44,800 --> 00:14:48,959
and the total number of

416
00:14:46,560 --> 00:14:51,279
passes completed by the pass maker so we

417
00:14:48,959 --> 00:14:53,519
see that it's 11 for all the

418
00:14:51,279 --> 00:14:56,079
cases rows where uh the pass maker is

419
00:14:53,519 --> 00:14:56,079
casimiro

420
00:14:56,160 --> 00:15:01,360
and same for liverpool

421
00:14:59,040 --> 00:15:03,040
uh so finally uh we will again perform a

422
00:15:01,360 --> 00:15:04,720
merge on these updated data sets for

423
00:15:03,040 --> 00:15:06,320
adding the average locations of the past

424
00:15:04,720 --> 00:15:08,720
receivers and the number of times the

425
00:15:06,320 --> 00:15:11,040
receiver receive the ball okay

426
00:15:08,720 --> 00:15:11,040
so

427
00:15:11,199 --> 00:15:14,000
that's what we do

428
00:15:12,399 --> 00:15:14,800
and uh

429
00:15:14,000 --> 00:15:16,480
like

430
00:15:14,800 --> 00:15:18,880
we added this column number of passes

431
00:15:16,480 --> 00:15:21,920
received by that particular uh

432
00:15:18,880 --> 00:15:24,320
player so casimiro i think received 24

433
00:15:21,920 --> 00:15:26,079
passes then lastly we will replace the

434
00:15:24,320 --> 00:15:27,199
players names with their jersey numbers

435
00:15:26,079 --> 00:15:29,600
okay

436
00:15:27,199 --> 00:15:31,360
and then we just visualize the pass

437
00:15:29,600 --> 00:15:33,519
pass network for both the teams so we

438
00:15:31,360 --> 00:15:36,240
use the pitch function to draw the page

439
00:15:33,519 --> 00:15:38,079
and then we use arrows to draw the edges

440
00:15:36,240 --> 00:15:42,000
from between the nodes

441
00:15:38,079 --> 00:15:44,639
and then pitch dot scatter uh

442
00:15:42,000 --> 00:15:46,320
so ps dot arrows function uh does that

443
00:15:44,639 --> 00:15:49,199
and ps dot scatter function draws the

444
00:15:46,320 --> 00:15:51,680
nodes the thicker the edge uh the number

445
00:15:49,199 --> 00:15:53,600
of more the number of

446
00:15:51,680 --> 00:15:55,839
passes were played so that's what it

447
00:15:53,600 --> 00:15:58,639
actually uh means

448
00:15:55,839 --> 00:16:00,240
and yeah so this is the

449
00:15:58,639 --> 00:16:01,600
first network for real madrid against

450
00:16:00,240 --> 00:16:05,839
liverpool

451
00:16:01,600 --> 00:16:08,320
and this is for liverpool okay

452
00:16:05,839 --> 00:16:09,920
and now as we have drawn the visualize

453
00:16:08,320 --> 00:16:13,920
the past networks we can

454
00:16:09,920 --> 00:16:15,279
perform some complex network analysis by

455
00:16:13,920 --> 00:16:16,560
analyzing our networks using some

456
00:16:15,279 --> 00:16:18,800
metrics

457
00:16:16,560 --> 00:16:18,800
and

458
00:16:19,040 --> 00:16:23,440
as uh let us first develop the

459
00:16:20,480 --> 00:16:26,079
isomorphic graph uh to the ones that we

460
00:16:23,440 --> 00:16:29,120
just saw uh and by using the network

461
00:16:26,079 --> 00:16:30,720
package network x package and the thing

462
00:16:29,120 --> 00:16:33,120
is uh we don't need the average

463
00:16:30,720 --> 00:16:35,040
locations here okay uh we are only

464
00:16:33,120 --> 00:16:36,880
interested on the

465
00:16:35,040 --> 00:16:38,959
nodes and the information about the

466
00:16:36,880 --> 00:16:41,040
nodes and the edges but not about the

467
00:16:38,959 --> 00:16:43,040
locations what we do is that

468
00:16:41,040 --> 00:16:44,959
we create this data set pass real new

469
00:16:43,040 --> 00:16:46,160
from password

470
00:16:44,959 --> 00:16:48,720
new where

471
00:16:46,160 --> 00:16:50,399
the pass maker we only take the columns

472
00:16:48,720 --> 00:16:52,800
pass maker pass receiver and number of

473
00:16:50,399 --> 00:16:54,880
passes use the nx dot digraph function

474
00:16:52,800 --> 00:16:56,560
because it's a directed graph so digraph

475
00:16:54,880 --> 00:16:59,120
function from network x

476
00:16:56,560 --> 00:17:01,040
and then we add the edge

477
00:16:59,120 --> 00:17:02,880
along with the weight yes so this is

478
00:17:01,040 --> 00:17:04,720
what it looks like okay now we will use

479
00:17:02,880 --> 00:17:06,079
network x to find out the node degrees

480
00:17:04,720 --> 00:17:08,559
from uh

481
00:17:06,079 --> 00:17:10,880
past network of real madrid so

482
00:17:08,559 --> 00:17:11,760
by using the nx dot degree function on g

483
00:17:10,880 --> 00:17:13,520
real

484
00:17:11,760 --> 00:17:14,880
and we see that these are the node

485
00:17:13,520 --> 00:17:16,400
degrees

486
00:17:14,880 --> 00:17:17,679
so node degrees means the number of

487
00:17:16,400 --> 00:17:20,319
passes

488
00:17:17,679 --> 00:17:21,760
the player was actually

489
00:17:20,319 --> 00:17:25,120
involved in

490
00:17:21,760 --> 00:17:26,559
so for real madrid we see uh that jersey

491
00:17:25,120 --> 00:17:28,079
number eight

492
00:17:26,559 --> 00:17:30,720
was involved in the maximum number of

493
00:17:28,079 --> 00:17:32,799
passes we can create the in degrees and

494
00:17:30,720 --> 00:17:34,640
output out degrees to so that means the

495
00:17:32,799 --> 00:17:37,919
number of in degrees means the number of

496
00:17:34,640 --> 00:17:40,240
passes a player received and out degrees

497
00:17:37,919 --> 00:17:42,480
means the number of passes the player

498
00:17:40,240 --> 00:17:44,400
gave to some other players we can also

499
00:17:42,480 --> 00:17:47,200
generate adjacency matrix

500
00:17:44,400 --> 00:17:48,640
and like we see the diagonal had uh all

501
00:17:47,200 --> 00:17:50,400
zeros that means

502
00:17:48,640 --> 00:17:52,240
there were no self loops so a player

503
00:17:50,400 --> 00:17:54,400
cannot pass to himself now we can work

504
00:17:52,240 --> 00:17:56,799
on a metric that focuses on geodesic

505
00:17:54,400 --> 00:17:58,559
distance between two players nodes uh

506
00:17:56,799 --> 00:18:00,720
two player nodes in a graph one way to

507
00:17:58,559 --> 00:18:03,679
implement this is to divide one by the

508
00:18:00,720 --> 00:18:05,840
weight column and that will give a a

509
00:18:03,679 --> 00:18:08,000
new graph this is the

510
00:18:05,840 --> 00:18:09,039
graph that we end up with

511
00:18:08,000 --> 00:18:10,799
and

512
00:18:09,039 --> 00:18:12,640
we'll perform the same operations to

513
00:18:10,799 --> 00:18:14,799
create a modify modified graph for

514
00:18:12,640 --> 00:18:17,039
liverpool 2

515
00:18:14,799 --> 00:18:19,039
and now using this modified graph we can

516
00:18:17,039 --> 00:18:20,640
calculate the all pair shortest paths

517
00:18:19,039 --> 00:18:22,720
between the nodes players for both the

518
00:18:20,640 --> 00:18:24,559
teams let us compute first for real

519
00:18:22,720 --> 00:18:26,960
matrix if you intend to find the

520
00:18:24,559 --> 00:18:29,520
shortest path between two players like

521
00:18:26,960 --> 00:18:31,679
what would have been the shortest

522
00:18:29,520 --> 00:18:34,160
path for a pass from one player to

523
00:18:31,679 --> 00:18:36,080
another so suppose we want to calculate

524
00:18:34,160 --> 00:18:38,320
the shortest path from the goalkeeper

525
00:18:36,080 --> 00:18:40,640
killer nervous to the forward christian

526
00:18:38,320 --> 00:18:43,679
ronaldo we'll just type the following so

527
00:18:40,640 --> 00:18:46,480
print uh this real one to seven and we

528
00:18:43,679 --> 00:18:48,960
see that from keller never sergio ramos

529
00:18:46,480 --> 00:18:50,799
and from sargeras to

530
00:18:48,960 --> 00:18:53,120
whoever the player

531
00:18:50,799 --> 00:18:56,240
was with jesse number 12 to cristina

532
00:18:53,120 --> 00:18:58,480
ronaldo i think uh

533
00:18:56,240 --> 00:19:00,080
okay it was marcelo and with him

534
00:18:58,480 --> 00:19:01,760
ultimately passing to uh christian and

535
00:19:00,080 --> 00:19:03,919
ronaldo would have been the

536
00:19:01,760 --> 00:19:05,200
uh

537
00:19:03,919 --> 00:19:07,679
like

538
00:19:05,200 --> 00:19:09,120
would have been the uh shortest path and

539
00:19:07,679 --> 00:19:10,559
this seems like a good post match

540
00:19:09,120 --> 00:19:13,360
analysis tool

541
00:19:10,559 --> 00:19:17,440
i got this idea from discussing with uh

542
00:19:13,360 --> 00:19:20,240
sarah babu and he is from iist india

543
00:19:17,440 --> 00:19:22,000
now we can calculate the eccentricity uh

544
00:19:20,240 --> 00:19:24,160
which is based on the shortest distance

545
00:19:22,000 --> 00:19:26,559
so eccentricity of a player note p tells

546
00:19:24,160 --> 00:19:29,679
us how far the furthest player note from

547
00:19:26,559 --> 00:19:32,480
a p is positioned in the past network so

548
00:19:29,679 --> 00:19:33,840
this is for all the players for for for

549
00:19:32,480 --> 00:19:35,840
real madrid

550
00:19:33,840 --> 00:19:38,480
and this is the average uh

551
00:19:35,840 --> 00:19:41,520
eccentricity and for liverpool also

552
00:19:38,480 --> 00:19:42,720
we calculate the average assembly speed

553
00:19:41,520 --> 00:19:43,919
we can also calculate the average

554
00:19:42,720 --> 00:19:46,160
clustering coefficient of a player

555
00:19:43,919 --> 00:19:48,320
calculated it's 0.182 and for liverpool

556
00:19:46,160 --> 00:19:50,000
it's 0.276

557
00:19:48,320 --> 00:19:52,640
so the average clustering coefficient

558
00:19:50,000 --> 00:19:54,400
lies in the range 0 and 1 where a value

559
00:19:52,640 --> 00:19:56,320
of 0 denotes the fact that none of the

560
00:19:54,400 --> 00:19:58,080
nodes are connected to each other and

561
00:19:56,320 --> 00:20:00,000
value of 1 denotes the fact that the

562
00:19:58,080 --> 00:20:01,840
network is a click that is each node is

563
00:20:00,000 --> 00:20:04,880
connected to all the other nodes of the

564
00:20:01,840 --> 00:20:07,840
network also can compute the centrality

565
00:20:04,880 --> 00:20:09,520
of each player uh and

566
00:20:07,840 --> 00:20:10,880
for each node in either team's first

567
00:20:09,520 --> 00:20:12,880
network i understand which player was

568
00:20:10,880 --> 00:20:14,720
the most important so sent more the

569
00:20:12,880 --> 00:20:17,280
centrality of a particular player more

570
00:20:14,720 --> 00:20:19,919
important he was in that match

571
00:20:17,280 --> 00:20:22,400
uh so for real madrid it was casimiro

572
00:20:19,919 --> 00:20:24,320
and for liverpool it was milner right so

573
00:20:22,400 --> 00:20:26,640
next we'll talk about uh pass map and

574
00:20:24,320 --> 00:20:27,760
the corresponding heat pumps so we will

575
00:20:26,640 --> 00:20:30,080
again uh

576
00:20:27,760 --> 00:20:32,080
from the same material again work on the

577
00:20:30,080 --> 00:20:33,280
fetching the pass data visualize the

578
00:20:32,080 --> 00:20:34,720
pass map

579
00:20:33,280 --> 00:20:36,559
and

580
00:20:34,720 --> 00:20:37,520
as we saw in so this is the events data

581
00:20:36,559 --> 00:20:38,799
set

582
00:20:37,520 --> 00:20:40,720
and

583
00:20:38,799 --> 00:20:42,240
so we'll

584
00:20:40,720 --> 00:20:44,480
like take the

585
00:20:42,240 --> 00:20:46,240
column stream type minute location

586
00:20:44,480 --> 00:20:47,520
password location pass outcome and

587
00:20:46,240 --> 00:20:49,760
player

588
00:20:47,520 --> 00:20:51,360
okay so here we will

589
00:20:49,760 --> 00:20:53,280
focus on a particular player and let's

590
00:20:51,360 --> 00:20:54,159
choose tony cruz because he is one of

591
00:20:53,280 --> 00:20:56,720
the

592
00:20:54,159 --> 00:20:58,480
best midfield builders of this uh era

593
00:20:56,720 --> 00:21:00,960
from events pass

594
00:20:58,480 --> 00:21:03,679
we only take the player whose name is

595
00:21:00,960 --> 00:21:06,000
tonic rose and

596
00:21:03,679 --> 00:21:08,320
this is what we end up with the data set

597
00:21:06,000 --> 00:21:10,880
like the type column in events past p1

598
00:21:08,320 --> 00:21:13,360
has event types other than passes but

599
00:21:10,880 --> 00:21:16,880
we'll only set to pass

600
00:21:13,360 --> 00:21:19,520
and that's what we do here and this is

601
00:21:16,880 --> 00:21:20,559
the data set where the type is only pass

602
00:21:19,520 --> 00:21:23,520
and

603
00:21:20,559 --> 00:21:26,960
like this is what the

604
00:21:23,520 --> 00:21:29,760
uh pass map looks for uh for tony close

605
00:21:26,960 --> 00:21:32,159
so if the pitch function and like for

606
00:21:29,760 --> 00:21:34,640
successful process we set the color

607
00:21:32,159 --> 00:21:36,240
green and for the ancestor process we

608
00:21:34,640 --> 00:21:37,440
for the unsuccessful process we set the

609
00:21:36,240 --> 00:21:39,280
color red

610
00:21:37,440 --> 00:21:41,760
and

611
00:21:39,280 --> 00:21:43,840
for the heat map we use the seaboard

612
00:21:41,760 --> 00:21:46,640
schedule plot

613
00:21:43,840 --> 00:21:50,320
yeah again i got this idea from uh

614
00:21:46,640 --> 00:21:52,640
like uh mickey jones tutorial and like

615
00:21:50,320 --> 00:21:53,840
if you

616
00:21:52,640 --> 00:21:55,039
want you can also calculate the

617
00:21:53,840 --> 00:21:57,360
percentage of successful and

618
00:21:55,039 --> 00:22:00,240
unsuccessful process for two nicoles and

619
00:21:57,360 --> 00:22:02,559
we see that he had around 91 percent

620
00:22:00,240 --> 00:22:04,640
successful passes so these are if you

621
00:22:02,559 --> 00:22:07,280
want you can draw the uh frequency

622
00:22:04,640 --> 00:22:08,880
distribution so now we will focus on

623
00:22:07,280 --> 00:22:11,039
applying some computational geometric

624
00:22:08,880 --> 00:22:12,559
concepts on

625
00:22:11,039 --> 00:22:14,320
uh the

626
00:22:12,559 --> 00:22:16,080
event and tracking data

627
00:22:14,320 --> 00:22:17,919
so for that we'll visualize the convex

628
00:22:16,080 --> 00:22:19,600
cells from players even data

629
00:22:17,919 --> 00:22:21,600
so we'll study how to develop a convex

630
00:22:19,600 --> 00:22:23,440
all around these points uh and from

631
00:22:21,600 --> 00:22:25,679
where a player had made a pass or had

632
00:22:23,440 --> 00:22:26,960
taken a shot in a particular game so

633
00:22:25,679 --> 00:22:29,440
mathematically if these points are

634
00:22:26,960 --> 00:22:31,440
contained in a set x then the convex l

635
00:22:29,440 --> 00:22:33,520
is the smallest convex set that contains

636
00:22:31,440 --> 00:22:36,000
x uh so this figure has been adapted

637
00:22:33,520 --> 00:22:37,120
from this wikipedia article uh if you go

638
00:22:36,000 --> 00:22:38,840
and you will see

639
00:22:37,120 --> 00:22:42,240
it has a lot of information about convex

640
00:22:38,840 --> 00:22:43,919
cells so we'll use uh the sci-fi package

641
00:22:42,240 --> 00:22:45,600
uh

642
00:22:43,919 --> 00:22:47,600
go download it

643
00:22:45,600 --> 00:22:49,120
uh so as we have been doing till now let

644
00:22:47,600 --> 00:22:50,480
us pick the important columns from the

645
00:22:49,120 --> 00:22:53,840
events data set

646
00:22:50,480 --> 00:22:56,240
like team location type and player

647
00:22:53,840 --> 00:22:58,159
here we only need the location of the

648
00:22:56,240 --> 00:23:00,720
players so it seems like we only need

649
00:22:58,159 --> 00:23:03,280
our four columns for now uh

650
00:23:00,720 --> 00:23:05,840
and we are focusing on pass and short

651
00:23:03,280 --> 00:23:08,640
events okay so we set the type to be

652
00:23:05,840 --> 00:23:10,400
either pass or short so that's what we

653
00:23:08,640 --> 00:23:12,080
do here then we'll next split the data

654
00:23:10,400 --> 00:23:13,679
into two data sets one for real media

655
00:23:12,080 --> 00:23:16,000
and one for liverpool we will now

656
00:23:13,679 --> 00:23:17,840
extract the even data for tony cruz uh

657
00:23:16,000 --> 00:23:20,559
from events all real

658
00:23:17,840 --> 00:23:22,159
uh so we set the player to be tony cross

659
00:23:20,559 --> 00:23:23,280
and this can be applied to all the other

660
00:23:22,159 --> 00:23:25,360
players

661
00:23:23,280 --> 00:23:26,159
and before computing and visualizing the

662
00:23:25,360 --> 00:23:28,240
past

663
00:23:26,159 --> 00:23:30,159
visualizing the con convex cell it is

664
00:23:28,240 --> 00:23:31,840
good practice to discard the outliers

665
00:23:30,159 --> 00:23:34,720
okay from the data sets

666
00:23:31,840 --> 00:23:36,240
so we use the interquartile range

667
00:23:34,720 --> 00:23:38,799
and we'll find the interquartile ranges

668
00:23:36,240 --> 00:23:40,559
for the columns location x location i

669
00:23:38,799 --> 00:23:42,159
from the event cell tony and then

670
00:23:40,559 --> 00:23:44,159
compute the upper and lower bounds for

671
00:23:42,159 --> 00:23:45,679
the data any points lying beyond these

672
00:23:44,159 --> 00:23:49,520
bounds will be considered as

673
00:23:45,679 --> 00:23:50,559
outliers and will be discarded okay

674
00:23:49,520 --> 00:23:52,240
so

675
00:23:50,559 --> 00:23:54,880
this is what we uh do here this is the

676
00:23:52,240 --> 00:23:56,000
box plot for uh tony cruz's uh location

677
00:23:54,880 --> 00:23:57,600
conditions

678
00:23:56,000 --> 00:24:00,159
and we see these are the so these are

679
00:23:57,600 --> 00:24:01,760
the whisker plots

680
00:24:00,159 --> 00:24:04,320
and for location x these are the

681
00:24:01,760 --> 00:24:05,039
outliers and for location why we don't

682
00:24:04,320 --> 00:24:07,440
have

683
00:24:05,039 --> 00:24:08,320
those outliers but yeah this

684
00:24:07,440 --> 00:24:11,360
uh

685
00:24:08,320 --> 00:24:13,440
we can discard the outliers uh

686
00:24:11,360 --> 00:24:15,120
and next let us look into the eventual

687
00:24:13,440 --> 00:24:16,559
tony data set

688
00:24:15,120 --> 00:24:19,200
and first we'll collect all the points

689
00:24:16,559 --> 00:24:21,120
from the two columns as a 2d matrix

690
00:24:19,200 --> 00:24:23,360
this uh comes in eight while drawing the

691
00:24:21,120 --> 00:24:25,279
convexal

692
00:24:23,360 --> 00:24:28,400
so poinsol equals to even solve only

693
00:24:25,279 --> 00:24:30,559
location x location y dot values give

694
00:24:28,400 --> 00:24:32,320
allows us to do that now let us use the

695
00:24:30,559 --> 00:24:34,880
convex hull function from scipy dot

696
00:24:32,320 --> 00:24:36,000
special uh

697
00:24:34,880 --> 00:24:36,960
package

698
00:24:36,000 --> 00:24:39,600
and

699
00:24:36,960 --> 00:24:41,520
we apply convex along this

700
00:24:39,600 --> 00:24:43,919
location action location

701
00:24:41,520 --> 00:24:46,320
y for events hall tony

702
00:24:43,919 --> 00:24:48,480
and then we just collect the useful

703
00:24:46,320 --> 00:24:50,799
information and we visualize the context

704
00:24:48,480 --> 00:24:52,240
so this is the convexel of tonic uses

705
00:24:50,799 --> 00:24:53,919
field coverage

706
00:24:52,240 --> 00:24:56,480
now we can draw the convex for other

707
00:24:53,919 --> 00:24:59,039
players too from either of the two and

708
00:24:56,480 --> 00:25:01,520
uh next what we do is that we uh draw

709
00:24:59,039 --> 00:25:03,600
the uh

710
00:25:01,520 --> 00:25:05,760
we learn how to get the tracking data

711
00:25:03,600 --> 00:25:08,480
for a particular match uh

712
00:25:05,760 --> 00:25:11,840
uh like for each for

713
00:25:08,480 --> 00:25:13,760
for some particular instances of a match

714
00:25:11,840 --> 00:25:15,279
and then draw the corresponding

715
00:25:13,760 --> 00:25:17,279
delineate triangulations and voronoi

716
00:25:15,279 --> 00:25:18,960
diagrams

717
00:25:17,279 --> 00:25:20,640
so we'll try to understand how to get

718
00:25:18,960 --> 00:25:22,880
the tracking data from particular game

719
00:25:20,640 --> 00:25:24,159
using stratform api so again it's given

720
00:25:22,880 --> 00:25:25,919
in their uh

721
00:25:24,159 --> 00:25:27,120
documentation please go and follow that

722
00:25:25,919 --> 00:25:29,919
if you want

723
00:25:27,120 --> 00:25:32,159
so we need to import useful classes from

724
00:25:29,919 --> 00:25:35,120
the mpl soccer dot strasbourn module and

725
00:25:32,159 --> 00:25:36,480
as you remember we use the match id18245

726
00:25:35,120 --> 00:25:39,360
and

727
00:25:36,480 --> 00:25:41,600
uh from mpl soccer.transform we import

728
00:25:39,360 --> 00:25:43,279
read event event

729
00:25:41,600 --> 00:25:45,360
the code is given here this is the

730
00:25:43,279 --> 00:25:48,559
events and tracking look at the event

731
00:25:45,360 --> 00:25:50,080
and tracking data sets so if we look

732
00:25:48,559 --> 00:25:51,840
closely into the tracking data set we

733
00:25:50,080 --> 00:25:54,240
understand that the column id represents

734
00:25:51,840 --> 00:25:56,720
an uh unique id for a short freeze frame

735
00:25:54,240 --> 00:25:58,480
that is a particular instance of a match

736
00:25:56,720 --> 00:26:00,799
now we'll only extract the relevant

737
00:25:58,480 --> 00:26:03,120
columns id player name the locations x

738
00:26:00,799 --> 00:26:05,600
and y and the team

739
00:26:03,120 --> 00:26:08,559
uh now let us try collecting the jersey

740
00:26:05,600 --> 00:26:09,760
numbers uh we you will use a different

741
00:26:08,559 --> 00:26:12,159
and easier approach from the one we have

742
00:26:09,760 --> 00:26:13,279
done here the blog that i have written

743
00:26:12,159 --> 00:26:14,799
so

744
00:26:13,279 --> 00:26:16,480
uh yeah but

745
00:26:14,799 --> 00:26:19,279
we can follow this to get the plane

746
00:26:16,480 --> 00:26:20,960
information use the following command by

747
00:26:19,279 --> 00:26:22,960
passing the match id so player info goes

748
00:26:20,960 --> 00:26:24,480
to sb dot lineups match ad equals to the

749
00:26:22,960 --> 00:26:26,799
one that we have been using

750
00:26:24,480 --> 00:26:28,240
here is the data set we see that it has

751
00:26:26,799 --> 00:26:30,320
just a number that country the player

752
00:26:28,240 --> 00:26:31,919
belongs to player nickname player name

753
00:26:30,320 --> 00:26:33,679
and player right now let us select a

754
00:26:31,919 --> 00:26:35,440
particular id from taking data set

755
00:26:33,679 --> 00:26:36,240
representing an instance of a particular

756
00:26:35,440 --> 00:26:38,720
game

757
00:26:36,240 --> 00:26:40,240
and we will filter tracking by a id

758
00:26:38,720 --> 00:26:41,520
value um

759
00:26:40,240 --> 00:26:43,039
which will give us the information of

760
00:26:41,520 --> 00:26:45,039
the locations of the players on the page

761
00:26:43,039 --> 00:26:47,840
at that moment

762
00:26:45,039 --> 00:26:50,720
uh so these are all the unique ids for a

763
00:26:47,840 --> 00:26:52,960
particular snap from a game

764
00:26:50,720 --> 00:26:55,440
let us filter the data set now so we use

765
00:26:52,960 --> 00:26:57,760
this short id to be this we just select

766
00:26:55,440 --> 00:26:59,600
a particular value from the id column

767
00:26:57,760 --> 00:27:02,400
uh and then tracking filter because the

768
00:26:59,600 --> 00:27:04,080
tracking tracking id where the id is the

769
00:27:02,400 --> 00:27:07,760
short id

770
00:27:04,080 --> 00:27:10,960
and same for event and uh the data

771
00:27:07,760 --> 00:27:12,080
filtered data set looks like this

772
00:27:10,960 --> 00:27:13,600
okay

773
00:27:12,080 --> 00:27:15,360
so we will now compute the dilute

774
00:27:13,600 --> 00:27:16,880
triangulations from our team's player

775
00:27:15,360 --> 00:27:18,320
locations to get an idea about the

776
00:27:16,880 --> 00:27:19,440
possible links created among the

777
00:27:18,320 --> 00:27:20,880
teammates

778
00:27:19,440 --> 00:27:22,320
so delay right triangulations help us

779
00:27:20,880 --> 00:27:24,159
understand the links

780
00:27:22,320 --> 00:27:26,720
uh between teammates of a particular

781
00:27:24,159 --> 00:27:29,120
team go look into the wikipedia article

782
00:27:26,720 --> 00:27:30,720
from where it took this

783
00:27:29,120 --> 00:27:32,559
diagram uh to understand more about the

784
00:27:30,720 --> 00:27:35,440
illinois triangulations

785
00:27:32,559 --> 00:27:36,559
so we will import that illinois uh

786
00:27:35,440 --> 00:27:38,799
class from

787
00:27:36,559 --> 00:27:40,320
sci-fi dot special

788
00:27:38,799 --> 00:27:41,840
we visualize the triangulations and the

789
00:27:40,320 --> 00:27:43,120
player positions of the instance on that

790
00:27:41,840 --> 00:27:45,760
page

791
00:27:43,120 --> 00:27:48,880
and this is the code to do that uh go

792
00:27:45,760 --> 00:27:52,159
check it out and yeah so this is the

793
00:27:48,880 --> 00:27:54,399
dilenoid triangulation for real madrids

794
00:27:52,159 --> 00:27:56,799
so i've linked to the real madrid

795
00:27:54,399 --> 00:27:58,720
players

796
00:27:56,799 --> 00:28:00,159
and the red nodes indicate the locations

797
00:27:58,720 --> 00:28:02,799
of liverpool's players and the white

798
00:28:00,159 --> 00:28:04,399
notes indicate the real matrix and the

799
00:28:02,799 --> 00:28:05,919
black lines indicate the direct links

800
00:28:04,399 --> 00:28:07,600
between the players one particular team

801
00:28:05,919 --> 00:28:09,600
at a particular moment forming that

802
00:28:07,600 --> 00:28:11,760
illinois triangulations so the book

803
00:28:09,600 --> 00:28:14,080
soccer matrix uh in his book soccer

804
00:28:11,760 --> 00:28:16,159
matrix by dr sumter he mentions that

805
00:28:14,080 --> 00:28:17,840
these lines have two useful indications

806
00:28:16,159 --> 00:28:19,039
first they portray the availability of a

807
00:28:17,840 --> 00:28:20,799
player

808
00:28:19,039 --> 00:28:22,159
uh availability of passes among the

809
00:28:20,799 --> 00:28:24,080
players from a particular team and

810
00:28:22,159 --> 00:28:25,840
second they also indicate the no man's

811
00:28:24,080 --> 00:28:28,240
lands for players from the opposition

812
00:28:25,840 --> 00:28:30,240
team meaning that if an player if an

813
00:28:28,240 --> 00:28:32,000
opposition player is on one of these

814
00:28:30,240 --> 00:28:34,480
linking lines

815
00:28:32,000 --> 00:28:36,799
then they are at the disadvantage

816
00:28:34,480 --> 00:28:38,559
uh so

817
00:28:36,799 --> 00:28:40,640
and we can also draw the voronoi

818
00:28:38,559 --> 00:28:42,399
diagrams uh again look into the

819
00:28:40,640 --> 00:28:43,600
wikipedia article to learn more about

820
00:28:42,399 --> 00:28:46,240
boronics

821
00:28:43,600 --> 00:28:48,720
we use uh cypher specials boronoid and

822
00:28:46,240 --> 00:28:51,600
voronoi plot 2d classes to draw the

823
00:28:48,720 --> 00:28:53,440
veronica diagrams

824
00:28:51,600 --> 00:28:55,039
and this is the

825
00:28:53,440 --> 00:28:56,640
voronoi diagram

826
00:28:55,039 --> 00:28:59,120
and the voronoi diagrams give us the

827
00:28:56,640 --> 00:29:00,799
zones of each and every player

828
00:28:59,120 --> 00:29:02,960
on the pitch at a particular moment by

829
00:29:00,799 --> 00:29:05,039
breaking the pitch into distinct regions

830
00:29:02,960 --> 00:29:06,799
belonging to the players indicating the

831
00:29:05,039 --> 00:29:08,000
field coverage of each player at that

832
00:29:06,799 --> 00:29:10,240
moment

833
00:29:08,000 --> 00:29:12,640
so yes this completes our section and i

834
00:29:10,240 --> 00:29:14,640
think that's it for my presentation so

835
00:29:12,640 --> 00:29:15,440
these are the references go check them

836
00:29:14,640 --> 00:29:17,919
out

837
00:29:15,440 --> 00:29:19,919
at the end thank you wear masks get

838
00:29:17,919 --> 00:29:22,919
vaccinated and stay safe

839
00:29:19,919 --> 00:29:22,919
thanks

840
00:29:26,640 --> 00:29:28,720
you