the known problem with average is the outliers
If you have extreme scores (outlier) it will distort the average number
let's see our player
livaja 19
juan 21
obi 21
frog 24
nagatomo 26
guarin 26
pereira 27
gargano 28
handanovic 28
palacio 30
mundigayi 31
chivu 32
rochi 35
zanetti 39
so for starter
19 21 24 26 26 27 28 28 30 32 39
N =11
mean : 27
list of deviaton -8 -6 -2 -1 -1 +1 +1 +3 +5 +12
square of deviation 64, 36, 4 ,1 ,1 ,1 ,1 , 9, 25, 144
sum of squared deviation : 286
divided by N-1 = 28.6
square root of 28.6 = 5.34
let's see whether JZ is outlier or not
G = (39-27) / 5.34 = 2.24
people usually assumes that "normal" data will all fall within around 2 standard deviations of the mean or referred as the 95% confidence interval
so 39 is has less than a 5% chance of being a true data point or more than 95% chance of being outliers
after subs come in
21 21 24 26 27 28 30 31 32 35 39
N=11
mean : 28
list of deviaton -7 -7 -4 -2 -1 +2 +3 +4 +7 +11
square of deviation 49, 49, 16 ,4 ,1 ,4 ,9 , 16, 49, 121
sum of squared deviation : 318
divided by N-1 = 31.8
square root of 31.8 = 5.63
let's see JZ
G = (39-28) / 5.63 = 1.95
this one seems better because jz is not an outlier
but as you can see it's because we put more older player
I don't think using average will be a good idea for this case as for starting line up we have outliers and for the after subs we have more older players
feel free to correct me if above calculation is wrong