[Football] No crowds - advantage away side

Harry Wilson's tackle said:

People have been using this as an argument for playing at neutral grounds.

Nothing to see here.

It's a statistical quirk derived from a small sample size.

Click to expand...

bn1&bn3 Albion said:

Having quickly skimmed through the results it looks like most the away wins came from teams that were favourites.. You'll need more than 4 rounds of games to make any judgement.

Click to expand...

Harry Wilson's tackle said:

Measuring the home-field advantage of a team (in a league with balanced schedule) requires a determination of the number of opponents for which the result at home-field was better ( k 1 {\displaystyle k_{1}} k_{1}), same ( k 0 {\displaystyle k_{0}} k_{0}), and worse ( k − 1 {\displaystyle k_{-1}} k_{{-1}}). Goals scored and conceded – in so called combined measure of home team advantage – are used to determine which results are better, same, and which are worse. Given two results between teams T 1 {\displaystyle T_{1}} T_{1} and T 2 {\displaystyle T_{2}} T_{2}, h T 1 : a T 2 {\displaystyle h_{T_{1}}:a_{T_{2}}} {\displaystyle h_{T_{1}}:a_{T_{2}}} played at T 1 {\displaystyle T_{1}} T_{1}'s field and h T 2 : a T 1 {\displaystyle h_{T_{2}}:a_{T_{1}}} {\displaystyle h_{T_{2}}:a_{T_{1}}}played at T 2 {\displaystyle T_{2}} T_{2}'s field, we can compute differences in scores (e.g. from T 1 {\displaystyle T_{1}} T_{1}'s point of view): d h , T 1 = h T 1 − a T 2 {\displaystyle d_{h,T_{1}}=h_{T_{1}}-a_{T_{2}}} {\displaystyle d_{h,T_{1}}=h_{T_{1}}-a_{T_{2}}}and d a , T 1 = a T 1 − h T 2 {\displaystyle d_{a,T_{1}}=a_{T_{1}}-h_{T_{2}}} {\displaystyle d_{a,T_{1}}=a_{T_{1}}-h_{T_{2}}}. Team T 1 {\displaystyle T_{1}} T_{1} played better at home field if d h , T 1 > d a , T 1 {\displaystyle d_{h,T_{1}}>d_{a,T_{1}}} {\displaystyle d_{h,T_{1}}>d_{a,T_{1}}}, and T 1 {\displaystyle T_{1}} T_{1} played better at away field if d h , T 1 < d a , T 1 {\displaystyle d_{h,T_{1}}<d_{a,T_{1}}} {\displaystyle d_{h,T_{1}}<d_{a,T_{1}}} (for example, if Arsenal won 3–1 at home against Chelsea, i.e. d h , A r s e n a l = 2 {\displaystyle d_{h,Arsenal}=2} {\displaystyle d_{h,Arsenal}=2}, and Arsenal won 3–0 at Chelsea, i.e. d a , A r s e n a l = 3 {\displaystyle d_{a,Arsenal}=3} {\displaystyle d_{a,Arsenal}=3}, then the result for Arsenal at home was worse). Same approach has to be used for all opponents in one season to obtain k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}}.

Values of k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}} are used to estimate probabilities as p ^ r = k r + 1 K + 3 , r = − 1 , 0 , 1 {\displaystyle {\hat {p}}_{r}={\frac {k_{r}+1}{K+3}},r=-1,0,1} {\displaystyle {\hat {p}}_{r}={\frac {k_{r}+1}{K+3}},r=-1,0,1}, where K {\displaystyle K} K is total number of opponents in a league (this is Bayesian estimator). To test hypothesis that home-field advantage is statistically significant we can compute P ( p 1 > p − 1 ) = 1 − I 1 / 2 ( k 1 + 1 , k − 1 + 1 ) {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(k_{1}+1,k_{-1}+1)} {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(k_{1}+1,k_{-1}+1)}, where I 1 / 2 ( ) {\displaystyle I_{1/2}()} {\displaystyle I_{1/2}()} is incomplete gamma function. For example, Newcastle in 2015/2016 English Premier League season recorded better result at home field for 13 opponents, same result with 4 opponents, and worse result for two opponents; therefore P ( p 1 > p − 1 ) = 1 − I 1 / 2 ( 14 , 3 ) = 0.998 {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(14,3)=0.998} {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(14,3)=0.998} and hypothesis about home team advantage can be accepted. This procedure was introduced and applied by Marek and Vávra (2017)[10] on English Premier League seasons 1992/1993 – 2015/2016.

Marek and Vávra (2018)[11] described procedure which allows to use observed counts of combined measure of home team advantage ( k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}}) in two leagues to be compared by the test for homogeneity of parallel samples (for the test see Rao (2002)[12]). The second proposed approach is based on distance between estimated probability description of home team advantage in two leagues ( p ^ r = k r K , r = − 1 , 0 , 1 {\displaystyle {\hat {p}}_{r}={\frac {k_{r}}{K}},r=-1,0,1} {\displaystyle {\hat {p}}_{r}={\frac {k_{r}}{K}},r=-1,0,1}) which can be measured by Jeffrey divergence (a symmetric version of Kullback–Leibler divergence). They tested five top level English football leagues and two top level Spanish leagues between 2007/2008 and 2016/2017 season. The main result is that home team advantage in Spain is stronger. Spanish La Liga has the strongest home team advantage and English football league two has the lowest home team advantage among analysed leagues.

Click to expand...

Icy Gull said:

Someone made the point that we groan quite a bit during home games when things aren’t going well so maybe an empty Amex will take away some of the pressure that moaning puts on players in home games. He had a point imo

Click to expand...

Mouldy Boots said:

I agree we have some grumpy home fans for years, the only time I remember them being grumpy at the goldstone was over Clive Walker.
The away fans always seem to have a more positive attitude imo.

Click to expand...

Harry Wilson's tackle said:

<sigh>

Measuring the home-field advantage of a team (in a league with balanced schedule) requires a determination of the number of opponents for which the result at home-field was better ( k 1 {\displaystyle k_{1}} k_{1}), same ( k 0 {\displaystyle k_{0}} k_{0}), and worse ( k − 1 {\displaystyle k_{-1}} k_{{-1}}). Goals scored and conceded – in so called combined measure of home team advantage – are used to determine which results are better, same, and which are worse. Given two results between teams T 1 {\displaystyle T_{1}} T_{1} and T 2 {\displaystyle T_{2}} T_{2}, h T 1 : a T 2 {\displaystyle h_{T_{1}}:a_{T_{2}}} {\displaystyle h_{T_{1}}:a_{T_{2}}} played at T 1 {\displaystyle T_{1}} T_{1}'s field and h T 2 : a T 1 {\displaystyle h_{T_{2}}:a_{T_{1}}} {\displaystyle h_{T_{2}}:a_{T_{1}}}played at T 2 {\displaystyle T_{2}} T_{2}'s field, we can compute differences in scores (e.g. from T 1 {\displaystyle T_{1}} T_{1}'s point of view): d h , T 1 = h T 1 − a T 2 {\displaystyle d_{h,T_{1}}=h_{T_{1}}-a_{T_{2}}} {\displaystyle d_{h,T_{1}}=h_{T_{1}}-a_{T_{2}}}and d a , T 1 = a T 1 − h T 2 {\displaystyle d_{a,T_{1}}=a_{T_{1}}-h_{T_{2}}} {\displaystyle d_{a,T_{1}}=a_{T_{1}}-h_{T_{2}}}. Team T 1 {\displaystyle T_{1}} T_{1} played better at home field if d h , T 1 > d a , T 1 {\displaystyle d_{h,T_{1}}>d_{a,T_{1}}} {\displaystyle d_{h,T_{1}}>d_{a,T_{1}}}, and T 1 {\displaystyle T_{1}} T_{1} played better at away field if d h , T 1 < d a , T 1 {\displaystyle d_{h,T_{1}}<d_{a,T_{1}}} {\displaystyle d_{h,T_{1}}<d_{a,T_{1}}} (for example, if Arsenal won 3–1 at home against Chelsea, i.e. d h , A r s e n a l = 2 {\displaystyle d_{h,Arsenal}=2} {\displaystyle d_{h,Arsenal}=2}, and Arsenal won 3–0 at Chelsea, i.e. d a , A r s e n a l = 3 {\displaystyle d_{a,Arsenal}=3} {\displaystyle d_{a,Arsenal}=3}, then the result for Arsenal at home was worse). Same approach has to be used for all opponents in one season to obtain k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}}.

Values of k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}} are used to estimate probabilities as p ^ r = k r + 1 K + 3 , r = − 1 , 0 , 1 {\displaystyle {\hat {p}}_{r}={\frac {k_{r}+1}{K+3}},r=-1,0,1} {\displaystyle {\hat {p}}_{r}={\frac {k_{r}+1}{K+3}},r=-1,0,1}, where K {\displaystyle K} K is total number of opponents in a league (this is Bayesian estimator). To test hypothesis that home-field advantage is statistically significant we can compute P ( p 1 > p − 1 ) = 1 − I 1 / 2 ( k 1 + 1 , k − 1 + 1 ) {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(k_{1}+1,k_{-1}+1)} {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(k_{1}+1,k_{-1}+1)}, where I 1 / 2 ( ) {\displaystyle I_{1/2}()} {\displaystyle I_{1/2}()} is incomplete gamma function. For example, Newcastle in 2015/2016 English Premier League season recorded better result at home field for 13 opponents, same result with 4 opponents, and worse result for two opponents; therefore P ( p 1 > p − 1 ) = 1 − I 1 / 2 ( 14 , 3 ) = 0.998 {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(14,3)=0.998} {\displaystyle P(p_{1}>p_{-1})=1-I_{1/2}(14,3)=0.998} and hypothesis about home team advantage can be accepted. This procedure was introduced and applied by Marek and Vávra (2017)[10] on English Premier League seasons 1992/1993 – 2015/2016.

Marek and Vávra (2018)[11] described procedure which allows to use observed counts of combined measure of home team advantage ( k 1 {\displaystyle k_{1}} k_{1}, k 0 {\displaystyle k_{0}} k_{0}, and k − 1 {\displaystyle k_{-1}} k_{{-1}}) in two leagues to be compared by the test for homogeneity of parallel samples (for the test see Rao (2002)[12]). The second proposed approach is based on distance between estimated probability description of home team advantage in two leagues ( p ^ r = k r K , r = − 1 , 0 , 1 {\displaystyle {\hat {p}}_{r}={\frac {k_{r}}{K}},r=-1,0,1} {\displaystyle {\hat {p}}_{r}={\frac {k_{r}}{K}},r=-1,0,1}) which can be measured by Jeffrey divergence (a symmetric version of Kullback–Leibler divergence). They tested five top level English football leagues and two top level Spanish leagues between 2007/2008 and 2016/2017 season. The main result is that home team advantage in Spain is stronger. Spanish La Liga has the strongest home team advantage and English football league two has the lowest home team advantage among analysed leagues.

Click to expand...

GOM said:

Short term and small sample but early evidence from the Bundesliga is showing that playing in empty stadiums could favour the away team.

Now if this is true, how will this affect our run in?

https://www.theguardian.com/footbal...ue-restart-will-not-be-football-as-we-know-it

Click to expand...