Does time of shot affect affect xG?
A major decision for managers of smaller clubs, who can’t just blow opposition teams away, is whether to go out quick at the start and try and get an early lead, then defend it or to keep it tight, before throwing everything at the wall for the last period of the match to try and snatch it at the death. Both methods have their benefits; the early attack gives more time to get the goal, but there’s no definitive end point as to when to start shutting up shop if the goal can’t be found as it’s unlikely a team would be able to keep up the intensity for a whole match. The late attack gives you more defensive security, as you prioritise not conceding in the early stages but can backfire if your opposition is fitter and have more in the tank at the end.
What if it was possible to tell when is best to attack based purely on xG, if it could be shown that the time in the match affected how likely a shot was to be scored? It could help to decide when to attack to get the most out of it, or when to sit back as it’s less worth it.
There doesn’t seem to be many reasons why this might be the case, but it is entirely possible. The main reason behind it is that xG could be affected by a players psychological state and perhaps at different stages of a match a player may be more or less confident - leading them to be more or less likely to score. For example; the actual xG distribution for early shots could be much lower than for an average shot taken at any time, due to a player believing it is unlikely that a goal would be scored that early in the match.
To test this theory of whether time affects xG, I have split the game into 6 x 15 minute sections and 90+ and will compare the distribution of shots taken in these times to the overall distribution of all shots. I will be looking at the distribution of aG-xG as this will allow me to just look at how well the xG model is fitting the aG data for each section, whereas xG or aG by itself could be influenced by a certain time period having more high quality chances. I will use a Z-test to compare the two distributions because the shots from a specific time period are a sample of the overall shot population which we assume has the underlying distribution. If we were comparing two different times a T-test should be used.
So statsy bit:
Null hypothesis - time of shot doesn’t affect expected goal value Use a two tailed test with 10% significance - so for a result to disprove the null hypothesis would need a likelihood <0 .05="" or="">0.95
Only 1 of the p-value gives a significant result which disproves the null hypothesis. The mean value for difference between aG and xG is different enough at this significance level to suggest that the shots taken between the 15-30 minute have a different xG distribution to an average shot.
If the current xG model which doesn’t take into account when the shot is taken correctly models shots then there shouldn’t be a time period where aG-xG varies significantly from the average value. As there is it suggests that the xG model needs to take into account when the shot was taken, and as for the 15-30 minute range the average aG-xG is significantly less than the average value for all shots, it would mean that goals are scored less frequently in this time period and thus the xG would need to be decreased to reflect that.
This information could be useful in one of two ways either it could be used to instruct teams to focus more on not conceding and growing into the game in that time period or if it’s a problem of belief causing the dip then by pointing it out to your players you could get them to belief in themselves more and increase your (aG-xG) in that period back up to the average giving you an advantage over your opponent in that period as your shots are worth more.
What if it was possible to tell when is best to attack based purely on xG, if it could be shown that the time in the match affected how likely a shot was to be scored? It could help to decide when to attack to get the most out of it, or when to sit back as it’s less worth it.
There doesn’t seem to be many reasons why this might be the case, but it is entirely possible. The main reason behind it is that xG could be affected by a players psychological state and perhaps at different stages of a match a player may be more or less confident - leading them to be more or less likely to score. For example; the actual xG distribution for early shots could be much lower than for an average shot taken at any time, due to a player believing it is unlikely that a goal would be scored that early in the match.
To test this theory of whether time affects xG, I have split the game into 6 x 15 minute sections and 90+ and will compare the distribution of shots taken in these times to the overall distribution of all shots. I will be looking at the distribution of aG-xG as this will allow me to just look at how well the xG model is fitting the aG data for each section, whereas xG or aG by itself could be influenced by a certain time period having more high quality chances. I will use a Z-test to compare the two distributions because the shots from a specific time period are a sample of the overall shot population which we assume has the underlying distribution. If we were comparing two different times a T-test should be used.
So statsy bit:
Null hypothesis - time of shot doesn’t affect expected goal value Use a two tailed test with 10% significance - so for a result to disprove the null hypothesis would need a likelihood <0 .05="" or="">0.95
| Minute Range | Z | p-value |
|---|---|---|
| 0-15 | -0.214854675976967 | 0.41683 |
| 15-30 | 0.774852991191252 | 0.77935 |
| 30-45 | -1.82306785234072 | 0.03438 |
| 45-60 | 1.00758682665302 | 0.84375 |
| 60-75 | 1.44903000180933 | 0.92647 |
| 75-90 | 0.446454022736825 | 0.67364 |
| 90+ | -0.409549308873886 | 0.34090 |
Only 1 of the p-value gives a significant result which disproves the null hypothesis. The mean value for difference between aG and xG is different enough at this significance level to suggest that the shots taken between the 15-30 minute have a different xG distribution to an average shot.
If the current xG model which doesn’t take into account when the shot is taken correctly models shots then there shouldn’t be a time period where aG-xG varies significantly from the average value. As there is it suggests that the xG model needs to take into account when the shot was taken, and as for the 15-30 minute range the average aG-xG is significantly less than the average value for all shots, it would mean that goals are scored less frequently in this time period and thus the xG would need to be decreased to reflect that.
This information could be useful in one of two ways either it could be used to instruct teams to focus more on not conceding and growing into the game in that time period or if it’s a problem of belief causing the dip then by pointing it out to your players you could get them to belief in themselves more and increase your (aG-xG) in that period back up to the average giving you an advantage over your opponent in that period as your shots are worth more.
Comments
Post a Comment