.png)
Looking Into Weighted Stats
Compiling Season Stats
For for any I collect stats for, I would often compile all the match stats throughout the season and see how the team and players are performing statistically. It gives you an idea of where the strengths and weaknesses lie within the team and the players themselves and it gives you a basis of how you can utilise those strengths (along your observation and knowledge as a coach). For example, I once had a decent City University London outside hitter that I put on the bench because I had two that I thought was better. As the stats rolled in, I saw she was a better player than I had originally thought. So backed up by the stats, I put her in the starting line-up. She ended up being my best passer by the end of the season (***Update 15 January 2021 | Katerina is actually ranked the third best passer on the all time statistical list)
However, one thing I have to take into consideration when compiling these season stats is who we have played. If we have played a number of lower level teams, those match stats will invariably be higher and this would over-inflate the overall season stats. There’s a danger I may choose a player over another to start a match based on over-inflated season stats. This is especially the case with City University London where difference in playing levels can be a chasm.
To combat this issue, I would print an addition set of season stats – one that is against tough teams only. I would define tough teams as any team where we had to work hard to win the match. This would take away the effect of overinflated stats against easy teams. I would look at both sets of stats and go from there. Below is an example of what I have done previously before an important match against an tough opponenet.
Issues With Seasons Stats According to Opposition Level
However, there are a number of issues when you look at stats in this way. The first is there is a lot of numbers to look at and which numbers do you use or trust, especially if you are going to make changes based on the numbers. The stats against tough teams has the advantage of excluding the over-inflated stats from the easier matches. It is also probably truer to how the team and players will perform against tough teams. However, using the stats against the tough teams has the disadvantage that you are excluding a lot of data. Limited data is volatile so more data is always better to establise a stabilised performance level. This is more problematic if you compile the season stats early in the season because you have limited data to draw upon in the first place (you always have new players on the team and you want to know what level they are playing at).
Using Weighted Statistics
So one way to overcome these issues is to have weighted statistics. It enables you to give different weights to the statistics according to the opposition level. So 1 kill against a tough team is worth more than a kill against a lower level team. We already have weighted statistics when we calculate the reception rating:
You can see the weight we give to perfect passes is 3, good passes is 2, poor passes is 1 and reception errors is 0. So when we when we take account all these weights, the rating ranges from 0.00 (all errors) and 3.00 (all perfect passes).
We can use a similar process to weight the other statistics. However, the question is how much weight should we give according to the opposition level? So we can give a weight like 1.25 against a higher level team, 1.00 against a similar level team and 0.75 against a lower level team. So 1 kill against a higher level team is worth 1.25. I don’t like this idea because this is quite arbiturary. Why not give a 1.50 weight against a higher level team? Or a weight of 10? How would you define what a higher, similar or lower level team is? Why is there only three opposition levels? Why not five or even more?
So I wanted a system that would give a better indication of an opposition team’s level. The idea eventually popped into the head that I could use their Points Efficiency Ratio (PER) as the best way to determine opposition level. It’s closely linked to success in volleyball and the actual number is a nice multiplier. Great teams will have a PER of around 1.50 so 1 kill against a good team is now worth 1.5. Poor teams will have a PER of around 0.50 so 1 kill against a poor team is now worth 0.50.
Calculating a Weighted Statistic
The process to calculate the weighted statistics is as follows: separate the season statistics by match, multiply each match statistics by the opposition PER in that match, sum up each the resulting weighted statistics and then recalculate the performance indicator. I did this for the team and repeated the process for each individual player. Below is how it looks like for team’s points scoring.
You can see that the weighted points scored and weighted points lost are not very useful. To say the team scored a weighted 22.03 points doesn’t mean much to anybody. The weighted points efficiency ratio for each individual match is exactly the same as the unweighted points efficiency ratio. This is because it is not exactly weighted – it’s more like a scaled. If you scale up a square to double its size, the sides of the square is still equal. So the only useful number is the weighted points efficiency ratio for all the matches.
The End Result
In the above, I’ve added the weighted statistics to the actual statistics. You can see the weighted statistics is generally slightly lower than the actual statistics. You would expect this because the statistics against the lower level teams, which are normally higher, have less weight and the statistics against the higher level teams, with is normally lower, has more weight. If a player has a higher weighed statistic, it would generally mean they played better against the better teams.
Is Weighted Statistics Worth It?
Right now, the answer is probably no. When I first created the spreadsheet in the last Christmas break, it took a lot of man hours do to and the result was much of the weighted statistics are very similar to the actual statistics. I didn’t see anything in the weighted statistics that would make me think I could make a change to exploit the numbers that the actual statistics didn’t already indicate. I also had the same issue of which number should you use and trust when making decisions to enhance performance. I may repeat the same process for the Polonia teams to see if something different happens but I suspect the a similar outcome will occur.
Comments