Point Distribution for Set Wins

Something intersting in Moneyball: The Art of Winning an Unfair Game is the idea that winning X number of games should put the team in the playoffs. I don't recall the specific number, and it will be highly specific to American major league baseball, but the idea in interesting to me. I think it is one that might have application in volleyball.

Some years ago I saw a coaching video by an NCAA Division I women's volleyball coach saying that on average W out of 25 points come from kills, X from blocks, Y from aces, and Z from opponent errors. I don't recall the exact numbers, but I believe the number of kills was around 16 or 17. I hadn't thought about it that much in the meantime, but Moneyball jostled things a bit in my brain. Because I don't remember the specific numbers, and I think the numbers will vary by competitive level, I decided to collect some data to see what these numbers are.

I'm starting with a spreadsheet to collect the data. It's pretty basic. It has columns for kills, blocks, aces, opponent errors, and just for comparison team errors. I'm just collecting that information for the winning team of the set. At the top of the column is the average points per set. At first I was just including sets that ended at 25 so the math would be simple, but I later decided that more data would give a better average. To incorporate more data, but still keep the data relevant and easy to parse and use, I included the score for the winning team in a new column, divided the sum of all kills, blocks, etc by the sum of total points, and multiplied by 25 to get the average kills, blocks, etc for a 25 point set win. I copied the sheet so there is one for FIVB men, FIVB women, NCAA men, NCAA women, high school boys, and high school girls. Thus far I have only entered data for 4 matches from this year's FIVB World League finals. They are USA/Serbia, France/Poland, USA/Poland, and Serbia/France. The FIVB matches have a handy end of set graphic that includes all this data, so it makes gathering the data relatively quick. I don't need to watch a whole match with a clipboard to get it.

The Numbers So Far
4 matches from 2015 FIVB World League

So in the average FIVB World League finals match, the winning team scored 13 points on offense, almost 3 on blocks, and 1 or 2 aces every set. They scored almost 8 points on opponent errors. Assuming a good distribution of sets, the two outsides and the opposite are looking at 3 kills per set each, and the two middles will get about one or two kills per set.

A few things stood out to me. First the number of kills is lower than I thought. I thought that at the level of play going on at World League the number of kills would be closer to 15. In a couple there were 17 and 18, but for the most part they were floating around 10-14. Even the longest set in the data so far with a score of 27 points, the kills are only 14. I thought blocks and aces should be higher as well. Blocking and serving is stronger than what I expect to see at the college level, but I guess covering hitters and serve receive are comparable to the blocking and serving. Opponent errors are higher than I thought it would be, but I think that is probably due to errors being an inclusive category for service and hitting errors.

Comparing team errors to opponent errors is interesting. I don't have anything handy that shows what's happening visually, but there were 5 sets where the winners had more errors than the losers. There were another 3 where the errors were the same. Out of a total of 16 sets, that leaves 8 where the winners have fewer errors than the losers. Conventional wisdom is that you should have fewer errors than your opponent. Yes, half the time that has happened, but almost 33% of the time the losing team committed fewer errors. The two highest kill sets (17 and 18) were two of the sets where the winners had more errors than the losers. Perhaps the more important comparison is between team kills and team errors than the comparison between team and opponent errors.

It will be interesting to see how the point distribution numbers change with more data and at other competitive levels. From what I have seen it looks like a goal of 13+ kills per set and a good ratio of kills to errors will win matches. I'm still debating whether I should go back and include all opponent data. When I collect data for matches that I have to watch and tally the points I will collect it by default. I just don't know if it should end up on the spreadsheet. We'll see. More to come.