In July I posted two articles about this concept of Expected Sideout. First, I introduced the concept:
And then, in this article, I fleshed it out a bit more and also introduced Expected First Ball Sideout as another possible metric to consider.
In this third article I’ll tie everything together and keep this as an evergreen reference for the topic as well.
Expected Sideout Tweet Summary
How To Calculate
This is for NCAA-level women’s volleyball. Good Passes = 3s or 4s on a 4-point scale. Medium Passes = 2s on a 3 or 4-point scale. Bad Passes = 1s on a 3 or 4-point scale. Shanked Passes means an overpass or shanked pass that leads to a freeball.
The 0.67, 0.60, 0.50, and 0.30 are approximate Sideout numbers on those quality of passes, respectively. Adjust as is relevant to your level.
Examples
The Lions receive 10 balls and win 5 of those rallies. They pass 4 Good Passes, 2 Medium Passes, 3 Bad Passes, and get Aced once.
Their Expected Sideout is:
and their Actual Sideout is 5 / 10 = 50%. They have “underperformed” their Expected Sideout, although by less than 1 real point.
The Tigers receive 10 balls. They pass 5 Good Passes, 2 Medium Passes, 1 Bad Pass, and get Aced twice.
Their Expected Sideout is:
And worth noting their 3-point passer rating is a 2.0 compared to the Lions’ 1.9 and their Good Pass % is 50% compared to the Lions’ 40%, but their Expected Sideout is lower because those metrics undercount the effect of getting Aced.
Expected First Ball Sideout Quick Summary Without An Obnoxious Self-Tweet-Screenshot
Some passing stats are deceptive. Evaluate passing based on correlation to First Ball Sideout, so that Transition performance doesn’t muddy the comparison.
How To Calculate (NCAA Women’s Level)
Examples
The Bears pass 10 balls and kill 4 of them in First Ball. They pass 5 Good Passes, 3 Mediums, 1 Bad, and get Aced once. Their xFBSO is:
The Bears had an actual FBSO of 40%, so they outperformed their passing.
Warning: YMMV on the Kill % by conference. I’d imagine there’s a pretty significant spread in Kill % by pass quality between the Big 10 and some smaller conferences, so check your numbers agaisnt conference averages.
Some Questions
Two recent messages I got from coaches using these calculations:
You have written about xFB on the blog but you never say how to calculate it. And how do you factor in aces for the server if it’s xFB for serving.
And
For expected FBSO reception, why is a passing error multiplied by -0.20 instead of just 0.00?
Well person-who-asked-the-first-question, there you go, see above.
But… that’s from a passing perspective and not a serving perspective. How do you calculate Expected Sideout or Expected First Ball Sideout for your servers?
Basically the same way, except you have to consider this fact:
When doing anything “expected” for your passers, you’re never considering Service Errors and when doing anything “expected” for your servers, you always are.
What this means is that when you calculate Expected Sideout for your passers, as shown above, you want to compare to actual Earned Sideout, not just Sideout, because if your opponents miss 12% of their serves, they are giving you a bunch of free Sideout that won’t show up in your Expected Sideout numbers. So you compare Expected Sideout to actual Earned Sideout, which is Sideout % only on balls that were served in.
Likewise, when calculating Expected First Ball Sideout for your passers, you compare to your First Ball Sideout on balls that were served in. Sometimes you might see this as First Ball Kill %.
Expected Sideout For Servers
Using DV codes to fit that all on one line. But it’s basically the same calculation while adding in the fact that a Serve Error always leads to a sideout.
Expected First Ball Sideout For Servers
Same basic deal. And remember that for both Expected Sideout and Expected First Ball Sideout, lower is better when you’re talking about your own servers.
And Now To Answer The Second Quoted Question
Expected First Ball Sideout treats, to use the DataVolley term, R/ the same as R=. You never First Ball Kill an overpass and you also never First Ball Kill a reception error. So, in theory, both coefficient values should be 0. But, clearly an ace is a worse result for the receiving team than an overpass. In women’s volleyball, you still win the rally almost 70% of the time on an overpass.
So, as the coach above noted, in my formulas for this stuff that are still floating around the DataVolleys, I added an additional term that subtracts an extra penalty for passers who get aced or bonuses servers who get aced:
This has the effect of:
Making your xFBSO number align slightly less with the comparison to actual
Making it a slightly better evaluator of server or passer quality.
So, if the primary reason you use xFBSO is to evaluate servers or passers, then keep that extra Ace coefficient in. If the primary reason you use it is to compare to actual to build an offense profile, exclude it.
Alright, I hope that was helpful. Which do you like better, Expected Sideout or Expected First Ball Sideout? What ways are you using this concept in your gym?