Velocity - Prediction techniques - Scrum Refined

As a scrum master, we always need to best estimate or predict on the amount of work which team can finish in upcoming sprint.
When we have already ran scrum with same team for 4-5 iterations, then we can consider the velocity of team while estimating work for future sprint. Sometime we can average total number of burned story points in number of sprints, to get an average velocity of team.

However, there are some holidays (national or public) in next iteration where everyone would be on leave. Or, someone from team is going on long sabbatical leave etc. we need to be more good in prediction on how much work a scrum team would be able to finish. 

• Ignore leaves. - if you going to planning for a year or so, then it would be feasible to consider holidays, however it would be best to acknowledge management that prediction is based on best of our knowledge. we may need to consider changes in plan later as we grow
• Adjust leaves to full team - Lets say, if we have team of 6 members and avg velocity in last 2 weeks sprint was 50 story points. Now we would like to predict velocity for next sprint where 1 public holiday and 2 person is on 1-1 week planned leaves, means total man working days = (10 days*6person)- (2person* 5 working days leave) -(5person* 1 public holiday ) =45 days

Predicted velocity = Avg. Velocity* (planned working days/ Avg. working days)

Predicted velocity for 45days = 50*(45/60) = 37.5 ~ 37 SP

• Adjust leaves based on skill - Let's say if 3 OSB developers out of 6 OSB developers are on leave for complete sprint. Also, 1 tester out of 3 testers is on leave for full sprint.

Since these are skill specific reduction in team, which ultimately will impact the through put of team in sprint. 

As 3 OSB developers out of 6 OSB developers on leave - means 50% reduction 

and 1 tester out of 3 testers is on leave - means 33% reduction

If we look at overall team, then 4 out of 9 member on leaves - means 44% reduction

Then in such case, velocity can be reduced in range from 33- 50% due to unavailability of resources. Here, we can consider larger proportion as an overall reduction for this sprint.