Rolling median by group

I know variations of this question have been asked and answered 1,000 times. I've read every post and tried to implement what I read without success.

I have a dataframe of basketball individual game statistics for multiple players. I want to calculate 1) the "season to date" rolling median for each player (excluding the current observation) and 2) The median for he last 3 previous games (excluding the current observation)

I've included samples of what I would expect the output to be in each case.

Here's a sample of the data:

# Create Dataframe
plyr_seas = c("J.J. Redick", "Andre Drummond", "Daniel Theis", "Andre Drummond", "Daniel Theis", "J.J. Redick", "Andre Drummond", "J.J. Redick", "Andre Drummond", "J.J. Redick", "Daniel Theis", "Andre Drummond", "J.J. Redick", "Daniel Theis")

plyrOREBpmin_gm = c(0, 0.1715686275, 0.1013513514, 0.02721088435, 0.2002224694, 0, 0.1217861976, 0.03552397869, 0.1044386423, 0.03984063745, 0.03573555688, 0.2264150943, 0, 0.03719776813)

plyrGmsPlydpcs = c(0, 0, 0, 1, 1, 1, 2, 2, 3, 3, 2, 4, 4, 3)

expected_last_3_median = c(NA, NA, NA, NA, NA, NA, NA, NA, 0.1217861976, 0, NA, 0.1044386423, 0.03552397869, 0.1013513514)

expected_overall_median = c(NA, NA, NA, 0.1715686275, 0.1013513514, 0, 0.0993897559, 0, 0.1217861976, 0, 0.1507869104, 0.1131124199, 0.01776198934, 0.1013513514)

Pbox = data.frame(plyr_seas, plyrOREBpmin_gm, plyrGmsPlydpcs, expected_last_3_median, expected_overall_median)

SOLUTION for above data! Provided by: G. Grothendieck

# Calc rolling median for last three games (excluding current)
Pbox = Pbox %>% 
 group_by(plyr_seas) %>% 
 mutate(plyrOREBrateMdnLst3 = 
   rollapplyr(plyrOREBpmin_gm, list(-seq(3)), median, fill = NA))

# Calc rolling median for season to date (excluding current)
Pbox = Pbox %>% 
  group_by(plyr_seas) %>% 
  mutate(plyrOREBrateMdnSeas = 
    rollapplyr(plyrOREBpmin_gm, seq_along(plyrGmsPlydpcs), 
      function(x) median(head(x, -1)), fill = NA))