Using a sliding window to identify and return the first success

I have and experiment where I am looking at bird visits to feeders. If they go to the correct feeder, it counts as a success. They are deemed to have learnt which feeder to visit once they are successful 16/20 visits (80%).

I want to write a piece of R code that goes through the full data frame by bird and identifies the first instance of 80% success with a moving window of 20 visits.

The code I have below works but it is clumsy. The end result is to obtain a tibble with each row representing a unique bird and the point at which they first reach 80% success.

# My attempt
library(tidyverse)

# Some sample data
data <- tibble(
  bird = rep("a121", 99), 
  success = sample(c(1,0), 99, replace = TRUE, prob = c(0.7, 0.3)),
  visit = 1:99) %>%
  bind_rows(tibble(
    bird = rep("b232", 99), 
    success = sample(c(1,0), 99, replace = TRUE, prob = c(0.75, 0.15)),
    visit = 1:99)) %>%
  bind_rows(tibble(
    bird = rep("c211", 99), 
    success = sample(c(1,0), 99, replace = TRUE, prob = c(0.7, 0.3)),
    visit = 1:99)) %>%
  mutate(observation = 1:297) %>%
  select(observation, everything())

# Identify first time 80% success rate met with a sliding window of 20

res <- NULL 

slide.funct <- function(data, window, step){ 
  birds <- unique(pull(data, bird)) # Identify the birds in the data
  for (j in birds) {
    sub_data <- filter(data, bird == j) # Work with one bird at a time

    # Might need a line to ensure the data is ordered by time

    total <- nrow(sub_data)
    spots <- seq(from = 1, to = (total - window), by = step)
    result <- vector(length = length(spots))
    for(i in 1:length(spots)){
      result[i] <- mean(sub_data$success[spots[i]:(spots[i] + window)]) # Success rate by window
      position <- Position(function(x) x >= 0.8 , result) + 19 # First point at which success > 80%
    }
    res <- bind_rows(res, sub_data[position, ]) # Build up info for each bird
  }
  res <- res %>% mutate(observation = observation + 19) %>% 
    select(observation, bird)
  return(res)
}

slide.funct(data, 19, 1)

#> # A tibble: 3 x 2
#>   observation bird 
#>         <dbl> <chr>
#> 1          53 a121 
#> 2         138 b232 
#> 3         237 c211

This is only the first stage of the process as there is more to do downstream. However, I'm not sure the above function robust enough. I looked at the 'zoo' package but didn't think it could help here. I'm also looking to use tidyverse code where I can, so if there are better options with respect to achieving the above, I'd be very happy to hear them.