Simulating a dice throw with threadlocalrandom does not converge to the expected value

I was trying to verify the expected value formula in java.
I did something very trivial, I simulated the roll of a dice, but I mapped each side to a specific value. So each value has probability of 1/6.
The expected value was determined as: 232.65
I run the dice simulation a very long number of runs, 10000000 to be precise but the number did not converge to 232.65 but was 232.74. Smaller runs also were fluxuating up to the value 233.97. Runs up to 2100000000gave 232.63.
So I am wondering if I am doing something wrong. My main premise here is that I shouldn't need to simulate 2 billion throws of a dice to eventually the expected value calculated via the formula.
So I am wondering if I am using the wrong random API. I can not get the seed so I can't verify if that changes per iteration. My code is the following:

Map<Integer, Float> map = new HashMap<>();
map.put(1, 17.2f);
map.put(2, 11f);
map.put(3, 128f);
map.put(4, 1f);
map.put(5, 1200f);
map.put(6, 38.7f);

double sum = 0;
int count = 0;

for(Map.Entry<Integer, Float> entry: map.entrySet()) {
     sum += (entry.getValue() * (1/6f));
}

System.out.println("EV " + sum);
sum = 0;

for(int j = 0; j < 2100000000; j++) {
    int dice = ThreadLocalRandom.current().nextInt(1, 7);
    sum += map.get(dice);
    ++count;
}
System.out.println(sum / count);

Output:

EV 232.65000109374523
232.63201358102154

I was expected after a much earlier number of runs I would consistently get something ~232.650...