Understanding model loss/accuracy and how not to leak information - tensorflow

Question

This question is related to the starting one posted here.
The problem is to classify rows so that the classification of row number i can rely on the data for all the previous rows including class membership. The linked post contains an answer which is posted bellow.
For the sake of experimentation I've used a set of randomly crafted data, where the classifying property is a 0,1 uniform random variable.
What strikes me is that the loss of the model in the above example is really low and the accuracy is 99% whereas I would expect something in the 50% range.
So I am assuming that the way the model is testing the classification is leaking information somehow.
Does anybody happen to see what's the issue? What would be the proper way to evaluate the accuracy in such scenario?
import tensorflow as tf
import pandas as pd
import numpy as np
from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
from random import randint
SIZE = 100
df = pd.DataFrame({'Temperature': list(range(SIZE)),
'Weight': [randint(1,100) for _ in range(SIZE)],
'Size': [randint(1,10000) for _ in range(SIZE)],
'Property': [randint(0,1) for _ in range(SIZE)]})
df.Property = df.Property.shift(-1)
print ( df.head() )
# parameters
time_steps = 1
inputs = 3
outputs = 2
df = df.iloc[:-1,:]
df = df.values
train_X = df[:, :-1]
train_y = df[:, -1]
scaler = MinMaxScaler(feature_range=(0, 1))
train_X = scaler.fit_transform(train_X)
train_X = train_X[:,None,:]
onehot_encoder = OneHotEncoder()
encode_categorical = train_y.reshape(len(train_y), 1)
train_y = onehot_encoder.fit_transform(encode_categorical).toarray()
learning_rate = 0.001
epochs = 50000
batch_size = int(train_X.shape[0]/2)
length = train_X.shape[0]
display = 100
neurons = 100
tf.reset_default_graph()
X = tf.placeholder(tf.float32, [None, time_steps, inputs])
y = tf.placeholder(tf.float32, [None, outputs])
cell = tf.contrib.rnn.BasicLSTMCell(num_units=neurons, activation=tf.nn.relu)
cell_outputs, states = tf.nn.dynamic_rnn(cell, X, dtype=tf.float32)
stacked_outputs = tf.reshape(cell_outputs, [-1, neurons])
out = tf.layers.dense(inputs=stacked_outputs, units=outputs)
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(
labels=y, logits=out))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
training_op = optimizer.minimize(loss)
accuracy = tf.metrics.accuracy(labels = tf.argmax(y, 1),
predictions = tf.argmax(out, 1),
name = "accuracy")
precision = tf.metrics.precision(labels=tf.argmax(y, 1),
predictions=tf.argmax(out, 1),
name="precision")
recall = tf.metrics.recall(labels=tf.argmax(y, 1),
predictions=tf.argmax(out, 1),
name="recall")
f1 = 2 * accuracy[1] * recall[1] / ( precision[1] + recall[1] )
with tf.Session() as sess:
tf.global_variables_initializer().run()
tf.local_variables_initializer().run()
for steps in range(epochs):
mini_batch = zip(range(0, length, batch_size),
range(batch_size, length+1, batch_size))
for (start, end) in mini_batch:
sess.run(training_op, feed_dict = {X: train_X[start:end,:,:],
y: train_y[start:end,:]})
if (steps+1) % display == 0:
loss_fn = loss.eval(feed_dict = {X: train_X, y: train_y})
print('Step: {} \tTraining loss: {}'.format((steps+1), loss_fn))
acc, prec, recall, f1 = sess.run([accuracy, precision, recall, f1],
feed_dict = {X: train_X, y: train_y})
print('\nEvaluation on training set')
print('Accuracy:', acc[1])
print('Precision:', prec[1])
print('Recall:', recall[1])
print('F1 score:', f1)