In this post, James R. Wootton, University of Basel, shows the effectiveness of Error code correction using the 16 Qubits Processor of IBM.
What is interresting to me is the practical results obtained and how the experiments was conducted
See quote of the article below
If we try to store our bit using just a single qubit, the probability of an error over the timescale of the experiment can be between 10% and 20%. The goal of using error correction is to reduce this. The bigger the error correcting code, the less likely an error should be.
The smallest case I considered was a repetition code with 3 code qubits (with an additional two qubits, the ancillas, helping to make measurements). When storing a 1, the probability of an error is around 4%. This is already better than a single qubit.
The largest case I considered had 8 code qubits (plus 7 ancillas). The probability of an error in this case was 0.025%, which is much better than a single qubit.
Worth reading and experiments by yourself !