A string of L and R characters is balanced when it has the same number of each. Given a balanced string, split it into the maximum number of balanced substrings. For "RLRRLLRLRL" the answer is 4: RL, RRLL, RL, RL.
Walk the string and keep a counter: +1 for R, -1 for L. Every time the counter returns to 0, the prefix since the last cut is balanced, so cut there and count it.
def balanced_split(s):
balance = 0
pieces = 0
for ch in s:
balance += 1 if ch == "R" else -1
if balance == 0:
pieces += 1
return piecesO(n) time and O(1) extra space.
Suppose an optimal split does not cut at the first point where the balance hits zero. Its first piece then contains that balanced prefix plus a remainder, and the remainder must also be balanced (its count difference is the piece's difference minus the prefix's, which is 0 minus 0). So you can split that first piece into two, which gives at least as many pieces. Repeating the argument shows the greedy cuts are optimal.
The part people stumble on is the exchange argument. Practise saying "any longer first piece can be split at the first zero without losing pieces" in one sentence before moving to code.
Disclosure: I work on Linkjob AI. If you want practice explaining solutions like this under time pressure, we build AI help for online coding assessments.