That's just how random chance works sometimes. Even if you only won the flip something like 6 out of 24 times, there's almost a 1% chance of that happening. So if 1,000 people did 24 battles since the change, 8 of them would be expected to have only gone first 6 of the 24 times.
If your sample size is small, it's not that unreasonable to see only a 25% success rate on the coin flip. As time goes on, you will probably see it stabilize closer to 50%.
You could also take a look at some of the alternative counters people have been experimenting with so you are not completely reliant on a single coin flip.