Two threads locking two mutexes in opposite order

threadA and threadB are launched concurrently (as shown) repeatedly, each run using fresh mutexes. What is the most accurate prediction across many runs of this program?

main.cpp
std::mutex mtx1, mtx2;

void threadA() {
    std::lock_guard<std::mutex> lockA(mtx1);
    std::this_thread::sleep_for(std::chrono::milliseconds(10));
    std::lock_guard<std::mutex> lockB(mtx2);
}

void threadB() {
    std::lock_guard<std::mutex> lockB(mtx2);
    std::this_thread::sleep_for(std::chrono::milliseconds(10));
    std::lock_guard<std::mutex> lockA(mtx1);
}

6 of 51 in Concurrency