As mentioned in the question, these are all techniques of preventing a Deadlock prevention as mentioned in the question. Please note the term Lock and Resources are used interchangeably.
Setting, 10 resources in the system 1, 2, 3, 4.. 10 and all the threads in the system needs all the resources.
1. In this case, the thread start acquiring the lock in random order and if it fails then it retries again. Here CPU cycles are wasted as it might be the case that one thread as acquired the the locks till 9 and then some other thread starts acquiring lock from 10. In this case the earlier thread which was almost done has to retry again and all the work done for acquiring 9 locks is wasted.
2. In this case the thread follows patter that it starts acquiring lock in increasing order. Here if a thread gets lock on resource 1, then it is sure to get all the locks ( This requires a condition of releasing order. Food for thought?).
3.This is same as case 2 here it start acquiring lock in decreasing order. Increasing or decreasing does not matter till all the threads follow some pattern. This will reduce wasted work ( might not even allow thread to do wasteful work).
4. This case is interesting. In current setting this is same as option 2 because every thread will start from Resource 1. This will be different when the requirements of each thread is not same. Here then it is not guaranteed that if you acquire Resource/Lock 1, you will succeed in acquiring everything. For example, suppose thread A need 1,2,3,4,5 and thread B needs 4,5,6,7,8 so when thread A and B starts, thread A will fail at lock 5 (assuming both get a chance to run and other scheduling operations are same).
Hope this helps.