what's the difference between different ways of iterate through a list

Question

I am currently taking cs188 courses offered by UCB. I encountered the problem when implementing the alpha-beta pruning algorithm. Two different ways of implementation is showed below.

# this one works
 def max_value(self, gamestate, depth, agentIndex, alpha, beta):
        v_max = float('-inf')
        max_action = ''
        # get the possible actions from gamestate
        legal_actions = gamestate.getLegalActions(agentIndex)
        for action in legal_actions:
            # get the successor corresponding to each action
            successor = gamestate.generateSuccessor(agentIndex, action)
            successor_value = self.value(successor, depth, agentIndex+1, alpha, beta)
            # get the maximum value
            if successor_value[0] > v_max:
                v_max = successor_value[0]
                max_action = action
                # pruning
                if v_max > beta:
                    return (v_max, max_action)
                alpha = max(v_max, alpha)
        return (v_max, max_action)

# this one fails to prune some nodes that should be pruned
def max_value(self, gamestate, depth, agentIndex, alpha, beta):
        v_max = float('-inf')
        max_action = ''
        # get the possible actions from gamestate
        legal_actions = gamestate.getLegalActions(agentIndex)
        # get the successor corresponding to each action
        successors = []
        for action in legal_actions:
            successors.append(gamestate.generateSuccessor(agentIndex, action))
        for i in range(len(successors)):
            successor_value = self.value(successors[i], depth, agentIndex+1, alpha, beta)
            # get the maximum value
            if successor_value[0] > v_max:
                v_max = successor_value[0]
                max_action = legal_actions[i]
                # pruning
                if v_max > beta:
                    return (v_max, '')
                alpha = max(v_max, alpha)
        return (v_max, max_action)

what's the difference between them. I guess there's something wrong with the alpha-beta updating in the second implementation, but I failed to locate the bug.