With human reaction speed comes the dynamic of abusing it to your advantage. Players create intentional guessing games. If we had a dice game where the object was to guess the number rolled as quickly as possible, this game would be entirely reaction based. When this is placed in a competitive environment the person with the best reaction doesn't always win because humans have a tendency to guess as well. So, even though you could rely entirely on your reaction speed, a little bit of guessing can come into play, which means luck plays a part in it (guessing right and wrong is luck).
This is where a lot of the guessing games comes into play for just about any real-time sport or game. Anything turn-based wouldn't need reaction speed at all. You have all the time in the world to take your turn and therefor it's a matter of decision making. When this is placed in a competitive environment a couple things can happen depending on the nature of the game.
With a turn based game like Poker there is quite a lot of guessing involved because the cards you get are randomly given to the players.
With Chess the absolute best play can be determined at any given turn from the first turn to the last turn. But most humans are not capable of memorizing every single correct play for every single possible occurrence on the board. Once again, humans have a tendency to guess in order to compensate for the lack of knowledge.
So even without reaction speed there's marginal amounts of guessing. But the difference between these types of guessing games is; with turns the guesses are entirely education based guesses. With reaction speed type games is comes down to simple RPS (Rock Paper Scissors).
There's also a significant difference between guessing and mind-reading, similar to above -- though many people tend to confuse the two. With guessing you are simply guessing at random or pseudo-randomly. With mind reading you are taking into account things like: risk and reward, security, probability.
Mind reading therefor isn't the simple act of guessing, but highly educated guessing. It is intentionally doing specific actions to create a situation in your favor, then performing your next action based on what you anticipate your opponent will do. In an controlled environment you know what to expect, and so you've read the opponent's mind.
For example, let's say you can put your opponent at a disadvantage with a move that puts you next to them, appropriate for a tick throw. However, the throw doesn't have a whiff animation, instead you attack if the opponent can not be thrown. Thus, if the opponent attempts to jump in order to avoid the throw they will be hit instead. This is forcing your opponent to make a decision, not a guess.
The "correct" option is to take the hit because it does less damage than the throw and gives the aggressor less mixup options after the hit. It's very little risk for the aggressor so there's hardly any reason not to use this option. And you don't need to react to what your opponent did because you've read the situation well enough to know your opponent will choose that option: the path of least pain.
This is only one example of many possible examples. There's countless ways for you to lead your opponent into a situation where they are forced to take damage, and knowing all their options they will choose the option that is going to hurt them the least. And because of the risks, the odds of countering, and the probability that both players are likely to follow that path, you can know in advance what your opponent will do. A predictable action that gets countered.
However, many players experience the "why would you do that?" moment when we fight a "bad" opponent that does not consider risk/reward and such. This happens when a player uses an option with ridiculously high risk and poor reward, such as a dragon punch or other punishable such move. When this happens, the player doing the dragon punch has risked being punished for huge amounts of damage, whereas if they had taken the safer option they would be hurt quite a lot less. The reward is also not particularly good; minor damage and a knockdown with limited mixup options at midscreen. And the security is usually rather poor, if they fail they'll probably be in the corner, whereas if they succeed they will probably only get the opponent at roughly mid-screen.
And that's why the aggressor needs to take into consideration probability. You need to be aware when the opponent is likely to be willing to bank on a huge risk. It's often when they are in trouble, but not always. There's also player personality and habits and such that you can familiarize yourself with.
In the end, simple random guessing would always lose to mind-reads (educated guessing) because the risks and rewards are not balanced. Mind reads are also less prone to failure. Rather, a true mind read isn't prone to failure at all. Instead your forcing your opponent into a situation where you can only succeed, and are therefor forcing your opponent to decide which option they want to get hit by. The trick is getting your opponent into the trap in the first place.