Many of us have likely travelled on planes before, arriving to our destination when vision is impaired, either through it being a cloudy day, or daylight has disappeared. As the plane begins to descend from the sky, bright lights begin to appear, to signal where the plane is aiming to land. These lights either side of the runway or landing strip are hugely important sources of information for the pilots to guide the plane back to the ground. In the same way, errors in training are hugely important sources of information for players and coaches, to guide performance to higher standards.
The process of altering behaviour to increase chances of success is called reinforcement learning (Sutton and Barto, 1998). Reinforcement learning occurs through prediction errors, which can be defined as the difference between expectation and outcome. For example
Team A is on a 3-game winning run, and they are about to come up against team B who have lost their last 4 games. Team A are anticipating a win, but Team B causes a surprise and dominate the game to win 2-0. This would be a prediction error for team A. This is a sign for team A that they need to alter their performance to achieve success.
From an individual standpoint, a centre forward may be working on his finishing outside the box. However he is not achieving much success against the goalkeeper. This is an indication that the centre forward needs to alter his behaviour to score more goals. This may be to shoot lower, shoot to the goalkeeper's weak side, or to aim more for the corners of the net.
One of the major drivers of reinforcement learning are reward-prediction errors - which refers to the degree an actual reward differs from the anticipated reward (Holroyd & Coles, 2002).
Positive reward-prediction: acts as a signal to stamp in that movement pattern. For example, you sink your first long putt - that movement produced an unexpectedly rewarding result. Unexpected rewards will produce the largest positive reward-prediction. This is why fantasy football is such a buzz for so many people.
Negative reward-prediction: acts as a signal to stamp out that movement pattern. For example, a player exhibiting a high shooting success, and then they miss - the movement produces an unexpectedly negative feeling, as the brain predicted a higher degree of success/probability of reward.
Large reward-prediction errors are common early in learning. Small prediction errors are common late in learning.
Feeling successful defines a rewarding experience in the brain, therefore subjective success is more important than objective success. One way coaches can enhance this feeling of subjective success is to provide positive feedback, whether that is after good attempts, near misses or significant failures.
The self-determination theory (SDT; Ryan & Deci, 2000) states that intrinsic motivation influenced by perceived competence in the task, autonomy, and/or relatedness. A successful outcome for a learner with perceived competence, autonomy and relatedness will result in a larger reward-prediction error than a learner without those factors. Coaches should aim to enhance 1) perceived competence, 2) autonomy and 3) relatedness in players to allow for greater learning to either consolidate successful behaviours (positive), or inhibit unsuccessful behaviours (negative).
Perceived competence can be enhanced by providing positive feedback to learners.
Autonomy can be enhanced by granting learners control over the order of practice exercises, when to receive feedback, the progression of difficulty during practice.
Reward-prediction errors facilitate learning. These errors should be maximised through a) the SDT, or b) appropriate level of difficulty for the individual. These errors can be positive or negative, with positive errors more common with novices and negative errors are more common with skilled performers.
Expectation to perform poorly + good performance = positive reward-prediction error => consolidate successful movement patterns.
Expectation to perform well + poor performance = negative reward-prediction error => inhibit unsuccessful movement patterns.
A key point for coaches is to optimally challenge players during practice, which means errors should be present, but also to support players through this optimal challenge. If a training task or session challenges players optimally, but players are not supported through this process and they leave feeling devoid of perceived competence, then the coach has not done a good job in my opinion.
Limiting errors in practice invokes implicit motor learning processes rather than effortful, explicit motor processes. Error-reduced learning promotes gradual changes and adaptations in movement patterns with a minimal discomfort and disturbance in attention requirements. However, error-reduced practice does not mean error-eliminating. Reducing errors might result in the optimal level of learning, which creates an autonomy-supportive environment and increases self efficacy (i.e. SDT), and enhances decision-making.
Errors are not something that should be avoided in practice. Rather they should be embraced as they can be utilised to guide skill acquisition and learning. Error-less practice means learning is not taking place. Practice needs to impose a level of desirable difficulty (Bjork) on the learner if learning is to be optimised. This difficulty is evidenced by errors. Errors are the signposts to optimal learning. Too much error leads to frustration, to little error leads to comfortable practice and ultimately poor performance.
Value of errors for players - informs players if they need to adapt, either to consolidate a successful movement, or inhibit an unsuccessful one.
Value of errors for coaches - informs coaches if the task is at the optimal difficulty: too few errors = task is too easy; too many errors = task is too difficult.