Practice
Explore vs Exploit
Searching for new options and using a proven solution serve different purposes. The right balance changes with the goal of practice.
- Explore
- Compare
- Exploit
- Reconsider
Several exploratory routes converge on a chosen solution, with room to reopen investigation.
Two uses of the same training time
Are you investigating a new entry or making the familiar pass dependable? Both can help, but you assess them differently. Exploration seeks information about an uncertain option. Exploitation uses something already reasonably reliable. Sutton and Barto discuss this tension in reinforcement learning. We borrow it for practice; people and partners are not a simple computer environment.
Always exploiting can produce skill inside a narrow area while leaving important alternatives undiscovered. Always exploring can collect options without developing any of them sufficiently. There is no ideal permanent ratio for every athlete. Experience, injuries, competition plans and available guidance all change what counts as a worthwhile experiment.
A known pass and an unfamiliar entry
A grappler has a dependable pass but struggles to obtain the required grip. During a training block, they investigate a different entry while keeping the eventual pass familiar. That is exploration without changing the whole game at once. Their partner understands which resistance is useful and keeps intensity appropriate.
Closer to competition, more practice may go towards exploiting the established route, including recovery when the first attempt fails. Curiosity need not disappear. A short exploratory block can still reveal an unfamiliar response. The question is not whether athletes should always be creative. It is what purpose the current round serves and whether the chosen task matches that purpose.
Exploration needs a question
Trying a different technique every round does not automatically produce useful information. Without a problem to investigate, you cannot easily assess why something worked. Explore with a question: does this entry create a usable angle when my partner closes the normal grip? Keep enough of the remaining situation familiar to interpret the answer.
Exploitation does not mean selecting only less skilled partners for comfortable victories. Dependable solutions still need appropriate resistance. An unsuccessful experiment does not immediately make the new option worthless either. It may be underdeveloped, unsuitable for this response or genuinely unhelpful. Distinguish these explanations before abandoning a direction. Both modes require judgement rather than a simple preference for novelty or familiarity.
Give each round a purpose
For one training week, choose a limited exploration question. Also schedule rounds in which you try to execute an existing solution reliably. In the exploratory block, record what you discovered. In the other block, examine recognition, timing and recovery. This prevents both kinds of practice being judged exclusively by whether you won.
At the end of the week, decide whether the new option deserves more time, should be set aside or can enter your ordinary game. Base the decision on what you observed rather than a wish to seem constantly inventive. Opportunity Cost then asks what this investment displaces. Exploration earns its place when information improves a later choice, while exploitation lets useful learning become dependable action.
Sources and further reading
- Reinforcement Learning: An Introduction, second edition
Sutton and Barto address exploration and exploitation computationally; the sport application is an analogy.
See an idea alongside another
Related models
Opportunity Cost
Time spent exploring displaces another possible investment.
Tokui Waza
A personal speciality needs time spent developing dependable use.
Scientific Method
A clear question makes exploration more informative than random attempts.
Bayesian Updating
New information can change confidence in a solution.