“I am more likely to repeat actions when I succeed, I am more likely to change actions when I fail.”
We take note of the outcomes caused by our behaviours. On the basis of these outcomes, we change the likelihood of performing these behaviours in the future. Behaviours that lead to positive outcomes are more likely to be repeated in the future (win-stay), while behaviours that lead to negative outcomes are more likely to be changed in the future (lose-shift) . In particular, subsequent decisions following the experience of negative outcome tend to be more impulsive  and hence sub-optimal.
In riding a bike, it is important to learn how to operate the pedals correctly. After a gentle push, you see whether you can maintain your speed by rocking the pedals backwards and forwards. You fall off your bike. Since this rocking motion led to a negative outcome, you will be more likely to try a different action next time (lose-shift). You get back on the bike again and decide to rotate the pedals only forwards this time. The bike maintains its speed. Since this forward motion led to a positive outcome, you will be more likely to repeat this action (win-stay). However, we might misapply these shortcuts to situations in which win-stay and lose-shift are not logical. For example, we might repeatedly play on a slot machine that is paying out (win-stay) and move to another machine if it is not playing out (lose-shift). This is despite slot machines being random systems over which we have no control.
The origins of win-stay / lose-shift have been examined at a number of levels. For individual and species survival, it is important that we avoid negative outcomes and seek out positive ones. Therefore, win-stay and lose-shift might represent natural tendencies that have evolutionary advantage.
Even though lose-shift appears to be the exact opposite of win-stay, there are differences in the ability to control and express these two shortcuts . Lose-shift behaviour appears more prevalent than the win-stay behaviour, probably due to the increased importance that we place on negative rather than positive outcomes .
The ability to win-stay and lose-shift can be beneficial in the context of learning. However, there are other environments in which these tendencies do not work in our favour such as in a competition where there can be only one winner. For example, during competition we need to avoid the risk of being exploited. Exploitation comes from our competitors finding predictable patterns in our behaviour and then performing in a way that will take advantage of our predictability. Therefore, in expressing win-stay and lose-shift in competitive environments, we run the risk of being exploited.
Thoughts on how to act in light of this bias
The only way to avoid being exploited is to try to make your behaviour as random as possible. For example, if you take a penalty in Soccer, you do not want to have a known bias in kicking to the left or the right. One type of action should not be expressed more frequently than another, and, the future actions should not be determined by the outcome of the previous action. This may require inhibiting the impulse to respond quickly (especially following failure) to allow for more decision-making time.
How is this bias measured?
Win-Stay / Lose-Shift can be measured in a number of ways. One way to do this is to create an environment where the number of options is limited. For example, this could be something as simple as the game Rock, Paper, Scissors, where three options are available. The number of times individuals repeat their option following a win (e.g., Rock to Rock) and change their option following a loss (e.g., Rock to Paper, or, Rock to Scissors) can then be tallied. These percentages can then be compared to other percentages that would be expected if the individual was playing randomly (e.g., 33% win-stay and 66% lose-shift) . Biases are represented by observed values that are different from expected values.
This bias is discussed in the scientific literature:
This bias has social or individual repercussions:
This bias is empirically demonstrated:
 Thorndike, Edward (1901). Animal intelligence: An experimental study of the associative processes in animals. Psychological Review Monograph Supplement, 2, 1–109.
 Eben, Charlotte, Zhang Chen, Luc Vermeylen, Joel Billieux & FrederickVerbruggen (2020). A direct and conceptual replication of post-loss speeding when gambling. Royal Society Open Science, 7: 200090.
 Kubanek, Jan, Lawrence H. Snyder & Richard A. Abrams (2015). Reward and punishment act as distinct factors in guiding behavior. Cognition, 139, 154-167.
 Kahneman, Daniel & Amos Tversky (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263-291.
 Forder, Lewis & Benjamin James Dyson (2016). Behavioural and neural modulation of win-stay but not lose-shift strategies as a function of outcome value in Rock, Paper, Scissors. Scientific Reports, 6: 33809.
Individual level, Need for security, Interpersonal level
Dr Ben Dyson is an Associate Professor in the Department of Psychology, University of Alberta, Canada. He graduated from York University, UK, in 2002 after completing his thesis on auditory cognition and went on to a postdoctoral fellowship position at the Rotman Research Institute, Canada (2002-2004) to learn about event-related potentials (ERPs). He had held academic positions within Psychology Departments at the University of Sussex, UK (2005-2008) and Ryerson University, Canada. His current interests surround the use of simple games in understanding decision-making at a behavioural and neural level.