How Should We Judge The Quality of a Sell Decision?

Determining whether a decision to sell an investment was correct seems like one of the simpler judgements that we make. If it continues to perform well after we have sold it then it was a mistake; whereas if it goes on to struggle it was a wise choice. Unfortunately, it is not always that easy. Focusing on the outcome of a decision alone can often be deceptive and lead us to draw entirely erroneous conclusions.

To explore why this is often the case, it is useful to offer some examples. The main issues we face when assessing the quality of a sell decision is dealing with known costs, hindsight bias, and tail risks:

Known Costs

The manager of a value fund sells a stock once it returns to its fair value. Following its sale, the stock continues to perform well and reaches a historically expensive price.

This can easily be viewed as a fund manager making the error of selling too soon or failing to run their winners. But is it? The manager has a philosophy and process built on buying undervalued securities, should we really expect them to continue to own a position when it becomes expensive?  This would invalidate the process. In such situations following the process is the correct decision even if it appears to be a mistake from a pure outcome perspective.  

If a manager persistently encounters situations where they are selling stocks that go on to outperform, they could adjust their process to capture this momentum. The alternative route is simply to state that missing out on these returns is a known and accepted cost of their approach.

Hindsight Bias

We are investing in an actively managed fund that undergoes significant changes; the asset management company is acquired, and the fund manager departs. An individual with minimal experience is appointed to run the fund. We decide to sell our position in the fund. It performs incredibly well over the next five years and is amongst the strongest in its peer group.

When we review a sale, we will inevitably incorporate information that was not available at the time the decision was made. The potent combination of hindsight and outcome bias means that not only we will think it was a mistake to sell the fund, but that it was obvious that we should not have done so.

In this instance, the new manager may simply have been lucky. Even if they do possess skill, we did not have sufficient knowledge at the time to make such a judgement. The choice to sell the fund was unequivocally the correct one. It will appear, however, as if we sold a talented fund manager at precisely the wrong time.

Tail Risks

We are investing in a fund which has a stellar track record but are becoming concerned about certain developments, particularly around the behaviour of the fund manager. We are not sure how serious these are but estimate that there is a 5% chance of the fund ‘blowing up’. Given this, we decide to sell the position. The fund continues to deliver outstanding returns in the years that follow.

This is a simplified example designed to portray the most difficult type of sell situation faced by an investor. Despite the decision being correct, we will likely be considered foolish for making it. As if we were worrying about nothing and have needlessly forgone performance.

When we judge the success of decisions solely on outcomes, we are likely to consistently penalise those choices that seek to avoid the cost of low probability, high impact failures. As, by definition, they do not often come to pass. Prudent, risk conscious choices will frequently appear unnecessarily cautious.



It is not that results are unimportant when assessing the quality of sell decisions. They can tell us a significant amount about our investment behaviour, particularly if we can build a large enough sample size. They do not, however, tell us everything. Successful investment is about creating a process that gets the odds on our side. In a system where randomness and chance can exert an overwhelming influence, we need to think less about the result of a given decision and more about whether it was consistent with a robust process.