Du Malone writes: There’s this striker who’s just had a phenomenal season. He’s been banging the goals in left, right, and centre. And he’s just finished as the league’s leading goalscorer.

An then this other striker who has never hit such heights. He’s been playing several seasons now, but has never finished as the leading goalscorer. Never even seriously been in contention.

He does, though, have a creditable record. For several years in a row, he’s been scoring an above-average number of goals each season.

So who should you sign?

Instinctively, most of us want the chart-topper.

Bur according to research by Dr Chengwei Liu of Warwick Business School, this preference may be irrational.

Research on performance

In an interview for the UK talent firm, FJ Wilson Talent Services, Dr Liu explains his findings as follows:

My main research program addresses a fundamental question in strategy: should we attribute performance differences to skill or to luck?

For example, the most successful are often perceived to be the most skilled and therefore receive the highest rewards and are imitated.

I show that the belief that the exceptional performers are the most skilled is flawed because exceptional performance is more likely to occur in exceptional circumstances and top performers are often the luckiest people who have benefited from rich-get-richer dynamics that boost their initial fortune


The misattributions of luck can imply opportunities for informed entrepreneurs and strategists because these mistakes can lead to systematic mispricing of talents. For example, most firms pursue the top performers, i.e., “stars”, believing that they have the highest talents or skills.

My research empirically shows that this practice is flawed because the top are likely the luckiest while the “second best” are in fact the most skilled.

To return to the goal-scoring example. It may be that the goalcorer was playing in a context that was peculiarly suited to them. At a different club, in a different team, they might not enjoy the same success.

Especially when you consider that, in the course of one season, factors don’t always balance out. You may strike lucky with, for example, offside decisions or defensive errors.

In contrast, if someone consistently performs well, even without ever performing outstandingly, that performance is unlikely to be the result of luck. Especially if they achieve this performance in a variety of contexts. If you buy that player, you will be buying not luck, but talent.

Dr Liu adds: “The bias that favours the top leads to the second best being seriously underestimated. This provides an opportunity for the more informed — through the acquisition of  superior talents with a price lower than their actual worth“.

Data analysis

Dr Liu’s research has major implications for data analysis in sport.

I don’t know whether clubs like Burnley and Sheffield Utd are familiar with Dr Liu’s research, but it seems to me that, one way or another, they’ve cottoned on to the gist.

To football managers the moral is clear: Rather than making decisions on exceptional performance, especially over the short-term, search for consistently strong (though not necessarily outstanding) performance over several seasons, ideally in diverse contexts.


Image credit: ‘Day 260‘, generously made available by vtbrak under a CC BY-NC-ND 2.0 licence.


5 thoughts on “Team selection and player acquisition: may be we’ve been doing it all wrong?

  1. Well, isn’t that what xG is about? Look for the performance rather than the result, because getting good chances is better reproducible than finishing (although, in FM, finishing is a real skill and reproducible, at least often). So I think Dr. Liu is looking at the noise/luck that’s there, and warning to not see that as the signal. But – signal is still there to find, you just have to look harder than at goals 😉

    So, if you look for performance in FM, what do you look at? With forwards, I would suggest number of shots and shots on target (be careful with free-kick takers and maybe penalties, because these help but don’t stack – they can be only taken once per team, not player). Headers are worse for scoring, so maybe try to look at number of goals from them. Haven’t really tried that for recruiting though so far.


    1. V. grateful for your thinking here. Am approving it straightaway, so that other readers can benefit. Will ponder more carefully as soon as I have some time. Will confess now, though, that the whole xG debate has passed me by: beyond knowing what the initials stand for, I know nothing!


    2. I’m not sure how useful a purely statistical basis would be. How suitable are the things that can be measured as proxies for what we want to measure? For example, if a forward continually whacks it at the goalie, they achieve a high shots-on-target stat. A forward who tries to place it in the corner will have a lower stat, but might be preferable?
      I think my protocol/scorecard for signing a forward might be:
      a. has he scored consistently over a run of seasons (say, at least 30? if so, good
      b. has he done so for more than one club? if so, even better.
      c. ditto assists
      d. how good is the club? look at other players but also context: coaches, training facilities etc.
      e. make a judgement: is his performance at best merely in line with what all this would lead you to expect? or better?
      I’ve been doing something similar with clean sheets data. Looking at raw numbers of clean sheets (or merely divided by number of appearances) isn’t much use. Looking for consistency in these things over a number of seasons is better. Then relating to context e.g. strength of team, goalkeeping coaching etc. can help to identify winners.


    1. Yes, exactly that. And that example is spot-on. I’ve observed a similar phenomenon elsewhere — in book publishing, for example.


Leave a response

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.