Optimistic about Optimisation

Two Medium posts this week (written by Thomas Oppong and David Weinberger) have returned me to think about optimisation as an organisational opportunity that goes beyond the rhetoric of ‘world best‘ and ‘world leading‘ aspirations.

Pursuit of progress

Thomas Oppong in his discussion of ‘good enough’ and the pursuit of progress quoted Seth Godin:

You’re not in the perfect business. Stop pretending that’s what the world wants from you.Truly perfect is becoming friendly with your imperfections on the way to doing something remarkable.

Thomas concludes his post with the observation “Pursuing progress allows you to celebrate each step that feels like an accomplishment”.

Machine Learning

David Weinberger‘s Medium post considered maximising the benefits of machine learning without sacrificing its intelligence. I was particularly interested in this suggestion:

Accept that we’re not always going to be able to understand our machine’s “thinking.” Instead, use our existing policy-making processes — regulators, legislators, judicial systems, irate citizens, squabbling politicians — to decide what we want these systems optimized for. Measure the results. Fix the systems when they don’t hit their marks. Celebrate and improve them when they do.

In his discussion of optimisation, David made three proposals:

1. Artificial Intelligence systems ought to be required to declare what they are optimized for.
2. The optimizations of systems that significantly affect the public ought to be decided not by the companies creating those systems but by bodies representing the public’s interests.
3. Optimizations always also need to support critical societal values, such as fairness.

David concludes:

The concept of optimization has built into it an understanding that perfection is not possible. Optimization is a “best effort.”


I have been thinking about how sport systems might flourish without the burden of winning edges or outcome measures in a highly competitive global sport network.

I do have an intrinsic connection with ‘good enough’ and ‘best effort’ approaches. They are invitational and permit us to fail as a learning opportunity.

My hope is that the acceptance of optimisation processes that offer better ways of performing in less effortful ways encourages and supports playfulness and joy. This I take to be a profoundly ethical undertaking that requires us to adhere to transparent fairness.

This kind of approach offers not an edge but a wide open space for the flourishing of our imagination and optimism.

Photo Credit

Climb (Efren, CC BY-SA 2.0)

Actionable insights: sport analytics



A post by Mary Hamilton (2017) about her time at The Guardian has sent me off thinking about actionable insights in sport analytics.

In her article, Mary shares thirteen lessons from her time as executive editor for audience at The Guardian. Three of the thirteen had a particular resonance with me.

Insight 1 is ‘Data isn’t magic, it’s what you do with it that counts‘. She notes “We make better decisions when we’re better informed, and all data is is information”. She adds that developing an in-house data resource, Ophan, “It’s not just about putting numbers into the hands of editorial people — it’s explicitly about getting them to change the way they make decisions, and to make them better”.

Insight 11 is ‘Radical transparency helps people work with complexity‘. Mary observes “In a fast-moving environment where everything is constantly changing … you have no way of knowing what someone else might need to know in order to do their job well. The only way to deal with this is to be a conduit for information, and not bottle anything up or hide it unless it’s genuinely confidential”.

Insight 13 is ‘What you say matters far less than what you do‘. Mary’s take is “This should be obvious, but it probably isn’t. It doesn’t matter what you say you want, it’s what you do to make it happen that makes a difference in the world”.


Mary’s thirteenth insight underscores the importance of action. In another context, Adam Cooper (2012) proposes:

Analytics is the process of developing actionable insights through problem definition and the application of statistical models and analysis against existing and/or simulated future data.

He adds “‘actionable‘ indicates that analytics is concerned with the potential for practical action rather than either theoretical description or mere reporting”.

Some of the key contributors to the sports analytics literature have focused on action.

In 2011, Ben Alamar and Vijay Mehrotra defined sport analytics as:

the management of structured historical data, the application of predictive analytic models that utilize that data, and the use of information systems to inform decision makers and enable them to help their organizations in gaining a competitive advantage on the field of play.

Three year later, Chris Anderson defined sports analytics as:

The discovery, communication, and implementation of actionable insights derived from structured information in order to improve the quality of decisions and performance in an organization.

In 2016, Bill Gerrard observed “Sports analytics is all about using data analysis to provide actionable insight for coaches and sporting directors”. He added “Analytics is analysis for purpose. It’s a servant function, there to help managers to make better informed decisions”. In his conceptualisation, “Analytics is decision-driven, domain-specific data analysis”.


In his essay on Concerning Human Understanding, John Locke asserts:

it is ambition enough to be employed as an under-labourer in clearing the ground a little, and removing some of the rubbish that lies in the way to knowledge

Bill’s comment about the servant function of analytics took me back to John Locke and under-labouring. I thought that any sport analyst would be keen to be such a labourer and contribute to “clearing the ground a little”.

One aspect of an analyst’s role is, I think, to reflect on the place granularity will play in the actionable insights we share. Another is to consider our creation of actionable insights from a user’s perspective (Kunal Jain, 2015).

But as Alexander Franks and his colleagues (2016) point out there are some important meta-issues at play here too. They consider “the metrics that provide the most unique, reliable, and useful information for decision-makers”. They employ three criteria to evaluate sport metrics:

  • Does the metric measure the same thing over time? (Stability)
  • Does the metric differentiate between players? (Discrimination)
  • Does the metric provide new information? (Independence)

Alexander and his colleagues note:

In general, any individual making a management, coaching, or gambling decision has potentially dozens of metrics at his/her disposal, but finding the right metrics to support a given decision can be daunting. We seek to ameliorate this problem by proposing a set of “meta-metrics” that describe which metrics provide the most unique and reliable information for decision-makers.

They add:

The core idea of our work is that quantifying sources of variability—and how these sources are related across metrics, players, and time—is essential for understanding how sports metrics can be used.

They conclude their discussion of meta-metrics by proposing a fourth meta-metric: relevance.

Relevance could simply be a qualitative description of the metric’s meaning or it could a quantitative summary of the causal or predictive relationship between the metric and an outcome of interest…

In Practice

Earlier this week, the Crusaders rugby union team, from New Zealand, advertised for a sport scientist. The job description provides a fascinating empirical focus for the discussion in this blog post.

The position description has these elements:

  • Reporting to the Crusaders’ Head Strength and Conditioning Coach, you will be responsible for overseeing and coordinating all aspects of the Crusaders’ performance monitoring systems including further enhancement of data collection, processing and reporting methods.
  • You will also be responsible for collating and reporting on all performance monitoring data to ensure optimal player loading for conditioning and recovery.
  • To be successful in this role, you will need to be appropriately qualified by training and/or experience, including a proven ability in research, data analysis and reporting including an outstanding level of understanding of performance monitoring and analysis tools ideally in a rugby environment.
  • You will also need to demonstrate extensive experience in the use of GPS Technology both hardware and software…

This kind of role is now becoming more and more frequent. It will be good to learn how the post holder adapts to this role and provides relevant, actionable insights that are domain specific whilst being mindful that in addition to structured data there are increasing opportunities for the analysis of unstructured data.

Photo Credits

Tree (Keith Lyons, CC BY 4.0)

Crusaders v Cheetahs (Geof Wilson, CC BY-NC-ND 2.0)

#cssia17 Connecting and Sharing

I have been following up on some leads shared by Mara Averick. Two recent suggestions caught my attention as I try to improve the ways I share and connect.

The first was a post by Joris Muller about reproducible computational research for R users. In it he explores ideas shared in a 2013 paper written by Geir Sandve and colleagues. In that paper, Geir proposes ten rules for reproducible computational research. These are very pertinent to those seeking to share and explore performance in sport using analytics insights.

The ten rules are:

  1. Keep track of how every result was produced.
  2. Avoid manual data manipulation steps
  3. Archive the exact versions of all external programs used
  4. Version control all custom scripts
  5. Record all intermediate results in standardised formats when possible
  6. For analyses that include randomness note underlying random seeds
  7. Always store raw data behind plots
  8. Generate hierarchical nalysis output allowing layers of increasing detail to be inspected
  9. Connect textual statements to underlying results
  10. Provide public access to scripts, runs and results

Joris concludes his post:

All the 10 rules proposed in the Sandve paper are reachable for a R user. Just by using R itself, the rmarkdown workflow and some organisational rules cover most of these rules. My basic reproductible workflow meet almost all the criterias with the notable exceptions of the software archive (but it’s work in progress with packrat) and the lack of public access (but I can’t share everything).

For an introduction to Joris’s workflow, you might find this post of interest.

The second lead from Mara focussed on reproducible behaviour too.  Jenny Bryan shared her ideas back in 2015 about Naming Things. This is one of the many resources Jenny has shared. I have found her GitHub repositories immensely helpful. In her 2015 paper, Jenny notes three principles for file names: machine readable, human readable and ‘plays well with default ordering’.

The two leads sent me off thinking about how I might improve my practice. I am fascinated by Joris’s transparency with his workflow and I see this approach as essential for sport analytics as we start to extend cumulative rather than ‘ab initio‘ research. I admire Jenny’s work immensely. I have tried to use some robust file naming conventions for the past fifteen years as I have sought to use cloud based storage for all my resources. I realise I am a long way from meeting Jenny’s three principles at the moment but this will be a work in progress.

Mara Averick’s Twitter recommendations are becoming a very important way for me to connect with a community of practice. These two leads discussed here are a way for me to make this process explicit … and to initiate a conversation about reproducible behaviours in sport analytics research and practice.

Photo Credits

Tree on campus (Keith Lyons, CC BY 4.0)