Precision

It has been a feast of delight this week on Clyde Street.

I have been following up on the ideas shared by David Zeevi and his colleagues about personalisation and prediction. One part of the paper has stayed with me:

Dietary interventions based on our predictor showed significant improvements in multiple aspects of glucose metabolism, including lower PPGRs and lower fluctuations in blood glucose levels within a short 1-week intervention period. It will be interesting to evaluate the utility of such personalized intervention over prolonged periods of several months and even years. (My emphasis)

I have been thinking about the implications of this for the learning and performance environments we build and maintain in sport. A week of investigating research and practice in precision medicine has encouraged me to contemplate the skills we might need to be engaged with long-term athlete and coach personal flourishing.

Leroy Hood and his colleagues have championed a systems biology approach to predictive, preventative, personalised and participatory healthcare. They noted:

Systems biology is a scientific discipline that endeavours to quantify all of the molecular elements of a biological system to assess their interactions and to integrate that information into graphical network models that serve as predictive hypotheses to explain emergent behaviours. (2004:640)

Leroy and Andrea Watson predicted:

a paradigm shift in medicine will take place within the next two decades replacing the current approach, which is predominantly reactive, to one that can increasingly predict and prevent cellular dysfunction and disease. (2004:179)

A decade later, researchers and practitioners are embedded in precision medicine and deliver treatments:

targeted to the needs of individual patients on the basis of genetic, biomarker, phenotypic, or psychosocial characteristics that distinguish a given patient from other patients with similar clinical presentations. (Larry Jameson and Dan Longo, 2015)

This shift requires a fundamental rethink of how to deliver personal care. Reza Mirnezami, Jeremy Nicholson and Ara Darzi (2012), for example observe:

Precision medicine will require handling of multi-parametric data and some proficiency in interpreting “-omics” data, placing new demands on medical professionals, who may be ill equipped to deal with the anticipated complexity and volume of new information. Addressing these challenges will require effective clinical decision support tools and new educational models.

These new educational models fascinate me, particularly in the context of understanding how analytics are embedded in coaches’ learning pathways in formal accreditation and in continuing learning.

Two posts this week added to my reflections about precision.

In the first post, Ricardo Tavares considered Why We Need Positional Data in conversations about football analysis. I really enjoyed the way Ricardo shared the process of analysis using a single example. I delight in n=1 studies and their resonance with other performances.

What made Ricardo’s post of particular interest to me was the open sharing he demonstrated. His post concludes:

You can download the the csv file with the data here (x and y coordinates on a scale of 0 to 100). The player data is here (player numbers and names aren’t filled yet, but they should be up soon).

If you know Python, you can also view (and download) the Jupyter Notebook that made the animations here (or here, for a more browser friendly version).

The second post, Protecting an NHL Player’s Greatest Asset is an interview with the San Jose Sharks’ trainer, Mike Potenza. In the interview, Mike notes:

The NHL is one of the longest seasons in professional sports. Each team will commonly play 13-16 games per month with no consistency to the format of days they play. West coast teams will travel more than the east coast teams due to the proximity of franchise locations. Given the compressed game schedule, travel schedule and requirements for mandatory days off per league rules, practice time is limited but still a valuable commodity.

In San Jose our goal is to monitor every practice during the season, which includes pre-season training camp. Team workload/intensity and duration of monitored practice times are shared with the coaching staff so they have useful information when planning the next workday and the yearly work to rest schedule.

Mike uses these data to monitor

  • Accumulative work load
  • Training effect
  • % of max HR
  • High intensity duration

Mike discussed the absence of HR monitoring in games (an NHL stipulation):

It is a major missing piece of the puzzle that we do not have game data from HR monitors or GPS units because we do not know the cost of a NHL hockey game and the stresses that go along with that. The frequency and physical component of games per week is very high both in the regular season and in the playoffs. This being the case, missing game data forces performance coaches to only draw conclusions from sub-maximal practice data that only can be compared to practice and not the main show! To further dissect the issue, by not monitoring games, performance coaches do not have a reference for the metabolic specific zones achieved in games. These are extremely useful pieces of data that would be used to assign HR training zones for players who earn substantially different time on ice (T.O.I.) accumulations.

This brought me back to think about the precision we bring to the observation about each player’s performance and the decision-support we use to modulate training, whilst having some clarity about long-term flourishing.

These are active debates in healthcare and I see a similar need to have these conversations in sport.

Ultimately, precision medicine should ensure that patients get the right treatment at the right dose at the right time, with minimum ill consequences and maximum efficacy. (Reza Mirnezami, Jeremy Nicholson and Ara Darzi, (2012)

Ricardo’s use of a single event to raise fundamental questions about what we observe and analyse in sport took me back to Ference Marton (1994:7) and the idea of phenomenography that aims “at a very specific level of description, corresponding to a level of experience believed to be critical as far as our capabilities for experiencing certain phenomena in certain ways are concerned”.

Considerations about precision in sport contexts require us, I believe, to make sense of digital records of performance in our everyday practice. As in medicine, we have immense opportunities to explore new paradigms. Phenomenography encourages us to reflect on “the question of people being capable, or not, of experiencing and acting in certain ways” (Ference Marton (1994:7).

This seems to me to be the start of an agenda to discuss our educational models.

Photo Credit

Dawn on Clyde Street (Keith Lyons, CC BY 4.0)

Real-time monitoring (Firstbeat)

Winning and Losing in the Regular Super Netball Season 2017

The regular season in the inaugural Suncorp Super Netball Competition concluded last weekend. The Vixens, Lightning, Giants and Magpies have progressed to the playoffs.

I have been following Champion Data’s coverage of the games played. There were two drawn games in the regular season (Firebirds v Lightning, week 1; Vixens v Swifts, week 3).

My median profiles for winning and losing teams in the remaining 54 games over 14 rounds were:

I used BoxPlotR to visualise some of the data too.

A comparison of Winners and Losers

The data for this visualisation:

These data are available in this GitHub Repository.

Photo Credit

Twitter (RSN927am)

Postscript

About BoxPloR

“This application was developed with Nature Methods as described in this editorial and this blog entry. Nature methods also dedicated a Points of View and a Points of Significance column to box plots.

This application allows users to generate customized box plots in a number of variants based on their data. A data matrix can be uploaded as a file or pasted into the application. Basic box plots are generated based on the data and can be modified to include additional information. Additional features become available when checking that option. Information about sample sizes can be represented by the width of each box where the widths are proportional to the square roots of the number of observations n. Notches can be added to the boxes. These are defined as +/-1.58*IQR/sqrt(n) which gives roughly 95% confidence that two medians are different. It is also possible to define the whiskers based on the ideas of Spear and Tukey. Additional options of data visualization (violin and bean plots) reveal more information about the underlying data distribution. Plots can be labeled, customized (colors, dimensions, orientation) and exported as eps, pdf and svg files.

On the radar?

I missed the Twitter exchange about radar plots yesterday.

I caught up with some of the exchanges through a Ted Knutson post about Revisiting Radars.

Ted noted “The fact that lots of people have reactions to this type of work is a good thing, not a bad one”.

I have been using radar plots in my work for some time. I am mindful of the issues that Luke and Sam (@stat-sam) raise.

Some time ago (2008), I was fascinated by the title of a Purna Duggirala (Chandoo) post You are NOT spider man, so why do you use radar charts? I enjoyed Graham Odd’s (2011) critique of radar charts too.

Primed with Graham’s observation:

… the overall shape presented for a series on a radar chart does not leverage any of the pre-attentive attributes we perceive quantitatively. In essence, this means we are unable to attribute much genuine meaning to the shape of a series. The only patterns our visual perception can really discern in a data set presented as a radar chart are similarity and extreme outliers.

I tried to use my radar charts as a stimulus for conversation. Like, Ted, over the years I have found radar charts a good way to hook attention and trigger conversation. I had similar experiences to Ted in a variety of sport contexts:

In situations like this, visuals go a long way toward opening the conversation. If you show a table of numbers to a coach who isn’t already on board, you’re dead. Bar charts? Only mostly dead. Radars? Interesting… Tell me more.

My use of the charts acknowledged the limitations of the visualisation. Once the coaches had started to discuss the issues raised by a fallible (flawed) visualisation, we inevitably started to discuss how performance might be re-presented (represented) which led in many cases to some fascinating second-order conversations about observation and the narratives we build around performance … and other forms of visualisation.

My willingness to use radar charts dates back to William Anderson’s (1971) discussion of descriptive-analytic research in physical education. He notes:

Their principal concern is to collect accurate descriptive records of events in actual classrooms and to analyze these records in a way that enables a better understanding of the events. (1971:2)

He adds:

The descriptive records of teaching which emerge are in many ways like the descriptive record of a basketball game contained in a basketball shooting chart. The shooting chart is a diagram of a basketball court on which is recorded the number of each player who took a shot, the place on the court from which the shot was taken, and whether the shot was made. The shooting chart is a record of a critical dimension of “real world events” (the game). A careful examination of the chart can lead to understandings and insights which were not possible during the game itself. In much the same way, descriptive records of teaching provide a picture of real world events (classroom interaction) which lead to a deeper understanding of the teaching process. (1971:3)

Ted makes the point “As I designed them, radars exist to help you open the door with statistical novices, and from that perspective they have been wildly successful”.

This happened in conversations about pedagogy too in William Anderson’s work.

Like Ted, I am acutely aware of the flaws in radar charts. We have unprecedented expertise in sport now. The stories we can produce have immense visualisation resources to share performances. The key for me will be how we work with a variety of audiences in sport, and particularly in decision support for coaches, to achieve the level of engagement Ted reports:

Radars start a conversation. They get a reaction. And for whatever reason, football people are often more comfortable talking about and digesting them than almost any other vis type I have encountered. (Original emphasis.)

These issues are why I have included Audiences and Messages in my open, online discussion of sport informatics and analytics. We have great opportunities for conversation in our community of practice about diverse practices.

Photo Credit

Crystal web (Wendy, CC BY NC-ND 2.0)