What is a T-Score?

T-Scores are displayed in the SMHP to improve the ability to interpret data. For user data presented in the form of population relative t-scores, a score of 50 is at the mean and a delta of 10 points corresponds to one standard deviation.

 

Normal Distro

The Short Answer:

There are a few reasons why T-scores are utilized to visor ualize and display data in the Sparta Platform. Each raw variable has different units and ranges, making it difficult to compare them to one another. For example, 5.7 is much lower than 4,358, but relatively speaking may be a much higher value when compared to normative data. Utilizing T-Scores allows us to understand significant and meaningful changes. For example, if 5.7 increases to 5.9, is that a larger (or smaller) relative change than 4,358 increasing to 7,690? Is either of those changes significant? How about meaningful? Perhaps the largest value of the Sparta Platform comes from this aggregated database of standardized data. This allows for data to be immediately meaningful as opposed to spending a few years collecting data in order for it to be useful.

It is also important to note that what is commonly referred to as "raw data," using our example from above of 4,358 N/s, is NOT the true raw data collected by the force plate sensors.  In order to provide different variables (ie. Eccentric RFD), calculations and algorithms are utilized to determine, for example, where the eccentric phase begins and ends.  Different systems can utilize different calculations to determine these variables as there is no single agreed-upon method. Unless the actual raw time-series data is collected and stored (which is rare) comparing variables from different systems is comparing apples to oranges. The true raw data which consists of 1000 raw force values for each second (1000 Hz) is rarely stored because the accumulation of this volume of data requires significant storage capabilities and software architecture.

The Long Answer:

Force plates, accelerometers, high-speed cameras, and global positioning systems give us thousands of different variables to measure and analyze. The easily understood data such as time (seconds) weight (lbs.) and distance (inches) unfortunately doesn’t show the validity that we would like. More advanced measures of force (N), velocity (m/s), and power (watts) show much more promise but are harder for practitioners and individuals to understand. These advanced concepts may be invaluable but most people will never have the knowledge (nor should they) of biomechanics or physiology required to truly understand them. We must be able to simplify and translate. As scientists continue to dive deeper, the ability to take action on this data will likely be limited by the ability to explain it to those who matter most.

Not only do we need to be able to explain these metrics in simpler terms, but we need to be able to immediately answer that first question we will no doubt get: “so… is that good?” To do this we need to collect enough data from across the population to understand the distribution and be able to identify what these norms are. This takes time. Unfortunately, we cannot simply compile massive amounts of unstandardized, subjective, unreliable data and expect to create a meaningful dataset. In computer science the saying “garbage in, garbage out” simply explains this concept that if the input data is flawed there is no amount of analyses that can be run to create a meaningful output.

For example, the subjectivity of squat depth will greatly influence the amount of weight that athlete is able to successfully lift. Simply type "Max Squat Testing" into YouTube and take a look at the wide range of squat depths that exist. Is my team better than yours because I have more 600 lbs squatters? Unlikely.

Attempting to compare these numbers across organizations is comparing apples to oranges, just as comparing hand-timed 40’s to laser times doesn't allow for accurate data interpretation. Little Johnny’s hand-timed 4.2 forty yard sprint on the track with blocks and spikes will hardly translate to an NFL players combine time. Even “normative” data taken from texts and articles can be dangerous as sample sizes are often small and methods can differ. For example, how you measure something as simple as vertical jump height (force plate, contact mat, Vertec) will greatly influence the results! Only with the standardization of equipment and protocols are we able to create data reliable enough that we can aggregate to find these norms.

Finally, we can answer our original question. Here at Sparta, we do this by utilizing a statistical tool known as a T-Score. The T-Score was first popularized when measuring Bone Mineral Density (BMD) to identify the risk of osteoporosis. The following is an expert from an article discussing the history of the T-Score:

 

“As bone density technology evolved, it became clear BMD expressed in raw units would be difficult to interpret. Ideally, for BMD measurements to be clinically useful, they should be presented in terms that are readily understandable by patients and clinicians, as well as independent of the densitometer used or the skeletal site measured.”

“Unlike common clinical measurements, such as blood pressure or cholesterol, the accepted normal values for BMD are not generally known. The T-score was suggested by researchers to simplify the interpretation of the bone density result and avoid the use of raw BMD values.” (1)


The challenges these practitioners faced are very similar to challenges in sports science today, instead of reinventing the wheel we can simply learn and apply, “standing on the shoulders of giants.”

A value of 50 represents the mean or average of the population, with 10 T-Scores in either direction (40 or 60) representing one standard deviation away from the mean. By normalizing data using T-Scores, practitioners can compare values between individuals and across populations using a standardized scale. The fact that these raw variables have different units and different scales doesn’t affect the ability to interpret the results, and MOST importantly quickly relay this information to those who matter most!

Without these T-Scores derived from a large and diverse database, the only way to evaluate scores would be within-individual percent change. Furthermore, while you could compare or rank individuals within a population there would be no way to know the potential biases of that specific population. For example, a group of elite basketball players might all have similar absolute scores for force plate variables, but these patterns may be a result of similar athletic background and training history. By incorporating a much larger and broader population, it is possible to better recognize strengths and weaknesses within a global movement assessment. While often overlooked, the power of Sparta’s database that provides T-Scores is one of the key advantages.

 

 

References:
1. Faulkner, Kenneth G. "The Tale of the T-score: Review and Perspective." (2005): 347-352.