1. Introduction
LITA is a company specialised in physical performance, particularly in everything around speed: acceleration, changes of direction, maximum velocity and gym. We work closely with elite athletes, national football and rugby teams, and numerous top-flight football and rugby clubs in France. We also have a sprint group, LITAthletics, training elite sprinters for the Olympic Games, World Championships and European/African Championships. gpexe has been supporting us since 2022, in the build-up for the Paris 2024 Olympic Games for 2 of our athletes: Mickael-Meba Zézé and Hélène Parisot. Now, our eyes are on the Los Angeles 2028 Olympic Games with a new squad of athletes, and we are happy to share with you how we use gpexe to help us improve our athletes’ performances.
2. GPS Data Collection Process
Objective of GPS Utilisation
The integration of GPS technology aims to model each athlete’s sprint profile, identifying strengths and areas for improvement. Using gpexe’s high-precision devices, we analyse the performances of our top athletes, obtaining detailed data on their capacity for acceleration, maximum speed, maintaining that speed, and deceleration. We call this analysis ‘race modelling’. It allows us to create individual profiles, describing how each athlete achieves his or her optimal performance.
Sequence Segmentation
Methods used to segment the race to obtain data accurately reflecting the actual race :
Using the API and the information it provides by selection:
We use the API and its data to make precise selections based on raw speed. In the image above, the run distance in the selection is approximately 100 m, ensuring that our data collection closely reflects the actual race.
Using raw speed:
We decided to set the start of the race when the raw speed exceeds 0.5 m/s. To determine this threshold on an individual basis, we used the average speed of the hip during the initial thrust phase at the end of the starting block, setting the threshold at 50% of this speed, using 2D biomechanical analysis. This captures the actual start of the athlete’s acceleration, avoiding false detections due to insignificant movements. The choice of a fixed speed allows us to compare performances both intra-individually and inter-individually.
In the following example, the start of the race is set from a raw speed of 0.56 m/s because it is the first speed greater than 0.5 m/s in the selection. We delete the data before this start speed, then reset the time to 0 from line 4 (see the following table).
3. GPS Data Analysis (CSV Export)
- Race Modeling: We model speed over time for two types of runs based on distance: Speed (races from 30 to 100 m) and Speed Endurance (races from 100 to 250 m).*
Speed: This model applies to races between 30 and 100 m, resembling short sprint efforts like the 100-meter race. This modeling provides an average curve based on all maximal-effort races, whether in training or competition. On this curve, hatching represents the standard deviation at each point, indicating the athlete’s performance variability. This allows us to visualise potential speed variations according to the athlete’s condition and race factors.
Speed Endurance: This model covers races between 100 and 250 m, corresponding to prolonged sprint efforts, like those in a 200-meter race. Speed Endurance allows us to observe how the athlete manages speed over longer distances. This model provides a detailed view of the athlete’s ability to sustain high speed over time. Here, hatching also shows the standard deviation, illustrating expected performance ranges and the athlete’s capacity to maintain effort over distance.
- Curve smoothing :
To interpret the race modelling more finely, curve smoothing with an exponential function (coefficient of 0.05 to 0.3) is applied, reducing the noise due to undesirable fluctuations and giving a more homogenous view of performance. This smoothing improves the readability of the data, making it easier to observe the general trend and understand the athlete’s acceleration and speed characteristics. This technique also makes it easier to compare different races or athletes. By making the curves comparable and easier to analyse, we obtain valuable indications for adapting and personalising training programmes.
- Race segmentation :
To further enhance our analysis and focus training efforts on specific phases of the race, we segment the data into 10-meter intervals.
We have developed a ‘Split Time Calculator’ which, using raw GPS data (time, distance and speed), determines the exact time for each 10-metre interval by adjusting the total distance and applying a triangulation method for accurate estimation.
In this example, to determine the exact time at 10 metres, the calculator takes into account the time and speed associated with the previous point, allowing an accurate estimate of the time needed to reach the tenth metre.
The calculated time is 1.93 seconds (without accounting for reaction time).
This segmentation can be applied to a specific race (e.g., the French Championships) or to our race model, which is based on the average of recorded runs in our database.
4. Interpretation and Application of an Optimal Model
- Comparison of Split Times with Reference Athletes: Performance analysis in a comparative table
Collecting data segmented by 10-metre increments from the world’s best athletes enables us to assess our athlete’s strengths and weaknesses in comparison. This segmented analysis helps to break down the running phases, making training more personalised and enabling detailed progress monitoring.
For example, if our athlete’s split-time from 0 to 10m is identical to that of the reference athletes, this indicates that the start is one of his or her strengths and that the difference in performance probably lies in other segments of the race.
Here are the data for the selected reference athletes :
A comparison between the typical performance of our athlete and the average performance of these references :
- Constructing an Optimal Race Model: The aim is to reproduce the running dynamics of reference athletes by applying an optimal model adjusted to our athlete’s performance level, enabling us to obtain a target speed per segment for our athlete.
The data collected from the reference athletes enabled us to develop two distinct optimal models:
- Model A : This model, based on the average of the world’s best performances, provides a stable and realistic reference for setting performance targets.
- Model B: This model, specific to each athlete, is based on the average of their various races in competition.
Below is a comparison between our adjusted optimal model (A) and the typical profile of our athlete.
- Correlation with Performance : Analysis of the curve to understand the athlete’s strengths and weaknesses.
Analysis of the GPS data allows the athlete’s current performance to be compared with an optimal model established from the world’s best female sprinters, highlighting the athlete’s strengths and weaknesses for each 10-metre segment.
By superimposing the speed curves, it is possible to identify specific gaps in performance, such as a good explosive start but a need to improve speed maintenance. This correlation guides the definition of targeted training objectives for each phase of the race. To deepen this analysis, a biomechanical assessment examines joint angles, posture and the forces involved, to understand in detail whether the weaknesses identified are linked to technical aspects, a lack of strength or a coordination problem, enabling training sessions to be adapted precisely.
- Reaching Maximum Speed and Deceleration Coefficient :
We are also able to obtain specific data on maximum speed (Vmax) and deceleration coefficient, key indicators for analysing an athlete’s ability to maintain high speeds and their resistance to fatigue.
This also allows us to compare several runs completed in a short space of time and to assess which part of our athletes’ runs is most affected by fatigue.
Thanks to Excel formulas, it is possible to quickly calculate the time and distance associated with reaching Vmax, as well as the deceleration coefficient. This coefficient is determined by comparing the maximum speed with the final speed reached in the last segment of the race.
In addition, we apply two types of analysis depending on the segmentation of the data:
- Point-by-point : This approach tracks speed variations at each moment, providing a highly detailed analysis of performance.
The deceleration coefficient is calculated using the following formula :
Deceleration Coefficient=100−(Maximum Speed/Final Speed)×100
- Segmentation: Here, we analyse deceleration over distance intervals. The deceleration percentage is calculated using the highest speed achieved in the fastest split and the final speed in the last split (90-100 m).
The deceleration coefficient is calculated using the following formula :
Deceleration Coefficient=100−(Split with highest speed/Final Split)×100
A sprinter’s deceleration coefficient indicates his ability to maintain a high speed after reaching his peak speed, revealing his resistance to fatigue and the effectiveness of his technique. A lower coefficient means better speed endurance, essential for optimising sprint performance.
The point-by-point method offers fine precision and enables instantaneous variations in speed to be detected. In contrast, the 10-metre split analysis provides a more stable overview and is better suited to tracking the general deceleration trend, making it easier to compare key segments of the race.
5. Practical Applications and Decision-Making
Speed and deceleration analysis allows training to be personalised by precisely targeting the athlete’s needs, such as explosiveness or maintaining speed, making sessions more effective. Tracking performance curves over time helps to assess the impact of training, detect progress and any plateaus, and adapt targets to ensure continuous improvement.
Author: Lita Performance