How to Combine Internal & External Training Load Monitoring

Summary

The video discusses approaches to combine internal and external training load metrics to gain insights into player fitness and fatigue without additional testing. It covers using simple ratios of the loads and more complex regression models to predict expected internal load based on external load. The video highlights research on training efficiency indexes that scale the loads for comparison. It advocates for individualized analysis, noting machine learning techniques that find different external load metrics most predictive of internal load by player. A case study shows patterns of higher actual than predicted internal load, indicating potential deconditioning to flag for intervention.

Takeaways

• 😀 Combining internal and external training loads may provide insights into fitness and fatigue
• 📊 A simple approach is to calculate the ratio between internal and external loads
• 🔬 Ratios may correlate with fitness measures but become weaker under fatigue
• 🧪 Uncoupling of loads could indicate fatigue without direct testing
• ⚖️ Issues can arise from numbers being on different scales, the training efficiency index aims to address this
• 📈 Training efficiency index plots loads on a logarithmic scale to enable comparison
• 😕 Ratios may oversimplify the complexity of training sessions
• 🤖 Machine learning is being used to relate loads on an individualized basis
• 😃 Case study: Compared actual and predicted internal load based on external load
• 💡 Flagged potential issues through this "invisible monitoring" approach

Q & A

What are internal and external training loads?

Internal training load refers to the physiological stress imposed on an athlete, often measured by heart rate. External training load refers to the work completed by the athlete, often measured by metrics like distance covered or accelerometry.

Why might combining internal and external loads be useful?

Combining internal and external loads may provide more insight into an athlete's fitness and fatigue status. This 'invisible monitoring' allows coaches to assess fatigue without extensive additional testing.

What was the first study to investigate ratios of internal to external loads?

A study by Ebi Acuba and his team first looked at ratios of individualized training impulse (iTRIM) to external load metrics like total distance. They found these ratios correlated with measures of fitness in soccer players.

What did Acuba's next study on internal/external load ratios find regarding fatigue?

In their next study, Acuba's team found that the relationships between internal/external load ratios and measures of fitness became weaker under fatigue. This suggests uncoupling of loads could indicate fatigue.

What issue needs to be considered when creating ratios of internal to external loads?

Internal and external loads use different scales - heart rate may be around 300, while distance could be thousands of meters. The numbers need to be scaled before taking a simple ratio.

How does the training efficiency index address the scaling issue?

The training efficiency index plots internal and external loads on a logarithmic scale before taking the ratio. This scaling allows the comparison of different load metrics.

What does the case study example demonstrate regarding internal/external load analysis?

The case study shows how regression equations can forecast athlete's internal load based on external load. Comparing predicted and actual internal load over time can indicate changes in fitness and fatigue.

What machine learning approach shows promise for relating internal and external loads?

A machine learning model developed by John Bartlett demonstrated the need to relate loads on an individualized, athlete-by-athlete basis, with different metrics being most predictive for each athlete.

What is the overall message regarding approaches that combine internal and external training loads?

There is no clear consensus, but analysis should be done on an individualized basis for each athlete to relate the right external metrics to internal load response.

What is the topic of the next video mentioned?

The next video will continue the discussion of invisible monitoring - using existing data to monitor fatigue without additional testing.

Outlines

00:00

🏃‍♂️ Defining internal and external training loads

The paragraph introduces the concepts of internal and external training loads. It mentions a previous study by Impellizzeri et al. that found ratios between internal and external loads correlated with fitness measures in athletes. However, these relationships weakened under fatigue, suggesting potential use of internal/external load ratios to detect fatigue.

05:02

😕 Criticisms and limitations of using ratios

The paragraph discusses common criticisms and limitations of using simple ratios between internal and external loads, like oversimplification. It argues we likely need more complex, individualized approaches relating varied external load metrics to internal load. A case study is shared on forecasting internal load for each athlete from their external loads.

10:02

🤖 Machine learning approaches relating loads

The paragraph mentions recent studies using machine learning to relate internal and external loads. It highlights Bartlett et al.'s work showing the need to take an individualized approach. The conclusion is that there are various ways to combine loads for "invisible monitoring", detecting fatigue from normal data.

Keywords

💡internal load

Internal load refers to the physiological stress imposed on an athlete's body during training or competition. In the video, it is quantified using heart rate exertion. Understanding internal load is key because it represents the athlete's physiological response to the external workload.

💡external load

External load quantifies the work completed by an athlete during training or competition, such as distance covered or high-speed running. Monitoring external load is important to understand the training stimulus and properly dose it for each athlete.

💡ratios

Ratios refer to simple models dividing measures of internal load by external load. The video discusses research using these ratios to detect fatigue when internal and external loads become uncoupled.

💡invisible monitoring

Invisible monitoring means estimating fitness or fatigue from existing load data, without needing extra fitness tests. This is appealing as it provides more insight without added athlete burden.

💡individualized

An individualized approach accounts for differences between athletes in the relationship between internal and external loads. This is emphasized in the video and studies showing players have different predictive load metrics.

💡regression

Regression refers to statistical models predicting internal load from external load metrics. The case study demonstrates this for individual players over a season to flag potential fatigue.

💡machine learning

Machine learning is an advanced statistical approachusing algorithms and artificial intelligence to model complex data. Recent studies are applying it to model the internal-external load relationship.

💡fitness

Fitness represents an athlete's conditioning and preparedness for optimal performance. Exploring the coupling of internal and external loads may provide insight into fitness without needing direct testing.

💡fatigue

Fatigue indicates declines in preparedness and impaired physiological state. Differences in internal vs external load may manifest as fatigue develops, even without performance changes.

💡conditioning

Conditioning means improving fitness through systematic training. In the pre-season period, athletes are deconditioned, resulting in higher internal loads for a given external load.

Highlights

Internal and external training loads can potentially be combined to gain insights into fitness or fatigue

When internal and external loads uncouple, it may indicate fatigue

Using load ratios to detect fatigue without additional testing is appealing and some studies show promise

Issues with load ratios include numbers being on different scales and oversimplifying complexity

The training efficiency index scales loads by plotting them on a logarithmic scale before calculating ratio

Criticisms of ratios - they may oversimplify, need more external metrics related to internal load

Case study: Compared predicted vs actual internal load based on regression of previous external load

In preseason, actual internal load was higher than predicted which matches deconditioning

During season actual internal load matched or was below predicted showing adaptation

Later in season internal load rose above predicted, flagging potential detraining

This analysis provided invisible monitoring to identify needs for individual players

Regressions simplify complexity, machine learning is now exploring internal/external relationships

Machine learning study showed need for individualized predictive modeling of loads

No consensus on best way to combine loads but should be individualized

Other ways to explore invisible monitoring by relating internal and external loads

Transcripts

so in my internal versus external load

video we Define these two different

types of training load and if you're not

comfortable with what they are then

check out that video first because now

what we're going to talk about is

potential ways that we can combine these

two measures to try and get some sort of

insight potentially about Fitness or

fatigue

we don't want to keep adding data

collection processes or testing to our

athletes so if we can combine these

measures to get more insight into our

athletes then that is a bonus for

everyone

so the simplest approach may be just to

look at a ratio between internal load

and external load and that is an

approach that ebi acuba and his team

previously looked into so this is their

first study on this topic and they took

individualized training impulse or eye

trim which is a heart rate so an

internal load measure from heart rate

and blood lactate profile and they

divided different measures of external

loads so total distance or high

intensity distance into this eye trim

for the different athletes to get their

ratios and they use this in this first

study with a soccer simulation

and then they found that these ratios

did significantly correlate with the

measures of Fitness that they also took

on the athletes

so they established those relationships

with measures of Fitness but what about

fatigue and in this the next study on

the topic they actually found that these

ratios the relationships became weaker

under fatigue now that is really

interesting because actually when

internal and external loads uncouple

then that may be a way that we can

measure fatigue and do some sort of

either fatigue testing without testing

which is often known as invisible

monitoring and this invisible monitoring

getting more out of the data that we are

already collecting without having to do

Fitness testing is really appealing to

practitioners and so we do see other

studies that are trying to use these

ratios such as this study by Conor

durbridge looking at the ratios during

standardized small gain to try to

determine if they can be used as a

detector of fatigue so there is perhaps

some promise in these ratios even if

they are very simplified and I'll come

back to that in a moment one of the

issues that we sometimes see with these

kinds of approaches is that actually the

numbers are on different scales so if we

think about a heart rate exertion number

maybe that is somewhere around 300 say

but if we're looking then at total

distances our measure of external load

that could be three thousand five

thousand eight thousand we want to be

able to scale for the different numbers

and there is a ratio again it is a

deemed an efficiency ratio that tries to

take this into account

so here in this article in science and

medicine in football Jace Delaney

presents his work comparing internal and

external loads and this is known as the

training efficiency index

though it is a ratio comparing internal

and external load because of this

scaling issue what Jace does is he plots

the internal and external loads on a

logarithmic scale and that then scales

these the numbers and then the slope of

the line gives us the scaling number

that we divide into the internal load

now this means whatever measure you are

using you will always be able to scale

it to compare against each other and

he's seen with some of the studies that

he's published some strong evidence in

support of this approach now Jace

actually used to on his blog provide a

calculator I don't think it's currently

available but what I'll do is I'll link

to a video in which Jace is explaining

this concept in more detail so if it is

something that you want to calculate for

yourself

and explore in your own sport or your

own population then you'll be able to do

that along with Jace's instructions

now as I said invisible monitoring is

really appealing but let's pause for a

moment and think about these ratios we

are taking two very

overarching

numbers that are simplifying the

complexity of a training session and

then dividing them into some something

and often this can be a limitation with

a ratio that we are over simplifying the

process and there has been a lot of

discussion and criticism of the use of

ratios most notably of course the acute

chronic workload ratio and perhaps if

I'm feeling brave one day I will do a

video that addresses that topic but

these ratios have come under criticism

are they oversimplified

in reality what we probably want to do

is be able to take a whole mixture of

external lobe metrics for an individual

athlete to try and assess the

relationship with internal load and I

want to share a case study of one way

that I've approached this previously in

the applied setting we took a season's

worth of internal and external load data

internal load being heart rate exertion

and the external load coming from player

tracking metrics notably GPS and

accelerometry derived measures and for

each individual athlete we determined a

regression equation

that for that individual

predicted or forecast the internal load

based on their external load so

each individual athlete had a different

formula that was based on last season's

data and going forward in that season

every day based on the various

combinations of external load outputs

from the session we would get an

estimated value for heart rate exertion

that we could then compare to what they

actually had and let me show you an

example of that data from one athlete

so here we can see in the blue line for

this individual athlete over roughly a

six months period

they're predicted

heart rate exertion value in blue So

based on their external load output from

each session what was predicted as their

internal load and we've plotted that

then against the actual heart rate

exertion that was seen

and of course there aren't

um

set points where someone suddenly goes

from adapted to non-adapted but I think

there are some general Trends here which

I'm gonna highlight so it's absolutely

not this clear-cut but just for ease of

viewing what we see here we in this is

the pre-season training period so

particularly early on but quite

consistently across this period we're

seeing the actual heart rate exertion

for this individual is higher than what

is predicted

now that is totally what I would expect

to see because we based these

regressions on a season's worth of data

and in the pre-season period players are

deconditioned and therefore

expect to have a higher internal load

physiological cost to the same amount of

external load the same work done so

that's a pattern that I would expect to

see

and then we move into the season

and again we see very little times where

the actual is higher than the predicted

if anything is lower or quite frequently

around about the same amount but then we

can see this shift and this trend here

in October November December period

where it reverts back to consistently

having a higher actual heart rate

exertion than what was predicted

now in that case that athlete was

someone who was a uh a non-starter and

although they were getting Top-Up work

this an analysis was a red flag to us

that something in their Fitness fatigue

status was changing it was not ideal and

actually they weren't getting the Top-Up

conditioning that that athlete needed to

maintain their level of conditioning

that they had earlier in the season so

this analysis served to provide a red

flag to us for the need to explore

further with that individual and again

this was all done essentially through

this invisible monitoring process

whereby we were just trying to use the

data we were already collecting as a bit

of a fitness test

in reality your regression approach is

is essentially quite a simple

statistical approach and of course now

we are seeing more and more approaches

with machine learning and artificial

intelligence and there's been a couple

of studies now start to use machine

learning to assess the relationship

between internal and external load

now one being this paper again Bartlett

John Bartlett's paper we talked about in

the internal external load video and if

the most important message from this

paper is it highlights the need for an

individualized approach so again

session distance for instance was most

predictive of rpe in 36 or 41 players

whereas high speed running was more

predictive in three players and meters

per minute in two players so again

whatever approach you take it's

important that you're looking at players

on an individualized basis

so a few different approaches there in

terms of combining internal and external

load for you to play around with no real

clear consensus other than it should be

done on an individualized basis but

there are other ways that we can try and

encapsulate these this Theory or

concepts sorry of invisible monitoring

and that will be the topic of my next

video