Seiler's Hierarchy of Endurance Training Needs.

Abstract

This is an invited lecture that proposes a prioritization hierarchy of endurance training measures and focus, based on numerous sources of data, both descriptive and experimental.

Full text

*With apologies to Maslow
Stephen Seiler
University of Agder
Kristiansand, Norway
Seiler’s Hierarchy*
of Endurance Training Needs

What new «training technology» is currently
capturing your fancy? Beware of the Hype Curve

Frequency/ Volume of training (VOL)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Well
established
Strength of Evidence/Effect

Source: Vickers and Vertosick, BMC Sports Science,
Medicine and Rehabilitation (2016) 8:26
(From survey data of ~2500
recreational
runners)

0
100
200
300
400
500
600
700
800
900
Treningstimer
I-sone 1
103
211
344
377
374,5
469,5
533,5
550
565
582,5
656
648
633
623
I-sone 2
176
135
84
45
70
60
52
78
93
70
54
59
44
47
I-sone 3
36,5
21
18,5
12,5
18,5
17,5
18
16,5
19
25
23,5
23,5
25
20
I-sone 4
18,5
30
31
22
26
40
34,5
40
48
50,5
37
29
47,5
43
I-sone 5
1,5
0,5
3
2,5
1,5
1
2,5
2,5
9
16,5
5,5
12
18 år
19 år
20 år
21 år
22 år
23 år
24 år
25 år
26 år
27 år
28 år
29 år
30 år
31 år
Bente Skari
5 time World Champion,
O-gold, 46 WC victories
(65) (67) (65) (67) (69)
(67) (70) (76) (73) (71) (73) (73) (74)
(Numbers)= best VO2max test result each season
NA
Data from Espen Tønnesen
Olympiatoppen with permission
Training hours
AGE >

Sandbakk Ø, Holmberg HC, Leirdal S,
Ettema G. The Physiology of World Class
Sprint Skiers. Scand J Med Sci Sports.
2011 Dec;21(6):e9-16

Main differences in training were that the world class skiers trained ~30%
more volume (hours), and performed more specific speed work.

10

Stay Healthy!
Elite British runners lose 49 days a year
to injury, on average (source, Barry Fudge
British Athletics’ Head of Endurance)
Mostly due to:
Achilles injuries
Calf strain
Stress fractures
Chronic training volume
Injury
Risk?
Acute training volume/chronic training volume

Frequency/ Volume of training (VOL)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Training
Taper
Well
established
Well
established
Strength of Evidence/Effect

“When I was young, I was too
slow. I thought I must learn to
run fast by practicing to run fast,
so I ran 100 meters fast 20
times. Then I came back, slow,
slow, slow.” Emil Zatopek
Gerschler’s Sessions: The purpose of each effort
was to get the pulse up to about 180 beats per
minute. After the effort, 90 seconds of recovery was
allowed to get the pulse back down to 120
-
125 beats
per minute. Then the next effort was undertaken. If
the recovery took longer than 90 seconds, the effort
had been too fast.
Rudolph Harbig
1939 Reindell and Waldomer Gershler Emil Zatopek
1948ish
Despite what you read on the internet or in fitness magazines,
interval training has been around a long time!

‘‘It is an important but unsolved question which
type of training is most effective: to maintain a
level representing 90% of the maximal oxygen
uptake for 40 min, or to tax 100% of the
oxygen uptake capacity for 16 min’’
(Åstrand & Rodahl, Textbook of Work
Physiology, 1970). Per Åstrand in 1963

0
10
20
30
40
50
60
70
85 90 95 100
Tolerable Accumulated Duration (min)
% HF max
64min 32min 16min
I’m cooked

Adaptations to aerobic interval training: interactive
effects of exercise intensity and total work duration.
S Seiler, K Jøranson, B V Olesen, K J Hetlelid
Scand. J. Med. Sci. Sports 23:74-83, 2013
1975-2016

1. LIT only 4–6 sessions/wk. Increase training time 15-20%
2. 4 x 16 –2.wk-1: 4 x 16 min intervals + 2-3 LIT sessions.wk-1
3. 4 x 8 2.wk-1: 4 x 8 min intervals + 2-3 LIT sessions.wk-1
4. 4 x 4 –2.wk-1: 4 x 4 min intervals + 2-3 LIT sessions.wk-1
4 training groups
Intensity prescription: Complete each interval session with the highest
AVERAGE intensity possible for the entire session!
EFFORT MATCHING MODEL

Distribution of individual response to training by group-averaged change in
VO2peak (l/min), power at VO2peak (W), and Power at 4mM blood lactate
concentration (W).
Averaged response for each subject was categorized as negative to trivial:
<4% improvement, moderate: 4–9% improvement, or large:>9%
improvement. The distribution of individual responses was significantly
different among the four groups (P<0.05).
Seiler S, Jøranson K, Olesen BV,
HetlelidKJ. Adaptations To
Aerobic Interval Training:
Interactive Effects Of Exercise
Intensity And Total Work
Duration. Scand. J. Med Sci
Sports 23, 74-83, 2013.

Effects of intensity and duration in aerobic high intensity
interval training in highly trained junior cross country skiers
Sandbakk, Ø. et al.
J Strength and Cond. Res.
27(7):1974-80. 2013
Øyvind Sandbakk

Methods
•Well trained junior XC skiers
•N= 20- 8 weeks high volume, low intensity
•Then divided into a long interval group (LIG) and
a short interval group (SIG) plus a control group
(CG) for
•CG= 2 additional low intensity sessions/wk
•SIG= 2 interval sessions/wk 94% HR max (2-
4min), (15-20 min total work), RPE 18
•LIG = 2 interval sessions/wk 91% HR max (5-10
min), (40-45 min total work), RPE 18

-2 0 2 4 6 8
12km
rollerski
7km hill run
VO2 max
VO2VT
Percent change
LIG
SIG
CG
*
*
*

Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Training
Taper
Well
established
Well
established
Well
established
Strength of Evidence/Effect

25
Exercise Intensity (%HRpeak)
[La-]Z1 Z3 Z5
55 80 87 100
~2mM
VT1~4mM
VT2
Z2 Z4
5 intensity zones?
93
72
Z6+

26
Exercise Intensity (%HRpeak)
[La-]
55 78 86 100
2mM
VT1
4mM
VT2
3 intensity zones?
Z1 Z2 Z3
Z2 Z4

How do elite marathoners distribute
their training intensity?
2:06-2:11 marathoners
Training data collected during 12-week
period prior to Olympic trials marathon
Billat et al. Physical and training characteristics of top-class marathoners.
Med. Sci. Sports Exerc. 2001: 33: 2089-2097.

0
20
40
60
80
100
120
140
160
180
<Vmarathon Vmarathon v3000m /v10000m
Training intensity
Dsitance run in km
These top performers rarely trained at
marathon racing speed!
78%
4%
18%

Elite Kenyan 5-10k runners?
0
10
20
30
40
50
60
70
80
90
under vLT vLT Over vLT
Percent of weekly kilometers
Intensity
Data redrawn from Table 3 in Billat et al. Training and bioenergetic characteristics in Elite Male
and Female Kenyan Runners. Med. Sci. Sports Exerc. 35(2), 297-304, 2003.

Training Intensity Distribution
International Rowing Medalists
0
10
20
30
40
50
60
Training hrs/ month
70s 80s 90s
Basic endurance
High intensity
Fiskerstrand & Seiler, SJMSS 14:303-310, 2004

Seiler & Kjerland. Quantifying training distribution in
elite endurance athletes: is there evidence of an
optimal distribution? Scand. J. Med. Sci. Sports.
16, 49-56, 2006.
75
719
0
10
20
30
40
50
60
70
80
90
X < VT1 VT1 < X > VT2 X > VT2
% end. training sessions
Intensity Zone
Glenn Kjerland

0
10
20
30
40
50
60
70
80
90
100
Low Intensity Threshold High Intensity
Percent of Endurance Training
Duration
Annual intensity distribution of 12 Olympic/
World champions- XC skiing
Espen Tønnessen, Øystein Sylta,Thomas A. Haugen, Erlend Hem, Ida S. Svendsen, Stephen Seiler
The Road to Gold: Training and Peaking Characteristics in the Year Prior to a Gold Medal Endurance Performance. PLOS
One July 14, 2014 DOI: 10.1371/journal.pone.0101796

Physiological Exposure Time (PET)
65% VO2max

Frequency of Sessions (FOS)
0
10
20
30
40
50
60
70
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5
Percent of Sessions
Intensity Zone
75%
85%
90%
95%
65% VO2max

Esteve-lanao, Foster, Seiler, & Lucia. Impact of training intensity
distribution on performance in endurance athletes. Journal of
Strength and Conditioning Research, 2007, 21(3), 943-949.

2000
2050
2100
2150
2200
2250
2300
2350
Pre Post
10 K X-Country Race Time (sec)
80/10/10
65/25/10
*
* p< 0.05 vs 65/25/10 group
Esteve-lanao, Foster, Seiler, & Lucia. Impact of training intensity distribution on
performance in endurance athletes. Journal of Strength and Conditioning Research, 2007, 21(3), 943-949.
Jonathon
Esteve-Lanao

Six weeks of a polarized training-intensity distribution leads to greater
physiological and performance adaptations than a threshold model in trained
cyclists. CM Neal et al. J. Applied Physiology 114:461-471, 2013
Pre screening
and controlled
detraining
(4wk)
Baseline
testing
Polarized OR
Threshold
training
model (6wk)
detraining
(4wk)
Polarized OR
Threshold
training model
(6wk)
Testing
week Testing
week
Testing
week
Well trained cyclists training 7-8 hr/wk for >6 months prior to study
Age= 37, PPP 4.7 W.kg-1

Does polarized training improve
performance in recreational
runners?
Iker Muñoz1, Stephen
Seiler2, Javier Bautista1,
Javier España1, Eneko
Larumbe1,
ePublished ahead of print, Int. J. Sport
Physiology and Performance 2013

Group PET
(n = 15)
Group BThET
(n = 15)
Age (yr) 34 ±9 34 ±7
Weight (kg) 71.4 ±8.9 67.0 ±10.4
Height (cm) 177±5 173±7
BMI 22.7 ±2.4 22.2 ±2.2
VO2max (ml·kg-1·min-1)61.0 ±8.4 64.1 ±7.3
HRmax (beats·min-1) 182 ±11 187 ±8
VT2 (%HRmax) 91 ±391 ±3
VT1 (%HRmax) 77 ±3 79 ±5
Training Experience (yr) 7.0 ±3.2 5.6 ±3.5
10k PRE Time (min) 39.3 ±4.9 39.4 ±3.9

Group PET
(n =15)
Group BThET
(n =15)
Total Running Time (hr)39.1 ±7.9 36.3 ±8.1
Total Time in Zone 1 (hr)28.5 ±6.3* 16.7 ±6.2
Total Time in Zone 2 (hr) 5.3 ±2.7* 13.9 ±8.8
Total Time in Zone 3 (hr) 5.3 ±1.7 5.6 ±1.6
Total Time in Zone 1 (%) 72.9 ±5.6* 46.8 ±15.2
Total Time in Zone 2 (%) 13.5 ±5.6* 37.3 ±16.1
Total Time in Zone 3 (%) 13.6 ±4.3 15.8 ±4.1
Total TRIMPs 3299 ±670 3691 ±982
Mean TRIMPS/wk 330 ±67 370 ±98

-10 -5 0 5 10 15
10kPerformanceImprovement(%)
Group
BThET
PET
5%
3.5%

-10.0 -5.0 0.0 5.0 10.0 15.0
10kPerformanceImprovement(%)
Group
BThET
PET
higher%inZone2
thaninzone1
Zone2lower
than11%
mean intensity distribution 78%
Z1, 11% Z2 and 11% Z3.
10k improvement was
7.0±3.6%.
Six subjects from BThET 32%
Z1, 53% Z2 and 16% Z3.
mean 10k improvement was
1.6±4%

Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Training
Taper
Well
established
Well
established
Well
established
Strength of Evidence/Effect
General Periodization Details (Annual) Unclear but likely overrated
Training VOL,
HIT, and overall
TID likely have
interactive
effects

International Journal of Sport Physiology and Performance
2012, 7, 242-250

Cool Periodization Names
•Reverse
•Block
•Non-linear
•Fractal
•Conjugate Sequence

Lots of scientific sounding ways to
say…..
•Variation?
•Variation?
•Variation?
•Variation?
•Variation?
(But not TOO much)

Basic periodization of a champion XC skier
0
10
20
30
40
50
60
70
80
90
100
Training sessions/ hours
Training hours
Total sessions
IZ3-5 sessions
Espen Tønnessen

Oktober November Desember Januar Februar Mars April Mai Juni Juli August
Treningstimer 88,22 101,68 89,13 123,78 95,92 95,02 92,52 101,77 84,57 110,23 58,37
Treningsøkter 56 58 57 53 53 52 51 61 53 52 52
0
20
40
60
80
100
120
hours/ number
Basic training periodization of a
2-time gold medal winning rower
Annual Total (gold medal year)
Total training hours: 1041
Total training sessions: 598
Rowing kilometers: 4911
Average hours per week: 21.7
Average # sessions per week 12.5

0
5
10
15
20
25
30
Oktober November Desember Januar Februar Mars April Mai Juni Juli August
Sessions
I-sone 5
I-sone 4
I-sone 3
Typical effective training time per session:
Zone 3: 60 min (e.g. 3 x 20 min, 5 min rec)
Zone 4: 60 min (6 x 10 min, 5 min rec)
Zone 5: 20 min (e.g 4 x 4 min, 3 min rec)
High intensity training sessions- champion rower
Data from Espen Tønnesen
Olympiatoppen with permission

Annual training characteristics of 12 champion XC skiers
Tønnessen E, Sylta Ø, Haugen TA, Hem E, Svendsen IS, Seiler S (2014) The Road to Gold: Training and Peaking Characteristics in
the Year Prior to a Gold Medal Endurance Performance. PLoS ONE 9(7): e101796. doi:10.1371/journal.pone.0101796
http://127.0.0.1:8081/plosone/article?id=info:doi/10.1371/journal.pone.0101796
Frequency of hard
sessions stable, but
intensity of HIT sessions
shifts towards more
Zone 5. Overall, training
becomes MORE polarized
near and during
competition season.
Overall intensity
distribution maintained
through season

0
10
20
30
40
50
60
70
Treningstimer
I-sone 1
34,00
38,00
46,50
46,50
41,50
31,00
30,50
35,00
30,50
29,00
29,50
16,50
I-sone 2
4,00
7,00
8,00
7,00
5,00
5,00
1,00
3,00
2,00
3,00
4,00
3,00
I-sone 3
0,50
1,00
1,50
3,00
3,50
4,00
1,50
1,00
1,50
1,50
3,00
3,00
I-sone 4
1,00
1,50
2,00
2,00
2,00
1,00
1,00
1,50
1,50
1,50
1,00
I-sone 5
1,00
1,00
0,50
0,50
2,00
2,00
1,50
1,50
0,50
November
Desember
Januar
Februar
Mars
April
Mai
Juni
Juli
August
September
Oktober
Preparation Competition
Ingrid Kristiansen
5 World Records
World Champion
Data from Espen Tønnesen
Olympiatoppen with permission

Contributors to a
multicenter training study

2016 Innovation Award winning study!
Øystein Sylta, PhD student- University of Agder
Kristiansand, Norway

There were no significant differences among groups. One-way between-groups ANOVA
analysis.
All
(N=63)
Traditional
(N=23)
Hybrid
(N=20)
Reversed
(N=20)
Age (years)
37.7 (7.8) 37.3 (9.3) 38.2 (6.8) 37.7 (7.2)
Body weight (kg)
79.9 (7.9) 80.3 (7.4) 79.7 (8.9) 79.7 (7.8)
Cycling experience
(years)
5.5 (4.2) 5.8 (4.5) 6.3 (4.1) 4.4 (3.8)
Training volume last
year (h
.wk-1)
9.7 (3.3) 9.9 (3.5) 9.3 (3.1) 9.7 (3.4)
Power@40min TT
(w
.kg-1)
3.6 (0.4) 3.5 (0.4) 3.6 (0.4) 3.6 (0.5)
Power@4mM (w
.kg-1) 3.6 (0.5) 3.5 (0.4) 3.6 (0.3) 3.6 (0.6)
PPO (w
.kg-1) 5.3 (0.6) 5.2 (0.5) 5.4 (0.7) 5.2 (0.7)
VO
2 PEAK (ml.kg-1.min-1) 61.3 (5.8) 61.9 (5.8) 61.3 (6.0) 60.5 (5.7)
VO
2 PEAK (L.min-1) 4.9 (0.5) 4.9 (0.5) 4.9 (0.5) 4.8 (0.4)
Baseline characteristics in all training groups (N=63). Values are mean (SD).

Interval session prescription
Zone
Session
Intensity
control
Zone 3 4 x 16min, 2min rest
All HIT
sessions were recommended to be executed
as «all
out», BUT:
1
- Within the session structure (eg. only 2min rest)
2
–With even or progressive power, without failure
3
–Be able to complete all 24 HIT sessions in this
prescribed
way
Zone 4 4 x 8min, 2min rest
Zone 5 4 x 4min, 2min rest

4x16
(N=63)
4x8
(N=63)
4x4
(N=63)
Power (w)
276 (25)* 308 (29)* 342 (33)*
Power (w
.kg-1) 3.5 (0.4)* 3.9 (0.4)* 4.3 (0.4)*
Power
% 4mM (%) 97 (8)* 106 (8)* 118 (9)*
Blood lactate (mmol
.L-1) 4.7 (1.6)* 9.2 (2.4)* 12.6 (2.7)*
HR
mean (% HRpeak) 86 (3)* 88 (2)* 89 (2)*
HR
max (% HRpeak) 89 (2)* 91 (2)* 94 (2)*
RPE average
15.0 (1.1)* 16.2 (0.8)* 17.1 (0.9)*
RPE
peak 16.6 (1.1)* 17.7 (0.8)* 18.6 (0.7)*
sRPE
(1-10) 30min post session 6.3 (1.0)*6.9 (1.0)*7.7 (1.2)*
HIT training characteristics of sessions in all subjects during the
intervention period.
*P<0.001 vs other groups, one-way repeated measures ANOVA.

0
5
10
15
Wk 2 Wk 3 Wk 4 Wk 5 Wk 6 Wk 7 Wk 8 Wk 9 Wk 10 Wk 11 Wk 12 Wk 13
Training volume (h .wk-1)
Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Other
Volume and intensity periodization - ALL GROUPS
6
7
8
TIRED
RESTED
No differences among groups. One-way between-groups ANOVA analysis.

0
1
2
3
Wk 2 Wk 3 Wk 4 Wk 5 Wk 6 Wk 7 Wk 8 Wk 9 Wk 10 Wk 11 Wk 12 Wk 13
HIT volume (h .wk-1)
HIT periodization- TRADITIONAL group
Zone 3 Zone 4 Zone 5

0
1
2
3
Wk 2 Wk 3 Wk 4 Wk 5 Wk 6 Wk 7 Wk 8 Wk 9 Wk 10 Wk 11 Wk 12 Wk 13
HIT volume (h .wk-1)
HIT periodzation –HYBRID group
Zone 3 Zone 4 Zone 5

0
1
2
3
Wk 2 Wk 3 Wk 4 Wk 5 Wk 6 Wk 7 Wk 8 Wk 9 Wk 10 Wk 11 Wk 12 Wk 13
HIT volume (h .wk-1)
HIT periodization- REVERSED group
Zone 3 Zone 4 Zone 5

RESULTS

Chi-square test for independence indicated no significant association between periodization groups and response distribution.
No clear evidence that specific HIT periodization order makes a predictable difference,
but no reason to leave the approach you are using if it is working either!

Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Training
Taper
Well
established
Well
established
Well
established
Strength of Evidence/Effect
General Periodization Details (Annual) Unclear but likely overrated
Sports-specific and
micro-periodization schemes Not established, but likely modest
Training VOL,
HIT, and overall
TID likely have
interactive
effects

M T W T F S S HI Th LI
Highly
Trained
AM
LI
LI
Rest
LI
HI
Th
LI
3
1
7
PM
HI
LI
HI
LI
Rest
LI
Rest
Highly
Trained
AM
LI
LI
HI
LI
LI
Th
LI
3
1
7
PM
HI
LI
Rest
HI
Rest
LI
Rest
Does this change make a consistent difference
in the adaptive response?
Redrawn after an example provided by J. Stanley
et al. 2013 Sports Med. and lecture video.
There is some new evidence supporting block training schemes,
and microcycle manipulation may make a difference over
time, but little research has been done on this so far.

Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
General Periodization Details (Annual)
Sports-specific and
micro-periodization schemes
Training Stimuli
Enhancement
(i.e. Altitude, Heat)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT) Well
established
Well
established
Well
established
Unclear but likely overrated
Not established, but likely
modest
Potentially important effects
but individual and condition specific
Strength of Evidence/Effect
Training VOL,
HIT, and overall
TID likely have
interactive
effects

Almost all International medal winning endurance athletes have done some training
at altitude (typically 2000-2500m). So, optimization of blood oxygen delivery by stimulating
enhanced red blood cell production may be critical at the world class level. BUT, this is the last
few percent, not the foundation of training, AND there is big individual response variation
and substantial risk of a negative training outcome.
Young athletes and their coaches should do all the other things right, and build a good
endurance training platform before investing lots of time and money on altitude camps.

http://www.csipacific.ca/wp-content/pdfs/pp/performance-point-phys-1305-heat.pdf
Heat acclimatization is very important if competitions are held under hot conditions.
Fortunately, a 5-7d acclimatization period is consistently shown to improve
heat removal capacity and heat tolerance. However, effects of «heat adaption» on
performance under normal ambient conditions are probably trivial.

Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
General Periodization Details (Annual)
Sports-specific and
micro-periodization schemes
Training Stimuli
Enhancement
(i.e. Altitude, Heat, Energy availability)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Race/
Pace Training
Training
Taper
Well
established
Well
established
Well
established
Unclear but likely overrated
Not established, but likely
modest
Potentially important effects
but individual and condition specific
Potentially decisive
if everything else is done right
Strength of Evidence/Effect
Training VOL,
HIT, and overall
TID likely have
interactive
effects

Pacing and racing are sometimes very different. Optimal pace for a given distance
requires both sensitivity to perceptual signals, and practice. Pack racing does not always
involve doing what is «optimal» for the fastest individual time. Therefore specific
«race» training may require a different strategy……
However, both pacing for optimal performance time and race training for optimal race
placement both build on your maximal capacity, which is a function of your committment
To the lower levels of the training hierarchy!

2008
2012
2016
Percent
Change
from 2008
5000m SB
13:09
12:56
12:59
1
-1.5%
Last 1k
2:27.5
2:25.2
2:25.6
1.3%
Last 400m
56.1
52.9
52.8
5.9%
Name this racer…

Total Frequency/ Volume of training (VOL)
Training Intensity Distribution (TID)
General Periodization Details (Annual)
Sports-specific and
micro-periodization schemes
Training Stimuli
Enhancement
(i.e. Altitude, Heat, Energy availability)
Seiler´s Hierarchy
of
Endurance Training Needs
High Intensity Training (HIT)
Race/Pace
Training
Training
Taper
Well
established
Well
established
Well
established
Unclear but likely overrated
Not established, but likely
modest
Potentially important effects
but individual and condition specific
Potentially decisive
if everything else is done right
Potentially decisive if you have one isolated competition...
and everything else is done right
Training VOL,
HIT, and overall
TID likely have
interactive
effects
Strength of Evidence/Effect

Showing up fit for fight on race day can be worth 0.5-3%, and a lot of research
has been done on the tapering process. Here is a nice article that discusses
that contemporary research, but also some other views (if it ain’t broke don’t fix it).
http://www.runnersworld.com/race-training/the-art-and-science-of-the-taper