3.1 Study Quality
Study quality was generally high, as indicated by the scores from the NIH's Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (Table 1); 13 studies were rated good, eight rated fair, and five rated poor. The main limitation amongst studies, which is common in sports research, was sample size. Especially in high-level athletes, researchers are often limited to samples of convenience, i.e. the one local team they have access to. To make up for this inherent limitation, the NIH recommends reporting effect sizes, confidence intervals, or some indicator of statistical power. Only 11 of the studies reviewed included some determination of statistical power, primarily using effect size; those which did not include any indicator of statistical power limited the veracity of their results. Another limitation of the current literature overall is the sparse amount of reliability testing done; only one study of a homogeneous group of international level French players has examined this quality of testing scrum force production [17]. In pilot testing prior to the main study, Lacome asked eight players to perform three individual maximal isometric scrums against a scrum machine; the scrums lasted 5 seconds, and 6 minutes of rest were given between repetitions. From this pilot testing, Lacome determined that maximal force could be reproduced under these conditions with an ICC = 0.8. Related to reproducibility and validity, some studies did not have subjects perform multiple scrum repetitions during testing to at least account for potential invalid trials, which may reduce the reliability of their results.
3.2 Demographic characteristics of samples
After applying inclusion and exclusion criteria, 26 research outputs were included. Of these, 20 were journal articles, four were theses or dissertations, and two were conference abstracts. Twenty five studies were conducted with rugby union players, and one did not specify the code of rugby. The majority were acute experiments (n = 11) or observational (n = 10), while four were cross-sectional comparison studies, and one was an intervention training study. Most studies sampled South African players (n = 6), with British (n = 5) and French (n = 5) teams also frequently sampled; three studies were conducted in New Zealand, two in Australia, and one study each came from the United States of America, Sweden, Canada, and Taiwan. Several studies were either cross-sectional or their sample included players from multiple playing levels without differentiation; thus, four studies sampled from international level players, 11 from elite/professional players, two from academy and semi-professional teams, 11 from amateur/community clubs, six from universities, six from high schools, and three from elite women’s teams. While the majority of studies did not specify the sex of their participants, we assumed all samples were male except for samples drawn explicitly from elite women’s teams. Most studies did not specify the time of season when testing occurred. Cohrane et al. [11] stated they conducted testing after the competitive season, Green, Kerr, Olivier, et al. [18] performed testing 4 weeks prior to the start of the inter-varsity tournament, Babault et al. [19] started their intervention 2 weeks after the mid-winter break, and Wu et al. [20] assessed players during the competition preparation phase. Sample sizes in each study ranged from only three players up to a study with 432 players drawn from 54 forward packs (and across five playing levels), though the majority of studies recruited fewer than 30 participants. Participants ranged in mean age from 16.6 to 34 years old, with the majority in their early 20s. Twelve studies used the full pack [3,4,7–9,12,14,15,21–24], while fourstudies used the front row [11,25,26] and two used the tight five only [10,13]. Two studies recruited a mix of forwards and backs (Dobbs 2017; Wu et al. 2006)[27]. Six studies did not specify the participants’ playing position [18,19,28–31]. Few studies quantified rugby playing experience of their subjects: Dobbs [27] had an inclusion criteria for university athletes to have played at least 6 months. Green, Dafkin, et al. [21] studied amateur players with an average of 11 years of playing experience, and Clayton [24] sampled players with an average of 8.4 years. The national players Wu et al. [20] studied had played for 9.1 years on average. No study reported the resistance training experience of their sample.
3.3 Equipment used for testing
Most of the studies (n = 21) had players scrum against a scrum machine [4,8,10–15,17–26,28,29,31]. Often the scrum machine was commercially purchased and then instrumented by the researchers, though some researchers built their own apparatus completely from scratch. Three studies used pressure pads or pressure sensor arrays affixed to players’ shoulders to capture forces during live scrums against an opposing pack [3,7,9]. One study had subjects push against a safety-squat bar and measured ground reaction forces using a force plate [27], while another study used a fixed yoke [30]. Nine studies had players perform scrums on natural grass turf [3,7,9,10,13,18,21–23], three used synthetic turf [11,24,29], two used rubber matting [12,14], and three were done indoors on lab or track floors [20,28,30]. One study had participants stand on a force plate, which the researchers had put skateboard tape on to reduce slippage [27]. Several studies did not report what surface players were tested on. Four studies had players support their feet on sprinter blocks or other wedges [10,19,28,30], though most studies just had players in shoes or cleats with no extra foot support. One study, however, had players go barefoot, though this choice was not explained [20].
3.4 Scrum test protocols
Only one study indicated that they provided or required subjects to perform a separate familiarization session [30]. Few studies provided details on the warm up used. Most studies utilized a general self-selected or coach-directed warm up (presumably of dynamic activities like high knees, butt kickers, jogging, etc) [3,4,9,10,15,23,25], and some studies provided athletes with warm up pushes on the scrum machine in addition to or instead of a general dynamic warm up [3,4,9,14,15,20,23,24,27–29]. The number of warm-up trials was not typically specified; the few that did stated that three to five submaximal trials [24] or eight 50% effort trials [29] were given.
Players were required to push for 5-10s in duration, usually performing two to five scrums per set [14,19,23,24,26,28,30], with a maximum of eights scrums per set [4,15,29]. In studies that evaluated scrums under multiple conditions (e.g. varying foot position, before and after a fatigue protocol), multiple rounds of two to five scrums were used [3,7,9–11,20,27,31]. In most cases players rested for 1-2 minutes between scrum trials [4,9–11,15,20,24,31], though some protocols restricted rest to only 15-24 seconds [14,25,30], and one allowed as long as 4 minutes between scrums [19]. In studies that tested the scrum under multiple conditions and reported between scrum-set rest periods, the rest between sets of scrums ranged from “at least one minute” to 10 minutes [3,4,9–11,20].
3.5 Body positioning in the scrums
The majority of studies allowed players to choose their positioning, though Clayton [24] reported that individuals were set up approximately 0.5 m away from the scrum pad, and Wu et al. [20] controlled the scrummaging height of players and tested them at 36%, 38%, and 40% of their body height. Most studies did not control or measure the joint angles of participants. Those that did had a wide variety of methodology. Cochrane [11] used a goniometer to set players into knee and hip angles of 120°; Dobbs [27] used a goniometer to set players into knee and hip angles of 90°; Morel [30] used starting blocks to control the position of the player’s feet, and set them far enough away from the yoke so that players pushed with approximate knee and hip angles of 130°. Other studies allowed players to self-select their scrum position, and the researchers measured the resulting joint angles. During the sustained pushing phase of the scrum, Green, Kerr, Dafkin, et al. [13] reported mean hip extension angles ranging from 152.6-155.0°, knee extension angles of 155-152.4°, and ankle angles of 70.5-74.6°, depending on the leg measured. Mensaert [26] measured players at professional, senior, and junior amateur levels during the engagement phase, and found hip extension angles ranging from 143-176°, knee extension angles ranging from 97-104°, and ankle angles of 90-93°. Quarrie & Wilson [14] found average angles within all forward positions of 123° of hip extension, 107° of knee extension, and 78° at the ankle. With these reported averages, it is worth noting that the standard deviations, i.e. inter-player variability, were sometimes quite high. For example, Green, Kerr, Dafkin, et al. [13] observed standard deviations of 19.6-20.5°, 17-23.9°, and 19.8-20.4° for the hip, knee, and ankle, respectively, while Mensaert and colleagues [26] found the range of hip extension angles to vary by 63-89°, depending on the playing level. The variation in joint angles both between studies and within studies highlights the difficulty in determining a so-called “ideal scrum position”, particularly when considering the large variety of body dimensions both across and within specific rugby playing levels.
Wu and colleagues [20] tested players in multiple positions (making shoulder contact at 40%, 38%, or 36% of body height), and found that professional rugby unionplayers produced significantly more force at 40% of body height (p < 0.01). However, it was interesting to note that there were no significant differences between the three scrum heights for hip, knee, or ankle angles. In their further analysis of what joint angles occurred at the highest pushing forces during a parallel stance, they found mean angles of 117.6° at the hip, 100.7° at the knee, and 53.9° at the ankle. An investigation by Green, Kerr, Dafkin, et al. [13] found similar results for both height and joint angles among university level tight five forwards. These players self-selected a position so their average back and pelvic height were at 49.2% and 41.5% of their standing stature, respectively, during the sustained push of the scrum. These positions changed little (less than 1%) across the different phases of the scrum. These players had slightly different joint angles in the left and right lower extremities, indicating they were not in a perfectly parallel stance, though likely were not in a purposefully staggered stance either. During the sustained push phase, mean hip extension ranged from 149.9-155.7°, knee extension ranged from 144.5-148.65°, and ankle angle ranged from 76.9-78.6°, depending on the leg measured. Furthermore, this study examined the correlations between the different examined body position characteristics and force production. They found that lower extremity joint angles did not significantly correlate with force production in any phase of the scrum; however, stance width, pelvic height, and back height did significantly correlate with force production, such that a lower body position and wider stance increased force production. Similarly, Quarrie and Wilson found no significant correlations between ankle, knee, and hip angle and individual scrum force [14]. In contrast, Bayne and Kat [10] found that knee angle had a large correlation and hip angle had a very large correlation with individual scrumming force, though they did not report body or pelvic height.
Aside from body position, the only individual scrum technique variation that has been examined is the use of a parallel or staggered foot position. The few research studies that have tried these two conditions have consistently found that the parallel foot stance allows for greater forward force generation against a scrum machine [10,20]. However, a staggered stance increases lateral forces, which may be of tactical importance or help with scrum stability [10]. The staggered foot position may also reflect the demands of a shifting scrum, as players reposition their feet. For whole-pack technique, the recent investigations have studied how opposing packs come together, with the goal of finding a technique that reduced peak engagement forces and increased the safety of the scrum. From the examined research, the consensus is that the entire pack should be fully bound onto each other before engaging the other team, as opposed to having the opposing front rows engage and then having the other five players engage (this was termed a staggered scrum engagement) in order to preserve scrum stability [7,8], and that the modern call of crouch-bind-set, where the props must reach out and maintain a full bind with their opposite prop, is effective at significantly reducing the peak engagement force [9]. Additionally, Hodge [8] mentioned that using a hip-bind technique for the locks to bind to the props may “indicate a safer technique for the prop” (page 36) than a crotch binding technique, though there was no difference in pack force production between the techniques used.
3.6 Force production in the scrum
3.6.1 Pack force production in live scrums
Reported average forces during live scrums can be seen in Figure 2a. One surprising result was that there was not an obvious trend for increase in average sustained pack force with increasing playing levels/combined pack weight (Figure 2a). The one study that tested forces in live scrums of high school players stands out in the graph as an unexpected outlier [7], with those players pushing with over 50% more force than the professionals. We double checked the manuscript to try to determine if there was some error in reporting, but the front row totals were consistent with the reported individual force production. Unlike Preatoni et al. [9] and Cazzola et al. [3], who calibrated their pressure sensors such that they would be optimized for the measurement of scrum forces [32], Du Toit and colleagues [7] failed to explain if a calibration process was used. Additionally, the average combined mass of opposing packs of high school players reported in Du Toi et al. [7] was 1361 kg , compared to an average combined mass of 1771 kg in elite [9] and 1708 kg in professional [3] players, making the finding that the high school players were both absolutely stronger and stronger relative to their mass highly unlikely. Therefore, while we cannot be entirely certain these are or are not plausible values for high school forwards, based on all the other data reported, it seems safe to say that this was an outlier. Ignoring this outlier, normalized average sustained force appears to increase with playing level (elite men were better than amateur and university men; Figure 2b) as would be expected, given the greater strength [16], lean body mass and overall body mass [33] observed with increasing playing level. We would expect this pattern to exist amongst female rugby players as well, given the high (53%) normalized average sustained force observed in elite female players, but unfortunately no data on non-elite female rugby players exist yet to confirm this.
3.6.2 Pack force production in machine scrums
While there was not as clear of an increasing force with playing level in live scrums (Figure 2a, 2b), when testing on a scrum machine, there was a clear trend for increases in average sustained pack force with increases in playing level/combined pack weight during machine scrummaging (Figure 2c). This trend remained when normalizing force by pack weight (Figure 2d). In a live scrum (Figures 2a and 2b), packs generally produce lower forces than they do against a scrum machine (Figures 2c and 2d). Assuming that the packs included in each of the studies summarized by Figures 2a and 2c are representative of their playing level, it appears that during live scrums, packs only exert about 50% of the force they produced on a scrum machine. While there was no direct correlation analyses between live and machine scrummaging forces, du Toit and colleagues conducted two studies with the same sample of high school players; one study measured full pack pushing forces during live scrums [7] and the other during machine scrummaging [12]. The resultant force during machine scrummaging (i.e. the force applied along the actual angle of push from the players) was 1.79-1.82x higher at max and mean, respectively, compared to the sustained force during live scrummaging.
The most dramatic differences between packs of different playing levels is seen in maximum force produced during the sustained pushing phase of the scrum when scrummaging against a machine (Figures 2e and 2f). The large absolute force differences between the men’s elite/professional packs and all other packs may be in large part due to greater mass (Figure 2e). Overall, there was a trend for greater maximum force and less variability (as indicated by smaller standard deviations) with increases in playing level and pack mass as expected. However, when normalized for pack weight, differences in maximum force during the sustained push phase greatly shrink between all playing levels except elite/professional, who still show a much greater ability to generate force. For example, the ratios of normalized maximum pack force between all comparisons of elite women, high school, amateur, and academy, range from 0.95-1.15, indicating that at these playing levels forwards are able to use their mass to generate force in the scrum to a similar extent. In comparison, the ratios of normalized maximum pack force between professionals and each other playing level range from 1.27-1.46, indicating that the players who make it to the highest level can use their mass much more effectively to generate forces in the scrum. These ratios hint at the likely impact of technique and other factors aside from pack mass on effectively producing pack scrum force. While relative forces can be useful for comparisons between ability level, it is important to note that in competition, it is ultimately the absolute forces produced that will dictate the movement of the scrum. Therefore, regardless of the relative forces produced by a pack, greater absolute forces are of primary importance: a pack able to generate greater absolute forces in the scrum has an advantage over an opposing pack with lower absolute forces but higher relative forces.
3.6.3 Relationship of other variables to scrum force production
In the research so far, total pack mass seems to be the greatest determinant of scrum force production. For example, elite women pushed less than semi-professional men, likely due to the lower mass of the women players rather than them being any lower in skill or body-weight-relative strength; indeed, in live scrummaging the women had a better normalized force production than any other group (except high school, which as previously discussed is likely an anomaly; Figure 2b). Studies that have correlated combined pack body mass to scrum force production have consistently found that the two factors are highly related [7,12–14]. Surprisingly, however, scrum force has not been unequivocally correlated with other measures of performance. For example, Green, Kerr, Dafkin, et al. [13] found no significant relationships between vertical jump height (both absolute and normalized to individual’s height and mass) and engagement (r = −.071, p = .738), peak (r = .084, p = .691) or sustained (r = −.072, p = .734) scrum forces among university players. Similarly, among males in the Dunedin premier rugby competition, average sustained force on a scrum machine did not correlate with vertical jump [14]. In contrast, Green, Dafkin, et al. [21] found that the scrum pack with the higher average vertical jump height of individual players in the pack won the scrum more often. An additional finding in the study by Green, Dafkin, et al. [21] was that combined pack body mass was not a significant determinant of winning and losing scrums. This was likely due to the similarity in mass between the two packs (difference in the average pack mass was only 4 kg). Furthermore, when looking at the individual scrums performed during the study and comparing the combined mass of the two packs, there were occasions when the winning pack weighed 75 kg less than the losing pack, demonstrating that while total body mass may allow for greater force production, greater mass does not guarantee scrum success.
3.6.4 Individual player contributions to total pack scrum force
Few studies have examined the contribution of individual players toward full pack force production. Those that have investigated this question have pursued it in two ways: 1) how much force does each unit (front row, second row/locks, back row/flankers and 8-man) contribute during a full pack scrum? And 2) do the forces of individual players scrummaging alone sum to the total force of a full-pack scrum? In answer to the first question, previous research has shown that players do not all contribute equally. Most studies indicate that back row players produce the least force in a full pack scrum, whereas front row players contribute the most force both due to their own force production and due to transfer of forces from the players behind them [7,12,22]. This is partly due to positioning, as flankers only bind on with one shoulder and therefore have less of a platform with which to transmit force straight forward into the pack unit. This disadvantage due to binding position is supported by the finding that in individual scrum trials (single players at a time) against a scrum machine, among amateur men’s players, the back row players did not create significantly less force, either absolute or relative to body mass, when compared to front row or second row players [23]. Du Toit and colleagues found that in both machine and live scrummaging, the front rows produce 40-51% of the average or maximum sustained pack force, locks produced 31-33% of these forces, and back rows contributed 18-27% [7,12]. These percentages are in line with a review Milburn published in 1993 [34], which indicated that the front rows contributed 42%, the second row 37%, and the flankers 25%. However, there is data to suggest very different contributions of the units within the pack. For example, Milburn [22] tested a high school pack on a scrum machine, starting with only the front row and then adding in players in subsequent trials. The front row produced 3290 N, while the full scrum produced 3370 N, indicating that the front rows alone could produce 98% of the scrum force. Additionally, adding the second rows made a minimal increase in total scrum force (310 N / 8 % increase), indicating a minimal rather than nearly equal contribution of the second row to the front row. It was surprising that the 2nd rows did not contribute as much force to the whole pack scrum, as 2nd rows often have as much mass as front rows and individually have been measured to produce as much if not more force than front rows [12,14]. However, based on the large swings in horizontal and vertical shear forces with each combination of units, it is possible that this single high school pack had trouble stabilizing with more players and effectively transmitting their forces in a synergistic manner [22]. Due to the much larger sample size in the Du Toit articles [7,12] (over 200 players), and the agreement with other reviewed research [34], it seems reasonable to conclude that front rows do produce the most scrum force (approximately 45%), with locks contributing approximately 35% and back rows contributing 20%. These data and summaries should be interpreted in light of the samples they represent (mostly high school) and that they were collected 10-40 years ago, and therefore may not represent the modern player, especially at different playing levels. Equally as important, these contributions need to be examined among female rugby players.
For examining if individual forces sum to total force during a scrum, Quarrie and Wilson [14] first showed that the sum of individual forces do not all get directed through the full pack by testing players on a scrum machine individually and then testing the whole pack on the scrum machine. They found that, on average, only 65% of the force produced by individuals is transferred into a scrum machine when the whole pack scrummed together. This is likely due to loss of force in other directions besides the forward direction (for example, lateral forces created by the angle of the flankers binding onto the props, or shear forces directed in the vertical axis), as well as general instability created by having eight players try to bind together rather than being able to optimally bind against a stationary and fully supportive scrum machine by themselves.Interestingly, while packs exerted 65% of the sum of their individual scrum results on average, there was significant variation between tested scrum packs (range: 52% to 74%). This variation indicates that there is substantial variability, even amongst professional scrum packs, in translating individual scrum performance to group scrum performance. This variability may indicate differences in individual and unit technique. Unfortunately, so far, no research has been conducted to operationally define individual or group technique, nor demonstrate a way to objectively measure it.
3.6.5 Force production and scrum success
Only one study compared forces of winning and losing teams [21]. Sixteen amateur players performed individual trials against an instrumented scrum machine to test their maximum individual scrum force. Then, players formed into packs using varying combinations of 8 players in their normal position (e.g. the two tight head props always had to tight head prop) and performed live scrums against each other. Winning was determined as pushing the other pack back about 1.5 meters; they did not put in the ball to contest actual possession. Green, Dafkin et al. [21] presented their data as the sum of individual scrum forces expressed as percentage of body weight (e.g. if each of the 8 player’s individual scrum force was 200% body weight, then the paper reported the summed value for the entire scrum as 1600% body weight). Green, Dafkin, and colleagues reported that there was an average difference of summed individual scrum force between winning and losing scrums of 182.1% body weight. To make results more comparable with other studies, we calculated absolute forces in Newtons using force data from Table 1 and scrum pack mass data from Table 2 [21]. We divided summed individual forces (in percentage of individual body weight) by 8 to calculate mean individual force (as percentage of body weight) then multiplied by mean pack mass to determine total pack force in kilograms, and finally multiplied by 9.81 to change the units to Newtons.The sum of individual machine scrum trial forces was 19,065 N for winning packs on average and 17,402 N for losing packs on average, for an average difference of 1,663N, or 9%, between winning and losing packs.
Summing individual scrum forces likely does not represent pack force production, because the sum of forces from individuals’ trials is greater than the forces produced during full pack scrummaging [14]. To try to account for this loss of force as well as how much each row in the pack contributes to total scrum force, Green, Dafkin, and colleagues [21] also calculated a pack total force by summing individual scrum force produced on the machine weighted for the percentage contribution expected from their position according to prior research. They performed separate calculations based on research from Du Toit et al. [12] and Milburn [22]. Du Toit [12] indicated that the front row contributes 42%, the second row 37%, and the back row 21% of total scrummaging force while Milburn found that the front row contributes 46%, second row 24%, and back row 30% of total pack force. Green, Dafkin, et al. [21] thus weighted player’s individual scrum machine results to create a weighted total. Using these position-specific scaling calculations, Green, Dafkin, et al. [21] found there was a 10-11% average difference in total force between winning and losing packs. Using the absolute force magnitudes we calculated from Green, Dafkin, et al. [21], and adjusting for the magnitude of force loss found by Quarrie and Wilson [14], the forces produced by winning packs in a live scrum are estimated as 1081 N or 9% higher on average (12,392 N for winning packs and 11,311 N for losing packs), closely matching the differences in percentage in row-weighted pack totals between winning and losing scrums [21]. While these magnitudes should be confirmed in future research, all these different methods of calculation indicate a 9-10% difference in force production between packs is associated with driving an opposing pack back 1.5 meters, which may provide an advantage in winning possession at the scrum.
3.6.6 Individual force production against a scrum machine
When testing players individually against a scrum machine for how much force they could produce, the expected trend of greater force with greater playing level was seen (Figure 3). However, especially when forces were normalized to individual player body mass, there was a lot of overlap between playing levels, especially for maximum force produced during sustained scrummaging (Figure 3d).
3.7 Fatigue during Repeated Scrums
Seven studies were found that examined the effect of fatigue, either from repeated scrums or other activities, on scrum force production. Lacome [17], Morel et al. [30], and Cochrane et al. [11] each examined the effects of repeated scrum trials on scrum force production. Lacome [17] had French National players perform twelve 5 second individual maximal scrums against a machine, with 15 seconds passive rest between trials (10 seconds standing, 5 seconds to get back into position). A statistically significant decrease of 11.7% of average scrum force was seen starting at the 4th scrum, and this lower force production was maintained through the twelfth repetition. Morel et al. [30] asked players competing in the U23 French championships to perform five 5 second maximal isometric individual scrums against a fixed yoke attached to a force sensor, interspersed with 20 seconds passive rest. In their analysis, they only compared repetitions one, three, and five, finding a significant decrease from trial one to three, and a significant decrease from trial one to five of a 23% reduction. In a follow up to their earlier study, Morel and Hautier [25] tested elite U-23 front row players on individual scrum force in response to fatigue. These players performed six maximal isometric pushes against a BabyScrum ergometer, each lasting six seconds. There were two inter-repetition recovery conditions, and each participant performed both in a two-session randomized crossover design. One condition was passive recovery (standing for 24 seconds), and the other was active recovery (run for 10s then get ready for next scrum for 24 seconds total). In this study, there was not a statistically significant decrease in average force across scrum trials, nor was there a difference between active and passive recovery conditions. Cochrane et al. [11] asked twelve front row players from academy, development, or semi-professional teams to perform three sets of five maximal effort isometric scrums, each lasting 10 seconds. Players were given 40 seconds rest between repetitions and 2 minutes rest between sets. When scrum trials within each set were averaged, there was not a statistical difference between sets one, two and three. However, there was a decrease in scrum force from the first rep to the third rep within set two and a decrease from the first rep to the third and fourth repetitions in set two, indicating growing fatigue across trials. Force data were only presented graphically, and therefore the percentage of loss experienced on the third repetitions of sets two and three can not be accurately estimated.
Other studies examined the effects of various fatigue protocols on scrum performance. Birch [28] examined the fatiguing effects of mental or physical stressors on scrum force. Ten local and/or national rugby union players attended two separate sessions. During each session, they performed a physical and mental fatigue protocol, with the order of protocols randomly determined in a crossover design. They performed one set of five repetitions of isometric maximal scrums before, between, and after the fatigue protocols. The length of rest between repetitions of the scrum test were not specified. The mental fatigue protocol consisted of 30 minutes of modified incongruent Stroop word tasks, and the physical fatigue protocol consisted of five sets of smith-machine squats to failure using 70% of the participant’s body weight, with 60 seconds rest between sets. Mean peak scrum force decreased by about 13% due to physical fatigue or combined physical and mental fatigue, though mental fatigue alone did not significantly reduce scrum force. Green, Kerr, Olivier, et al. [18] used the Bath University Rugby Shuttle Test (BURST) as a game-simulation protocol to estimate the effects of fatigue due to game play on scrum force in individual university forwards. The BURST protocol requires players to do 5 minute rounds of activity, including running, jogging, or walking and performing simulated contact situations such as mauls or tackles, with a very short rest period at the end of each round. In Green, Kerr, Olivier, et al.’s [18] study, university level players performed 16 cycles of the BURST protocol, divided into two halves with a 10 minute rest period in between to simulate a real game. For data collection, participants performed two repetitions of 6 second maximal isometric scrummaging against a machine before and after the game-simulation protocol, with the trial for each testing period in which they produced the maximal force kept for analysis. Contrary to their hypothesis, they saw no significant reduction in scrummaging force after the game simulation, despite other markers of fatigue (blood lactate and rate of perceived exertion) significantly increasing. In a different study, Green, Dafkin, et al. [21] examined individual scrum force before and after players from local amateur clubs performed 10 live full-pack scrums, and also found no significant reduction in force despite two subjective indicators of fatigue (a visual analog scale of fatigue and ratings of perceived exertion scores) significantly increasing (p < 0.001) from the start to the end of the experiment.