Phibbs, PJ, Jones, B, Roe, G, Read, DB, Darrall-Jones, J, Weakley, J, Rock, A, and Till, K. Organized chaos in late specialization team sports: weekly training loads of elite adolescent rugby union players. J Strength Cond Res 32(5): 1316–1323, 2018—The aim of this study was to quantify the mean weekly training load (TL) of elite adolescent rugby union players participating in multiple teams and examine the differences between playing positions. Twenty elite male adolescent rugby union players (17.4 ± 0.7 years) were recruited from a regional academy and categorized by playing position: forwards (n = 10) and backs (n = 10). Global positioning system and accelerometer microtechnology was used to quantify external TL, and session rating of perceived exertion (sRPE) was used to quantify internal TL during all sessions throughout a 10-week in-season period. A total of 97 complete observations (5 ± 3 weeks per participant) were analyzed, and differences between positions were assessed using Cohen's d effect sizes (ES) and magnitude-based inferences. Mean weekly sRPE was 1,217 ± 364 arbitrary units (AU) (between-subject coefficient of variation [CV] = 30%), with a total distance (TD) of 11,629 ± 3,445 m (CV = 30%), and PlayerLoad (PL) of 1,124 ± 330 AU (CV = 29%). Within-subject CV ranged between 5 and 78% for sRPE, 24 and 82% for TD, and 19 and 84% for PL. Mean TD (13,063 ± 3,933 vs. 10,195 ± 2,242 m) and PL (1,246 ± 345 vs. 1,002 ± 279 AU) were both likely greater for backs compared with forwards (moderate ES); however, differences in sRPE were unclear (small ES). Although mean internal TLs and volumes were low, external TLs were higher than previously reported during preseason and in-season periods in senior professional players. Additionally, the large between-subject and within-subject variation in weekly TL suggests that players participate in a chaotic training system.