KINESIOLOGIASLOVENICA3 (1997)1 : 3-7 3 THE VALIDITY OF A PREDICTION MODEL OF COMPETITION PERFORMANCE IN DOWNHILL WHITE WATER KAYAKING AntonUšai * VELJAVNOST NAPOVEDI MODELA TEKMOVALNE ZMOGLJIVOSTI V KAJAKU NA DIVJIH VODAH - SPUST ABSTRACT To ascertain whether pred icted competition results in a white water downhill sports event are val id if they are predicted by usingtests on fiat water, 11 top ranked national level kayakers performed three tests, twice in the period of two international com- petitions (one month). The predictors were: pad- dling speed and heart rate determined by Onset of Blood LactateAccumulation (v08LAand HR0 8LA), av- erage speed, HR and Lactate concentration in the test over 4000 m (v 4000, HR4000, LA4000) and 500 m (v500, HR500, LA500) . A combination of VosLA, V4000 and v500 was selected as the best prediction combi- nation. This combination predicted a competition tirne of 902±20.5 s in the first experiment, similar to the real competition results (903±20.6 s). In the second experiment, the calcu lated competition re- sult (862±45 s) is also sim ilar to the real one (862 .8 ± 48.4 s). Regardless of this precise prediction of the specific competition results, it is not possible to use each of both prediction models as universal in order to calculate competition results in both expe- riments with satisfactory accuracy and precision. The very different duration of competitions caused an unpredictabilityofthe universal model. lrrespective of this flaw, which was characteristic for both of the models, the basic princi p les by wh ich the models predicted competition results were verysimilar. This general characteristic may be used for estimation of qualitative changes in performance during the com- petition season. Key words: kayak, white water, downhi/1, predictive tests, multiple regression mode/s • faculty of Sport, Universily of Ljubljana, Ljubljana, SLOVENIA IZVLEČEK Enajst najboljših kajakašev, državn ih reprezentantov v spustu na divjih vodah, je opravilo tri teste, dvakrat, v obdobju enega meseca, med dvema mednarodni- ma tekmama. Namen tega preizkusa je raziskati možnost dovolj zanesljivega predvidevanja tek- movalne zmogljivosti v kajaku na divjih vodah, dis- ciplini spust, na osnovi rezultatov testov na mirni vo- di . Izbrani napovedni kazalci so: hitrost veslanja in frekvenca srca, ki ju določa kriterij OBLA (Onset of Blood Lactate Accumu lation) (v08LA in HRosLA), povprečna hitrost veslanja, frekvenca srca in vseb- nost lakta ta v testu na 4000 m (v 4000, H R4000 in LA4000) in 500 m (v500, HR500 in LA500) . Kombinacija kazal- cev VosLA, v4000 in v500 je bila izbrana kot najboljši napovedni model. S pomočjo tega modela je izračunan teoretičn i čas 902 ± 20,5 s v prvem eksperimentu, kar je zelo podobno kot realno doseženi tekmovalni časi (903 ± 20,6 s). Podobno je bilo tudi v drugem eksperimentu, kjer je izračunan tekmovalni čas (862 ± 45 s) zelo podoben doseže- nemu (862.8 ± 48,4 s) . Ne glede na to natančnost pri predvidevanju specifičnih rezultatov, ni mogoče uporabiti katerega koli od obeh modelov za uni- verzalnega: izračunati tekmovalne rezultate v enem od obeh eksperimentov, z uporabo modela iz druge- ga eksperimenta. Zelo različne tekmovalne proge namreč povzročajo napake v izračunih . Ne glede na to, pa je razvrščanje po izračunanih in dejanskih časih ohranjeno, ne glede na uporabljeni model, kar pomeni, da se mogoče le ohranja princip, po katerem se različno kvalitetni tekmovalci razvrščajo med seboj, vsebuje pa ga izračunan model. Tak način uporabe izračunanih modelov je lahko pot, po kateri je mogoče spremljati zmogljivost posamez- nika skozi tekmovalno sezono. Ključne besede: kajak, divja voda, spust, prediktor- ski testi, regresijski modeli 4 AntonUša1 THE VALIDllY OF A PREDICTION MODELOF COMPETITION PERFORMANCE IN DOWNHILL WHITE WATER KAYAKING INTRODUCTION The competition result in each sport is the single and most i mportant characteristic of the sportsmen's per- formance at the particular moment. It refiects the in- fluence of the simultaneous effect of several biolog- ical, psychological, technical and tactical character- istics. These characteristics influence the competi- tion results by thei r complex combination. It is of great importance for the train ing process that the most important characteristics should be known. However, changes of values of those characteristics throughout the competition season asa result of trainingshould be assessed, and th is can be done by repeating specific tests. The value of each test should be affected by a si ngle factor. Therefore, a multifac- toral system could be achieved in order to asses the competition performance of a particular sportsman at a particular moment. The idea of the study was to ascertain if the construction of a very simple multi- factorial model can successfully predict competition performance in this sports event. The described idea can be realised by using a multiple regression mod- el (3). Two prel iminary condit ions should be met be- fore usingthis mathematical-statistical method. The firstcondition demands that the number of subjects should be large enough and the second one de- mands linear correlations between the tests and the competition results. These conditions have limited the appl ication ofthe described method, because in stud ies in top level sports, there exists only a small number of subjects. In such situations the linearity of correlations is also in question. Regardless of such cond itions, this method can be used for an assess- ment of how compet ition results depend on several basic characteristics of sportsmen determined by usingspecific tests. It is possible to calcu late theoret- ical competition results with a prediction model al- so in the period out of the competition period. This is very important for planning and selecting proper methods in the training process. The estimation of the valid ity of prediction should be done before this kind of model can be used for that purpose. Validity seems to be especially critical, because of its depen- dence ona number of subjects used in calculations in the period of construction of the model. This is a specific problem in top level sports, wherethe num- ber of subjects has been reduced to a team of less than 15. The compet ition res uit (tirne of rac ing) is very difficultto predict because of the different char- acteristics of water in the rivers, even in the same riv- er. Nevertheless we would like to estimate how much variance can be predicted in white water kayaking on the basis of resul ts of specific tests per- formed on flatwater. We have not found any data in 1 iterature related to th isto pic except the results of Fry and Morton (4) who useda similar idea, but on fiat water kayakers. We tested the hypothesis in spite of the non-standard competition conditions and small number of subjects. It is possible to predict an im- portant part of the variance of competition results in white water downhill kayaking by results of tests per- formed on fiat water in spite of the small number of subjects. METHODS Eleven top level kayakers from the national wh ite water downhill team (also international ly well ranked competitors) voluntarily participated in the study, after they confirmed in writing their consent to take part. The average height was 177±5 cm, body mass 70±6 kgand age 22±5 years. Each subject performed three tests on fiat water, one week after the World Cup competition. The whole procedure was repeated twice, over a span of one month, both times a week after the same range com- petitions. The OBLA TEST consisted of 5 repetitions of 1000 m on fiat water with a predetermined increase in paddling intensity. For this pu rpose, continuous monitoring of Heart Rate (HR) was used. It was pre- viously determined at five values: 11 O, 1 25, 140, 155 and 170 6/min. Resting periods between each 1000 m trials consisted of 8-1 O min of slow intensity return to the starting position. The average speed was calculated by using the tirne for each 1000 m . The endurance test over 4000 m (TEST 4000) con- sisted of one maximum effort trial on fiat water. The average speed was calcu lated by usingthe tirne over the 4000 m distance. The speed endurance test over a 500 m distance (TEST 500) consisted of one maximum effort trial on fiat water. The average speed was calculated by using the tirne over the entire d istance. For competition results (CR) the official results were used (tirne of race+ penalty tirne). Heart rate (HR) was measured continuously usinga PE3000 monitor (Polar Electro, Finland). Average HR in the OBLA TEST was calculated usingthe last2 min interval of each 1000 m distance. In the TEST 4000 the last 1 O min interval was used for calculating av- erage HR. The interval of the last 30 s was used for calculating the average H R in the TEST 500. Blood sam ples (20 µI) were taken from the hyper- emied earlobe and analysed usingan ANALOX GM7 Anton Ušaj THE VALIDITY OF A PREDICTION MODEL OF COMPETITION PERFORMANCE IN DOWNHILL WHITE WATER KAYAKING 5 (Analox, England) analyser. Blood samples were col- lected as soon as possible (usually 10-1 5 s after ar- rival through the finish line) after each 1000 m d is- tance. Similarly, blood samples were collected be- fore t he test start and as soon as possible after the TEST 4000. By contrast, blood sa mples were col- lected after the 3 m in resting period in the TEST 500. The criterion OBLA applied the method descri bed by S jod in et al. (5), additionally adapted for kayaking by using the two-compo nent model (8) . The pad- dlingspeed (voBLA) and heart rate (HRoBLA) were de- termined as potential pred ictors in the OBLA TEST. The average paddling speed (v4000), heart rate (HR4000) and lactate concentration (LA4ooo) were used as the potent ial p red ictors in the TEST 4000. Similarly average speed (vsoo), heart rate (H Rsoo) and blood lactate concentration (LAsoo) were selected as the predictors in the TEST 500. RESULTS There were 6 members of the national team partici- pating in the fi rst experiment (EXPERIM ENT 1 ), w hich consisted of the competit io n and tests. N ine kayakers participated in the second experiment (EX- PERIMENT 2) which consisted of another competi- tion and tests. The inte rval between the two experi- ments was one month. Ave rage padd ling speed, voBLA is similar to v 4000, but significant ly lower (P< O.OS) t han v500 in EXPE- RIMENT 1 (Table 1 ). By contrast, HRoBLA is signifi- cantly lower than H R4000 (P< O.OS). Surprisingly HR500 is similar to H R4000, i rrespective of different paddl ing speeds (Table 1 ). The situation in EXPERIMENT 2 is sim ilar to that in EXPERIMENT 1 (Table 1 ). Paddling speed v00LA was practically identica l to v 4000. Both were clea rly lower than v500 (P< O.OS) . Heart rate HRoBLA was lower Table 1 Basic statistical data of both experiments. CHARACTERISTICS UNITS EXPERIMENT 1 EXPERIMENT 2 N=6 N=9 VnRIA m/s 3.29±0.09 3.04±0.26 HRnRIA b/min 153+7 151 + 7 V snn m/s 3.75±0.1 7 3.61 ±0.23 HRsN'I b/min 178±5 177±4 LAsnn mmol/1 7.6±1.3 6.6±1.1 V•nnn m/s 3.30±0.18 3.04±0.26 HR.nnn b/min 178+6 179+9 LA•nnn mmol/1 8.8+4.7 5.9+1 .1 CR s 903±21 863±48 LEGEND: Values are means ( standard deviations) Table 2 Correlat ions between characteristics and competi- t ion results in both experiments. CHARACTERISTICS EXPERIMENT 1 CR Vrrn, A -0.754? HRoBLA 0.252 Vrnn -0.908* HRrnn -0.116 LAsnn -0.856* V4nN'I - 0.776? HR4nnn 0.614 LA4000 - 0.501 LEGEN D: * - P< O.OS ** - P < 0 .01 ? - P ( O.OS (0.06 - 0 .08) EXPERIMENT 2 CR -0.920** - 0.302 -0.879** 0.693 0.429 -0.912** -0.503 0.190 than H R4000 (P< O.OS) but was similar to H R500. A lso [LA] in both tests, LA500 and LA4000 did not show any significant di fference. The one-month period of t raining between EXPE- RIM ENT 1 and EXPERIMENT 2 affected the results in t he tests applied (Table 1 ). Paddling speed v00LA de- creased by abo ut 0.25 m/s (P<0.05), and v500 showed an insign ificant tendency to decrease. Sim ilarly v4000 also showed an insigni ficant tendency to decrease. None of the heart rate characteristics used showed any significant change from the first to the second experiment. The average t irne necessary to finish the race in EX- PERIMENT 1 was signi ficantly longer (P< O.OS) than in the second competition (EXPERIMENT 2) (Table 1) because of the different rivers and competition dis- tance. Correlations between selected characteristics and competition results (CR) in both experi ments showed that on ly the speed v500 correlated sign ificantly (P= 0.01) (Table 2). Both of the other selected speeds VosLA and v4000 corre lated to CR significant ly in EX- PERIM ENT 2 and were a little above the limit of sig- ni ficance at EXPERIMENT 1 (P=0.08 for VoBLA and P=0.07 for v4000) (Table 2) . According to the correlations between selected char- acteristics and CR we selected four combinations of predictors (Table 3). Linear mult iple regression was used for selecting the best prediction combination of competit ion resu lt in w hite water downhill kayaki ng (Table 3). The most powerful prediction combination in both experiments was selected accord ing to mul- t iple correlation values, adj usted for a small number of subjects in both experiments and accord ing to their level of signi fica nce. That combinat ion was rep- 6 Anton Ušaj THE VALIDITY OF A PREDICTION MODEL OF COMPETITION PERFORMANCE IN DOWNHILL WHITE WATER KA YAKING Table 3 Pred iction quality assessments using the mult iple correlation values of different combination o f char- acteristics as predictors and CR as criterions in both experiments. EXPERIMENT 1 EXPERIMENT 2 N=6 N=9 Combination of Rmultip. p Rmultip. p predictors adjust. adjust. vOBLA + v50Q + V40QO 0.965 0.02 0.784 0.01 voBLA + HRoBLA 0.385 0.22 0.870 <0.01 v 500 + H R500 + LAsoo 0.583 0.239 0.951 0.029 v 4000 + H R4000 + LA40()( 0.947 0.031 0.917 o.oso resented in all three paddling speeds used in the study (Table 3). To ascertain if significant d ifferences existed in pre- diction modelsofboth CR in both experiments, both regression equations were compared : CR, = 1554-97.5 * v08LA - 86.6 * v 500 - 1.6 * V4000 (Equation 1) CR2 = 1444 -112.2 * v08LA - 44.3 * v500 - 26.5 * v4000 (Equation 2) The calculated competition results in EXPERIMENT 1 (CR1) were 902±20.5 s. This is a similar va lue to the real competition results (903 ± 20.6 s). Both were in close correlation (r = 0.99, P 1 O min in classical downhill competi- tions) and very similar technique and tactics among top performance competitors may influence the en- durance and aerobic and anaerobic energetic pro- cesses that have become essential for competition success (1,6). This may be very similar to the situa- Anton Ušaj THE VALIDITY OF A PREDICTION MODEL OF COMPETITION PERFORMANCE IN DOWNHILL WHITE WATER KAYAKING 7 tion in fiat water kayaking of sim ilar duration (2,4,7). The relatively high correlations between test results on fiat water - vO8LA, v 4000, and v500 -demonstrated in our study, proved this hypothesis. A caution should be added in generalising over whether this conclu- sion should be applied to technically very difficult and short competitions, which have also become very popular lately. According to tirne of competi- tion (903 ±21 and 863 ±48 s), white water compe- titions were simi lar to the 4000 m distances on fiat water. The relationship between endurance perfor- mance assessed byv4000 and by vO8LA, and speed en- durance assessed by v500, w ith competition resu lts (CR) were similar to the Venti latory Threshold, Vo2 max and 1 min all-out test correlation w ith 1000 m and 10000 m competit ions on fiat water (4). Prediction of about 90% of the known variance of competition results in both experiments was similar to the prediction of fiat water competition results over 1000 m (92%) and over 10000 m (90%). The calculated prediction times showed satisfactory pre- diction only with specific models. The tirne differ- ences of differentdownhill races were too large (140 s) and unpredictable. When a specific model was se- lected, then it predicted accurately only those com- petition results which were used for calculation of that model, but not if competition results of another experiment were used. Therefore the pred iction of competition results cannot be sufficiently precise and accurate w ith a universal and simple model. lrrespective of these differences, the very good cor- relations between calculated and observed compe- tition results have shown that models obtained in two d ifferent samples of subjects and in different parts of the competition season preserves the basic princi ples whereby kayakers were different in t heir performance. In this case, one of either models can be used for calcu lation of the theoretical competi- tion results, based on tests results performed throughout the competit ion season. The purpose of such calculations can beto assess possible qualita- tive changes influenced by training. However, both models should be additionally approved before their use in practice. The present situation shows that the TEST 500 is not necessary in the group of tests. But this is not conven ient, for two reasons. The first one is that the new trends in white water downhill show that competitions should be also shorter ( <5 min), and with a parallel sta rt. This means that speed and speed endurance (anaerobic lactate power and ca- pacity) may dominate in those competitions, which was similar in thewhite water slalom (1,6). Therefore it is useful that one test assesses that performance. The second reason is related to the requirements of the training process. It should be necessarythat each kayaker develops his overa ll performance with in- creased endurance, speed, speed endurance, strength and so on. Therefore the TEST 500 should be necessary as an indication of the effect of training on that particular performance. REFERENCES 1. Astrand, PO: Textbook ofWork Physiology. McGraw-Hill Book com- pany, 1986. 2. Colli, R, Faccini P., Schermi C, lntraini E, Dal Mante A.: Valutazione funzionale ed allenamento del canoista. Scuola dello Spori 1990; 18: 27-37. 3. Draper N, Smith H.: Appl ied Regression Analysis, Second Edition. John Willey & Sons, 1981. 4. Fry R.W., Morton A.L : Physiological and kinanthropometricattributes of elite fiat water kayakists. Med Sci in sports and Exer 1991; 23: 1297- 1301. 5. Sjodin B, Schele R, Karlsson J, Linarsson D, and Wallensten R: The Physiological Background of Onset of Blood Lactate Accumulation (OBLA). In: Exercise and Spori Biology 1982; Vol. 12, 43-56. 6. Sydney K, Shephard R. Y: Physiological Characteristics and Performance of the White-Water Paddler. Europ J Appl Phisiol 1973; 32: 55-70. 7. Tesch P, Piehl K, Wilson G, Karlsson J: Physiological investigation of Swedish el ite canoe competitors. M ed. and Sci. in Sports and Exer. 1976; 8: 214-218. 8. Ušaj A, Starc V: Two concepts of Anaerobic Threshold and Running Endurance. In: Sports, Medicine and Health (GPH Hermans ed.) Elseiver Science Publishers B.V. 1990; 753-758.