Relationships between general physical performance and judo-specific power outputs in elite judo athletes: an allometric approach

Abstract

Background

This cross-sectional study examined whether associations between general physical performance and peak power in hikite, tsurite, and morote-seoi-nage differ when variables are expressed in absolute terms or scaled to fat-free mass (FFM).

Methods

Sixty Chinese national-team judoka completed body-composition, judo-specific peak-power, vertical-jump, isometric mid-thigh pull, one-repetition maximum, and Wingate assessments. Spearman correlations were evaluated using a global Benjamini-Hochberg false-discovery-rate correction across 93 primary tests. Sex-controlled explanatory regression models were fitted, and an exponent of 0.75 was examined as a sensitivity analysis to the primary FFM^0.67 scaling.

Results

Forty-four of 54 absolute associations remained significant after false-discovery-rate correction. After FFM^0.67 scaling, significant associations remained for eight hikite and seven tsurite comparisons, whereas only scaled countermovement-jump peak force was associated with scaled morote-seoi-nage peak power (ρ = 0.369, q = 0.006). The best-fitting absolute multivariable models explained 55.8%–79.9% of adjusted sample variance. Using 0.75 did not change any false-discovery-rate significance classification.

Conclusions

Associations between general physical tests and judo-specific peak power were action-dependent and changed after scaling. Absolute values characterize total output, whereas scaled values provide complementary information relative to FFM. Direct task-specific testing remains necessary, particularly for the multi-phase morote-seoi-nage action.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13102-026-02003-1.

Background

Judo performance requires repeated high-intensity actions during grip dispute, unbalancing, throwing execution, transition from standing combat to groundwork combat, and during groundwork combat. These actions depend on rapid force production, maximal strength, short-duration power output, and repeated sprint ability, which are recognized components of the physical profile of high-level judo athletes [1–3]. In high-performance settings, general physical performance tests are used not only to describe strength and power capacities, but also to monitor training readiness, neuromuscular fatigue, recovery status, and adaptation to training loads. Tests such as one-repetition maximum strength assessments, vertical jumps, isometric force tests, and anaerobic power tests are routinely used and have well-established procedures in athlete monitoring [4–8]. Determining whether these general tests are related to judo-specific power may therefore improve the interpretation of general testing data alongside more task-specific assessments. However, this relationship should be examined in relation to the specific judo action being assessed rather than treated as a single general construct.

Previous studies have established that judo athletes differ in morphological, strength, power, and anaerobic characteristics according to competitive level, sex, and weight category [1, 2, 9]. Other work has developed or validated judo-specific testing approaches, including cable-resistance or ergometer-based assessments using judogi- or kimono-like grips, and has shown that such tests can capture task-relevant neuromuscular information in judo athletes [10–12]. In parallel, biomechanical and skill-comparison studies of seoi-nage techniques have described the coordinated lower-limb, trunk, and upper-limb contributions required during complete throwing actions, as well as phase-specific differences between skill levels [13–15]. However, these lines of evidence have rarely been integrated to determine whether routinely used general physical tests relate similarly to different judo-specific power outputs, or whether these relationships are preserved when performance is interpreted relative to fat-free mass.

The associations between general physical qualities and sport-specific outcomes may differ according to the demands of the target action. Movement pattern, force direction, contraction characteristics, velocity, and inter-segmental coordination are relevant when interpreting these associations [16, 17]. In the present study, hikite (sleeve-side pulling action), tsurite (lapel-side lifting and pulling action), and morote-seoi-nage (a te-waza, i.e., hand/arm technique in the traditional judo classification) were selected to avoid treating judo-specific power as a homogeneous outcome. Hikite and tsurite represent distinct kuzushi-related components, whereas morote-seoi-nage integrates pulling with entry, body rotation, lowering of the center of mass, and forward-downward completion. These actions therefore impose different task constraints for examining associations with general physical performance.

A second interpretive issue concerns the weight-category structure of judo. Absolute strength and power are practically meaningful because athletes must generate mechanical output against an opponent. However, body mass, fat-free mass, body composition, and strength-related characteristics differ across sex, competitive level, and weight category in judo athletes [1, 2, 9]. As a result, absolute associations between general physical tests and judo-specific power may reflect both neuromuscular capacity and body-size-related differences. Allometric scaling has been recommended when comparing strength and power among individuals with different body dimensions, and previous work in judo has shown that simple ratio standards may not adequately account for body-size effects in strength assessment [18–21]. Thus, absolute and allometrically scaled analyses provide complementary information: absolute values describe total mechanical output, whereas scaled values help examine whether relationships remain after accounting for fat-free mass.

Therefore, the primary aim was to compare associations between general physical performance and judo-specific peak power across hikite, tsurite, and morote-seoi-nage when variables were expressed in absolute and FFM-scaled terms. Secondary aims were to describe between-sex differences and quantify sex-controlled sample variance using explanatory regression models. The morote-seoi-nage tokui-waza analysis was designated exploratory and is reported in the supplementary material. We hypothesized that absolute morphological and physical variables would be positively associated with absolute judo-specific peak power, that scaling would attenuate these associations, and that the retained scaled associations would differ by action.

Methods

Participants

Sixty elite judo athletes from the Chinese National Team participated in this cross-sectional study. The sample comprised 31 males and 29 females and included athletes from all male and female weight categories defined by the International Judo Federation. Inclusion criteria were: active national-team roster status, elite domestic or international competitive experience, and absence of performance-limiting musculoskeletal injury during the three months preceding data collection. The distribution of athletes across weight categories is reported in Supplementary Table S1.

All athletes were informed of the study procedures, potential risks, and benefits before participation and provided written informed consent. The study protocol was approved by the Ethics Committee of Shanghai University of Sport (approval no. 102772025RT256) and was conducted in accordance with the Declaration of Helsinki.

Study design

Figure 1. Study design and testing workflow. Day 1 comprised morning fasting body-composition assessment, the standardized warm-up, and judo-specific peak-power testing in the fixed order hikite, tsurite, and morote-seoi-nage. Day 2 occurred 48 h later and comprised the same warm-up followed by SJ, CMJ, DJ, IMTP, squat 1RM, prone-row 1RM, and one Wingate test. The warm-up consisted of a 10-min low-intensity jog followed by Knee Hug to Calf Raise (10 m), World’s Greatest Stretch with Rotation (5 per side), Dynamic Standing Quadriceps Stretch (10 m), Inchworm Walkout (5 repetitions), Lateral Lunge Mobility (8 per side), and Dynamic Good Morning (10 repetitions). Recovery was 2 min between repeated trials and 5–10 min between testing modalities.

Fig. 1.

shows the study design and testing workflow

Data were collected from 23 to 26 December 2025 using two testing sessions separated by 48 h; age was calculated using 23 December 2025 as the reference date. Athletes completed two familiarization sessions one week before formal testing. They were instructed to avoid high-intensity training for 24 h and caffeine and alcohol for 12 h before each session. All tests were conducted at the same time of day by the same research team using standardized instructions and verbal encouragement.

On Day 1, body composition was assessed before the warm-up. On both days, the same warm-up comprised a 10-min low-intensity jog followed by Knee Hug to Calf Raise (10 m), World’s Greatest Stretch with Rotation (5 per side), Dynamic Standing Quadriceps Stretch (10 m), Inchworm Walkout (5 repetitions), Lateral Lunge Mobility (8 per side), and Dynamic Good Morning (10 repetitions). On Day 2, testing proceeded in the following order: SJ, CMJ, DJ, IMTP, squat 1RM, prone-row 1RM, and Wingate. SJ, CMJ, DJ, IMTP, hikite, tsurite, and morote-seoi-nage each comprised two valid trials separated by 2 min; the higher metric-specific value was retained. Recovery between testing modalities was 5–10 min. The 1RM and Wingate procedures are described separately below.

Body composition assessment

Body composition was assessed in the morning before performance testing using a multi-frequency bioelectrical impedance analyzer (InBody 770; InBody Co., Ltd., Seoul, South Korea). Athletes were instructed to arrive after an overnight fast and to avoid exercise before assessment. No additional standardized water-intake protocol or pre-assessment voiding requirement was recorded. Body mass, body fat percentage, and fat-free mass (FFM) were recorded, and FFM was used as the scaling denominator. The measurement limitations of this field method are addressed below [22, 23].

Judo-specific power tests

Judo-specific power was assessed using a Keiser A300 pneumatic resistance system (Keiser Corporation, Fresno, CA, USA). A standard judogi grip was attached to the cable terminal to allow athletes to perform the actions using sport-specific gripping patterns. Although previous device-based judo studies did not use the same pneumatic system, they have used cable-resistance systems, judogi- or kimono-like grips, and sleeve- and lapel-side pulling actions to assess judo-specific force or power production [10, 11]. The present setup adopted a similar task-constraint rationale using a pneumatic cable system and a judogi grip attachment. The judo-specific test setup and action endpoints are illustrated in Fig. 2.

Fig. 2.

Representative action-demonstration sequences of the judo-specific power test actions. Panels A-C show hikite, D-F show tsurite, and G-J show morote-seoi-nage performed from a static start using a judogi grip attached to a pneumatic cable-resistance system. Morote-seoi-nage illustrates entry, body rotation, lowering of the center of mass, and forward-downward completion against resistance; the endpoint did not include opponent projection or landing. These photographs are intended to illustrate the task setup and action endpoints only and do not depict the formal experimental data-collection sessions. The demonstrator was a retired high-level judo athlete who was not part of the study sample

Athletes performed three judo-specific actions on their dominant throwing side from their preferred combat stance. The 20% body-mass load was selected in a pilot assessment of morote-seoi-nage involving 12 athletes (six males and six females); within each sex, two athletes represented light, middle, and heavy weight-category groupings. Loads of 10%–35% body mass were tested in ascending order, with two trials per load and 2 min of rest between loads. A national-team judo coach judged whether the prescribed technical sequence was maintained. 20% of body mass was the highest load at which all 12 athletes completed a valid morote-seoi-nage action; at 25% and above, at least one athlete could not maintain the prescribed technique. The pilot evaluated morote-seoi-nage only. The same relative load was subsequently applied to hikite and tsurite to standardize testing across actions and body sizes; it was not assumed to be the optimal power load for either isolated pulling action.

Formal testing followed the fixed order hikite, tsurite, and morote-seoi-nage. For hikite, athletes performed a maximal diagonal sleeve-side pull from approximately shoulder height. For tsurite, athletes performed a maximal lapel-side lifting and pulling action from slightly below shoulder height. For morote-seoi-nage, athletes performed entry, body rotation, lowering of the center of mass, an upward pulling component, and forward-downward completion against the pneumatic resistance. All actions began from a motionless, pre-tensioned cable position. Each action comprised two valid trials separated by 2 min. According to the manufacturer, the A300 Power Display expresses real-time power as force × velocity in watts and displays peak power for each repetition; peak power is the highest power attained during the set [24]. Each formal trial comprised one maximal action, and a researcher transcribed the displayed peak-power value by hand and subsequently entered it into Excel; no software export file was retained. The investigators did not apply a separate equation, averaging window, or post-processing algorithm. The higher peak-power value from the two valid trials was used in analysis.

Reliability was calculated from the two same-session trials completed by the same 60 athletes, with 2 min between trials and the fixed action order described above. A two-way mixed-effects, single-measure consistency coefficient [ICC(3,1)] was used. Reliability was acceptable to excellent: hikite ICC = 0.924 (95% CI 0.876–0.954), CV = 9.02%, SEM = 64.82 W; tsurite ICC = 0.931 (95% CI 0.887–0.958), CV = 8.10%, SEM = 50.28 W; and morote-seoi-nage ICC = 0.900 (95% CI 0.838–0.939), CV = 9.89%, SEM = 40.77 W. Full results are reported in Supplementary Table S8.

General physical performance tests

General physical performance was assessed using vertical jumps, the isometric mid-thigh pull (IMTP), one-repetition maximum (1RM) strength tests, and the Wingate anaerobic test. These tests were selected to assess lower-limb force production, maximal isometric strength, maximal dynamic strength, and anaerobic power and capacity.

Vertical-jump tests included the squat jump (SJ), countermovement jump (CMJ), and drop jump (DJ), all performed on dual force platforms (ForceDecks FD Lite; VALD Performance, Brisbane, Australia) sampling at 1000 Hz. Athletes kept their hands on their hips. The SJ was performed from approximately 90° knee flexion after a 2-s static pause, the CMJ used a rapid self-selected countermovement, and the DJ was performed from 40 cm for males and 30 cm for females with instructions to minimize ground contact time and maximize jump height. Two valid trials were completed for each jump, with 2 min between trials; the higher metric-specific value was retained [4].

The IMTP was performed using the same force-platform system and a rigid testing frame. Knee and hip angles were approximately 130°–140° and 140°–150°, respectively. Athletes pulled vertically against the immovable bar for 3–5 s and were instructed to pull ‘as fast and as hard as possible.’ Two valid trials were separated by 2 min, and the higher value for each IMTP metric was retained [6].

Maximal dynamic strength was assessed using 1RM squat and prone-row tests following standardized procedures [25]. Athletes first completed 5–10 repetitions with a light load and rested 1 min. The load was then increased by 10%–20% for the squat or 5%–10% for the prone row for 3–5 repetitions. After 2 min, the load was increased by the same relative amount for 2–3 repetitions. Single 1RM attempts then followed 2–4 min of rest. After a successful attempt, the load increased by 10%–20% for the squat or 5%–10% for the prone row; after a failed attempt, it decreased by 5%–10% or 2.5%–5%, respectively. Typically, 3–5 maximal attempts were completed, and the highest valid load was retained. A valid squat required the superior surface of the thigh to reach at least parallel with the floor; for the prone row, the barbell had to contact the underside of the elevated bench.

Anaerobic power and capacity were assessed once using a 30-s Wingate test on a Wattbike Trainer cycle ergometer (Wattbike Ltd., Nottingham, UK). No additional test-specific warm-up was performed immediately before the Wingate test. Resistance was set by the manufacturer’s software according to athlete height, body mass, and sex. Athletes were instructed to maintain maximal effort throughout, and genuine Wingate peak and mean power outputs were retained [7].

Data processing and calculated variables

Force-time data from jump and IMTP tests were processed using the manufacturer’s software. SJ and CMJ peak force and jump height were extracted. DJ reactive strength index was calculated as flight time divided by ground contact time and is dimensionless. The eccentric utilization ratio was calculated as CMJ height divided by SJ height. CMJ stiffness was obtained from the force-platform output and expressed relative to body mass.

For the IMTP, peak vertical force was the maximum force value, and impulse was calculated over the 0–50 ms and 0–200 ms epochs from force onset. For the judo-specific tests, only the manually transcribed peak-power values described above were analyzed. Wingate peak and mean power were obtained from the Wattbike software.

Variables were analyzed in absolute and allometrically scaled forms. Allometric scaling was applied to strength, force, and power variables using FFM as the scaling denominator:

where Y represents the absolute performance value. The theoretical exponent of 0.67 was chosen according to geometric-similarity principles and previous normalization recommendations, rather than estimating variable-specific exponents from this sample, to limit sample-specific overfitting [18–20]. Jump heights remained in their original units, whereas CMJ stiffness was normalized to body mass. An exponent of 0.75 was examined in a restricted sensitivity analysis of the same 39 principal scaled correlations and the aligned explanatory models.

Statistical analysis

Final analyses were reproduced from the validated master dataset using Python 3.12 (SciPy 1.13.1 and statsmodels 0.14.2). Continuous variables are presented as mean ± standard deviation. Participant-characteristic sex differences were examined using independent-samples t tests with Cohen’s d and 95% confidence intervals. Extended physical and judo-specific variables were compared using Mann-Whitney U tests because several performance variables deviated from normality; rank-biserial correlation was reported as the effect size. Statistical significance was set at α = 0.05.

Normality was assessed using the Shapiro-Wilk test. Spearman rank-order correlations were calculated for absolute variables and absolute judo-specific peak power, for scaled or relative variables and FFM-scaled peak power, and among the three peak-power outcomes. The Benjamini-Hochberg procedure was applied across the complete set of 93 primary physical-performance correlations, and q < 0.05 defined significance. For the 0.75 sensitivity analysis, the same family structure was retained: the 54 absolute-test p values were combined with the 39 corresponding 0.75-scaled p values and corrected together as a 93-test family. Raw p values and FDR-adjusted q values are reported in the supplementary material. The six correlations among action outcomes were treated as a separate secondary family.

For concise main-text presentation, Table 3 was constructed using a domain-based reporting rule. The displayed variables were selected to represent the measurement domains included in the testing battery: morphology, maximal dynamic strength, maximal isometric force, jump-derived force and mechanics, and anaerobic power. Selection for Table 3 was independent of the observed correlation coefficient, raw p value, and FDR-adjusted q value. No correlation-strength threshold was used because threshold-based presentation would make inclusion conditional on the observed results. Complete results for all 93 primary correlations are reported in Supplementary Tables S3A and S3B.

Table 3.

Representative Spearman correlations between physical variables and judo-specific peak power

Variable	Hikite PP, ρ (q)	Tsurite PP, ρ (q)	MSN PP, ρ (q)
Absolute variables
Body mass	0.708 (< 0.001)	0.751 (< 0.001)	0.715 (< 0.001)
Fat-free mass	0.805 (< 0.001)	0.819 (< 0.001)	0.695 (< 0.001)
Squat 1RM	0.851 (< 0.001)	0.806 (< 0.001)	0.526 (< 0.001)
Prone row 1RM	0.848 (< 0.001)	0.791 (< 0.001)	0.564 (< 0.001)
IMTP peak force	0.781 (< 0.001)	0.681 (< 0.001)	0.546 (< 0.001)
SJ peak force	0.791 (< 0.001)	0.831 (< 0.001)	0.661 (< 0.001)
CMJ peak force	0.784 (< 0.001)	0.807 (< 0.001)	0.691 (< 0.001)
Wingate peak power	0.749 (< 0.001)	0.740 (< 0.001)	0.540 (< 0.001)
Wingate mean power	0.739 (< 0.001)	0.774 (< 0.001)	0.600 (< 0.001)
CMJ stiffness	0.502 (< 0.001)	0.624 (< 0.001)	0.636 (< 0.001)
FFM^0.67-scaled/relative variables
Scaled squat 1RM	0.676 (< 0.001)	0.544 (< 0.001)	0.026 (0.893)
Scaled prone row 1RM	0.604 (< 0.001)	0.496 (< 0.001)	0.003 (0.979)
Scaled IMTP peak force	0.433 (< 0.001)	0.221 (0.134)	0.030 (0.879)
Scaled SJ peak force	0.474 (< 0.001)	0.519 (< 0.001)	0.194 (0.194)
Scaled CMJ peak force	0.474 (< 0.001)	0.513 (< 0.001)	0.369 (0.006)
Scaled Wingate peak power	0.459 (< 0.001)	0.480 (< 0.001)	0.167 (0.267)
SJ height	0.595 (< 0.001)	0.410 (0.002)	-0.214 (0.147)
CMJ height	0.569 (< 0.001)	0.384 (0.004)	-0.204 (0.168)
Relative CMJ stiffness	-0.118 (0.440)	0.108 (0.473)	0.151 (0.314)

Values are Spearman’s ρ, with global Benjamini-Hochberg FDR-adjusted q values in parentheses. Variables were selected for concise main-text presentation using a domain-based reporting rule covering morphology, maximal dynamic strength, maximal isometric force, jump-derived force and mechanics, and anaerobic power. Selection was independent of ρ, raw p, and FDR-adjusted q values. No correlation-strength threshold was applied because threshold-based inclusion would make the main-text display conditional on the observed results. Complete results for all 93 primary correlations are reported in Supplementary Tables S3A and S3B

PP peak power, MSN morote-seoi-nage, FFM fat-free mass

Same-session reliability for the judo-specific tests was assessed using ICC(3,1), representing a two-way mixed-effects, single-measure consistency coefficient. For each athlete, the individual CV was calculated as the absolute difference between the two trials divided by √2 and by that athlete’s two-trial mean, multiplied by 100; the reported CV was the mean of the 60 individual CVs. SEM was calculated as the square root of the residual mean square from the two-way mixed-effects ANOVA used for the ICC. Full results are reported in Supplementary Table S8.

Sex-controlled single-predictor models were retained in the supplementary material to quantify each physical variable’s incremental sample variance while limiting simultaneous collinearity. To provide one multivariable model per outcome, all two-predictor combinations among ten theoretically relevant strength, force, stiffness, impulse, and anaerobic-power variables were screened with sex included in every model. To limit overfitting at n = 60, no model included more than two physical predictors. Combinations with VIF ≥ 5 were excluded, and the eligible model with the highest adjusted R² was retained for each action. This selection was exploratory and does not constitute external prediction.

Regression coefficients were estimated by ordinary least squares and reported with HC3 heteroscedasticity-consistent standard errors for all final models; standardized β, 95% confidence intervals, adjusted R², ΔR², and VIF are reported. Scaled models used the corresponding FFM^0.67 predictors, and aligned models were repeated using 0.75. Morote-seoi-nage tokui-waza status was self-reported and examined in sex-adjusted explanatory models for the three absolute and three scaled peak-power outcomes, with FDR correction across the six coefficients; this analysis was designated exploratory and hypothesis-generating.

Results

Participant characteristics

Participant characteristics and formal comparisons are presented in Table 1. Males were taller and had greater body mass and FFM than females (all p ≤ 0.006), whereas females had a higher body-fat percentage (p = 0.001). Age and training experience did not differ between sexes (both p ≥ 0.729).

Table 1.

Participant characteristics by sex

Variable	Males (n = 31)	Females (n = 29)	t	p	Cohen’s d [95% CI]
Age (y)	23.6 ± 3.5	23.7 ± 3.3	-0.15	0.880	-0.04 [-0.55, 0.47]
Height (cm)	180.1 ± 9.3	168.5 ± 9.0	4.93	< 0.001	1.27 [0.72, 1.83]
Training experience (y)	9.7 ± 3.5	10.0 ± 2.9	-0.35	0.729	-0.09 [-0.60, 0.42]
Body mass (kg)	91.0 ± 23.8	73.3 ± 24.4	2.85	0.006	0.74 [0.21, 1.26]
Fat-free mass (kg)	73.4 ± 13.8	54.5 ± 11.6	5.73	< 0.001	1.48 [0.91, 2.06]
Body fat (%)	18.0 ± 6.3	23.6 ± 6.7	-3.35	0.001	-0.87 [-1.40, -0.33]

Values are mean ± SD. Cohen’s d is signed as male minus female

CI confidence interval, FFM fat-free mass

Descriptive performance data

The full set of valid performance variables and between-sex comparisons is presented in Table 2. Males had higher absolute and FFM-scaled strength, force, Wingate, hikite, and tsurite outputs for most comparisons. Scaled hikite and tsurite peak power remained higher in males (both p < 0.001), whereas scaled morote-seoi-nage peak power did not differ between sexes (p = 0.906).

Table 2.

Physical and judo-specific performance variables and between-sex comparisons

Variable	Males (n = 31)	Females (n = 29)	U	Z	p	Rank-biserial r
Absolute strength and power variables
Squat 1RM (kg)	165 ± 35	107 ± 24	841.0	5.806	< 0.001	0.871
Prone row 1RM (kg)	110 ± 15	77 ± 16	836.5	5.749	< 0.001	0.861
IMTP peak force (N)	3241 ± 597	2381 ± 465	784.0	4.941	< 0.001	0.744
IMTP impulse 0–50 ms (N·s)	54.6 ± 18.7	45.6 ± 18.0	613.5	2.419	0.016	0.365
IMTP impulse 0–200 ms (N·s)	294.2 ± 89.3	232.6 ± 76.8	657.0	3.062	0.002	0.462
SJ peak force (N)	4434 ± 863	3189 ± 734	785.0	4.956	< 0.001	0.746
CMJ peak force (N)	4362 ± 774	3190 ± 1048	782.0	4.911	< 0.001	0.740
Wingate peak power (W)	1167 ± 174	786 ± 223	817.5	5.436	< 0.001	0.819
Wingate mean power (W)	704 ± 104	471 ± 140	809.0	5.311	< 0.001	0.800
Hikite peak power (W)	897 ± 153	495 ± 124	879.0	6.346	< 0.001	0.956
Tsurite peak power (W)	700 ± 159	407 ± 104	839.0	5.754	< 0.001	0.867
Morote-seoi-nage peak power (W)	449 ± 115	377 ± 146	619.5	2.508	0.012	0.378
Allometrically scaled variables (FFM^0.67)
Scaled squat 1RM (kg·kg FFM⁻⁰·⁶⁷)	9.23 ± 1.03	7.33 ± 0.98	820.0	5.473	< 0.001	0.824
Scaled prone row 1RM (kg·kg FFM⁻⁰·⁶⁷)	6.23 ± 0.48	5.28 ± 0.56	809.0	5.311	< 0.001	0.800
Scaled IMTP peak force (N·kg FFM⁻⁰·⁶⁷)	182.59 ± 23.78	163.84 ± 18.22	676.0	3.343	< 0.001	0.504
Scaled IMTP impulse 0–50 ms (N·s·kg FFM⁻⁰·⁶⁷)	3.03 ± 0.74	3.07 ± 0.75	430.0	-0.281	0.779	-0.043
Scaled IMTP impulse 0–200 ms (N·s·kg FFM⁻⁰·⁶⁷)	16.48 ± 3.89	15.82 ± 3.52	498.0	0.710	0.478	0.108
Scaled SJ peak force (N·kg FFM⁻⁰·⁶⁷)	248.65 ± 26.62	218.05 ± 25.05	710.0	3.846	< 0.001	0.580
Scaled CMJ peak force (N·kg FFM⁻⁰·⁶⁷)	245.05 ± 21.41	216.07 ± 38.58	733.0	4.186	< 0.001	0.631
Scaled Wingate peak power (W·kg FFM⁻⁰·⁶⁷)	66.13 ± 8.09	53.55 ± 10.05	744.0	4.349	< 0.001	0.655
Scaled Wingate mean power (W·kg FFM⁻⁰·⁶⁷)	39.75 ± 4.12	32.01 ± 5.67	781.0	4.896	< 0.001	0.737
Scaled hikite peak power (W·kg FFM⁻⁰·⁶⁷)	50.79 ± 7.47	33.90 ± 6.19	861.0	6.080	< 0.001	0.915
Scaled tsurite peak power (W·kg FFM⁻⁰·⁶⁷)	39.57 ± 8.21	27.80 ± 4.08	806.0	5.266	< 0.001	0.793
Scaled morote-seoi-nage peak power (W·kg FFM⁻⁰·⁶⁷)	25.37 ± 5.79	25.35 ± 6.53	458.0	0.118	0.906	0.019
Relative jump-derived variables
SJ height (cm)	34.4 ± 5.6	27.4 ± 3.8	787.5	4.993	< 0.001	0.752
CMJ height (cm)	35.2 ± 5.3	28.5 ± 3.6	771.5	4.757	< 0.001	0.716
Relative CMJ stiffness (N/m/kg)	83.5 ± 26.5	84.2 ± 16.8	380.0	-1.021	0.307	-0.155
DJ RSI (dimensionless)	1.52 ± 0.56	1.44 ± 0.48	490.5	0.599	0.549	0.091
Eccentric utilization ratio (dimensionless)	1.03 ± 0.09	1.04 ± 0.09	396.5	-0.777	0.437	-0.118

Values are mean ± SD. Between-sex comparisons used Mann-Whitney U tests. Rank-biserial r is the effect size; positive values indicate higher values in males. FFM-scaled variables were calculated as Y/FFM^0.67. DJ RSI is dimensionless

1RM one-repetition maximum, IMTP isometric mid-thigh pull, SJ squat jump, CMJ countermovement jump, MSN morote-seoi-nage, FFM fat-free mass

Associations between physical variables and judo-specific peak power

After global FDR correction, 44 of 54 absolute correlations remained significant: 15 each for hikite and tsurite and 14 for morote-seoi-nage. Absolute body size, maximal strength, jump-derived force, and Wingate outputs were broadly associated with all three judo-specific peak-power outcomes (Table 3 and Supplementary Table S3A).

The scaled pattern was more selective. Sixteen of 39 scaled correlations remained significant after global FDR correction: eight for hikite, seven for tsurite, and one for morote-seoi-nage. Scaled CMJ peak force was the only retained correlate of scaled morote-seoi-nage peak power (ρ = 0.369, p = 0.004, q = 0.006). Repeating the 39 scaled correlations with an exponent of 0.75 while retaining the same 93-test FDR family produced no change in significance classifications; 16 of 39 associations remained significant and the largest absolute change in ρ was 0.099 (Supplementary Table S6A).

Relationships among hikite, tsurite, and morote-seoi-nage power outputs

All three absolute action-to-action correlations were significant after correction: hikite versus tsurite (ρ = 0.746, p < 0.001, q < 0.001), hikite versus morote-seoi-nage (ρ = 0.556, p < 0.001, q < 0.001), and tsurite versus morote-seoi-nage (ρ = 0.585, p < 0.001, q < 0.001).

After FFM^0.67 scaling, hikite and tsurite remained associated (ρ = 0.534, p < 0.001, q < 0.001), whereas hikite was not associated with morote-seoi-nage (ρ = 0.141, p = 0.281, q = 0.281) and tsurite was not associated with morote-seoi-nage (ρ = 0.183, p = 0.162, q = 0.194). All six correlations were corrected as one secondary FDR family (Table 4).

Table 4.

Relationships among judo-specific peak-power outputs before and after FFM^0.67 scaling

Relationship	Absolute ρ	Absolute p	Absolute q	FFM^0.67-scaled ρ	Scaled p	Scaled q
Hikite vs. Tsurite	0.746	< 0.001	< 0.001	0.534	< 0.001	< 0.001
Hikite vs. morote-seoi-nage	0.556	< 0.001	< 0.001	0.141	0.281	0.281
Tsurite vs. morote-seoi-nage	0.585	< 0.001	< 0.001	0.183	0.162	0.194

Absolute and scaled coefficients, raw p values, and Benjamini-Hochberg FDR-adjusted q values are shown explicitly. The three absolute and three scaled correlations were corrected together as one six-test secondary family. N = 60; MSN morote-seoi-nage, FFM fat-free mass

Explanatory regression models for judo-specific peak power

The best-fitting absolute multivariable model for hikite included sex, squat 1RM, and CMJ peak force (adjusted R² = 0.799, ΔR² = 0.126, p for ΔF < 0.001); both squat 1RM (β = 0.328, p = 0.005) and CMJ peak force (β = 0.172, p = 0.034) were retained. The tsurite model included sex, SJ peak force, and CMJ stiffness (adjusted R² = 0.723, ΔR² = 0.188, p < 0.001); SJ peak force (β = 0.356, p = 0.004) and stiffness (β = 0.223, p = 0.016) were retained. The morote-seoi-nage model included sex, CMJ peak force, and CMJ stiffness (adjusted R² = 0.558, ΔR² = 0.508, p < 0.001); CMJ peak force (β = 0.543, p = 0.001) and stiffness (β = 0.325, p = 0.003) were retained, whereas sex was not significant.

For scaled hikite, the physical predictors did not improve the sex-only model (adjusted R² = 0.607, ΔR² = 0.017, p = 0.279). The scaled tsurite model improved beyond sex alone (adjusted R² = 0.503, ΔR² = 0.073, p = 0.018), with scaled SJ peak force retained (β = 0.244, p = 0.014). The scaled morote-seoi-nage model also improved beyond sex alone (adjusted R² = 0.191, ΔR² = 0.232, p < 0.001), with scaled CMJ peak force retained (β = 0.458, p < 0.001). The 0.75 sensitivity models yielded the same substantive interpretation (Supplementary Tables S5–S6B).

Table 5 presents the sex-controlled explanatory multivariable model estimates for the absolute judo-specific peak-power outcomes.

Table 5.

Sex-controlled explanatory multivariable models for absolute judo-specific peak-power outcomes

Outcome / term	B	95% CI	β	p (HC3)	VIF	R²	Adj. R²	ΔR²	p for ΔF
Hikite peak power: Intercept	164.265	[53.549, 274.980]	—	0.004	—	0.809	0.799	0.126	< 0.001
Hikite peak power: Sex (male = 1)	244.665	[167.294, 322.035]	0.502	< 0.001	2.00
Hikite peak power: Squat 1RM	1.930	[0.619, 3.241]	0.328	0.005	4.67
Hikite peak power: CMJ peak force	0.039	[0.003, 0.075]	0.172	0.034	3.36
Tsurite peak power: Intercept	72.754	[-29.471, 174.978]	—	0.160	—	0.737	0.723	0.188	< 0.001
Tsurite peak power: Sex (male = 1)	187.021	[107.488, 266.554]	0.472	< 0.001	1.66
Tsurite peak power: SJ peak force	0.070	[0.024, 0.116]	0.356	0.004	2.21
Tsurite peak power: CMJ stiffness	0.018	[0.003, 0.032]	0.223	0.016	1.43
Morote-seoi-nage peak power: Intercept	53.503	[-27.811, 134.817]	—	0.193	—	0.580	0.558	0.508	< 0.001
Morote-seoi-nage peak power: Sex (male = 1)	-26.122	[-85.377, 33.132]	-0.098	0.381	1.52
Morote-seoi-nage peak power: CMJ peak force	0.067	[0.031, 0.104]	0.543	< 0.001	2.73
Morote-seoi-nage peak power: CMJ stiffness	0.017	[0.006, 0.029]	0.325	0.003	2.01

B unstandardized coefficient, β standardized coefficient, HC3 heteroscedasticity-consistent standard error, VIF variance inflation factor, R² coefficient of determination, ΔR² change beyond the sex-only model, PP peak power, MSN morote-seoi-nage. Models are explanatory and sample-specific, not forecasts of future performance

Across the six exploratory sex-adjusted models, standardized β values for self-reported morote-seoi-nage tokui-waza status ranged from − 0.078 to 0.306; no coefficient remained significant after FDR correction (Supplementary Table S7).

Discussion

Main findings and absolute versus allometrically scaled interpretation

The main finding was that associations between general physical performance and judo-specific peak power depended on both the action and the expression of performance. Absolute associations were widespread across all three actions. After FFM scaling, multiple associations remained for hikite and tsurite, whereas scaled CMJ peak force was the only retained correlate of scaled morote-seoi-nage peak power. The same significance classification pattern was observed with the 0.75 sensitivity exponent when the identical 93-test FDR family was used. Thus, the data support complementary interpretation of absolute and scaled outcomes rather than a single uniform physical profile.

This distinction is relevant because judo is organized by weight categories, and body mass, body composition, strength, and anaerobic power differ according to sex, competitive level, and weight category [1, 2, 9]. The strong absolute associations therefore describe a meaningful sample pattern in which greater FFM and force-production capacity coexist with higher total judo-specific output; they should not be interpreted as evidence that increasing any one general test would cause an equivalent change in a judo-specific action.

However, absolute values do not distinguish whether higher performance is due to superior neuromuscular function relative to body size or simply to greater body dimensions. Methodological work has shown that normalization is required when comparing strength and power across individuals with different body dimensions, and that allometric scaling may be more appropriate than simple ratio standards in many contexts [18–20]. Similar concerns have been reported in judo-specific strength assessment, where simple ratio scaling did not adequately account for body-size effects in handgrip strength [21]. In the present study, the attenuation of several associations after allometric scaling suggests that part of the relationship between general physical qualities and judo-specific power was body-size dependent.

Using fat-free mass as the scaling denominator is also physiologically defensible because force and power production are more directly related to contractile tissue than to total body mass. This issue may be relevant in judo, where body composition varies across weight categories [26] and fat mass does not directly contribute to force production. Evidence in judo athletes suggests that total and regional fat percentage can be negatively associated with physical and judo-specific performance measures, supporting the need to distinguish body mass from fat-free mass when interpreting performance data [27].

Thus, absolute and allometrically scaled outcomes should be considered complementary rather than competing approaches. Absolute values describe the total mechanical output an athlete can produce, whereas scaled values help clarify whether physical qualities remain associated with judo-specific performance after accounting for fat-free mass. This distinction provides the basis for interpreting why hikite and tsurite retained several associations after scaling, whereas morote-seoi-nage showed a marked attenuation of associations with general physical variables.

Task specificity across isolated pulling actions and morote-seoi-nage

The different patterns for hikite, tsurite, and morote-seoi-nage indicate that these peak-power outcomes are not interchangeable. Hikite and tsurite are more constrained kuzushi-related pulling actions, whereas morote-seoi-nage is a multi-phase action requiring entry, positioning, upper-limb pulling, trunk rotation, lower-limb drive, and forward-downward completion. Hikite and tsurite remained associated after scaling, suggesting a shared task-specific pulling component that was not fully accounted for by FFM [11, 28, 29].

In contrast, the absolute associations of hikite and tsurite with morote-seoi-nage were not retained after scaling. This pattern suggests that their absolute covariation partly reflected shared body-size and total-output characteristics. It does not imply that either pulling action is unimportant; rather, isolated pulling power alone was not closely associated with the remaining between-athlete variation in the multi-phase task after FFM scaling.

These results are consistent with the principle that associations between general physical capacities and sport-specific outcomes depend on task similarity. Movement pattern, force direction, contraction characteristics, velocity, and inter-segmental coordination affect how closely two tests correspond [8, 16, 17]. The retained scaled CMJ peak-force association with morote-seoi-nage may reflect a relevant whole-body force component, but the cross-sectional design cannot identify a causal mechanism.

This distinction is relevant for interpreting judo-specific testing. Previous validation work indicates that judo-specific tests and ergometer-based assessments can capture meaningful neuromuscular and task-specific information, but different tests assess different performance dimensions and should not be interpreted as equivalent measures of complete technical ability [11, 12]. Thus, hikite and tsurite power may be useful indicators of kuzushi-related pulling capacity, whereas morote-seoi-nage power should be interpreted as a more complex task outcome that depends on physical qualities as well as technical skills.

Morote-seoi-nage task complexity

The exploratory sex-adjusted tokui-waza analysis did not retain any FDR-significant coefficient; the result is therefore hypothesis-generating and cannot isolate technique preference, practice history, competitive use, or proficiency.

The selective scaled association for morote-seoi-nage should be interpreted in relation to the technical and coordinative demands of the throw. Biomechanical evidence indicates substantial lower-limb and trunk contributions, whole-body rotation, and coordinated energy transfer across segments [13]. The present pneumatic-resistance assessment quantified whole-action peak power but could not identify how power was distributed across phases or segments.

This interpretation is supported by studies comparing seoi-nage execution across skill levels. Ishii et al. [15] reported that high-level seoi-nage performance was related to the forward drive of the thrower relative to the opponent and to rotational movement characteristics. More recently, Choi and Song [14] showed that elite and non-elite judo athletes differ across the unbalancing, positioning, and throwing phases of seoi-nage, suggesting that performance is influenced by the organization of the entire movement sequence rather than by isolated force production. These findings are relevant to the present study because the pneumatic-resistance assessment quantified whole-action power, but did not identify whether power was generated effectively through the lower limbs, trunk, or upper limbs.

Taken together, the findings characterize morote-seoi-nage peak power as a complex task outcome. General strength and jump-force tests provide associated information, but they do not replace direct observation of the technical sequence or direct judo-specific testing.

Practical applications

For routine monitoring, practitioners should retain absolute outputs when the applied question concerns total force or power capacity, especially within similar sex and weight-category contexts. FFM^0.67-scaled outputs should be considered alongside absolute values when comparing athletes of different body sizes or when body-composition changes could alter interpretation. The identical significance classification obtained with 0.75 under the same 93-test FDR family indicates that the present association pattern was not dependent on the single theoretical exponent chosen.

A parsimonious general testing panel can prioritize squat 1RM and force-platform jump measures. In the present sample, squat 1RM and CMJ peak force were retained in the hikite model, SJ peak force and CMJ stiffness in the tsurite model, and CMJ peak force and stiffness in the morote-seoi-nage model. These are sample-specific associations rather than evidence that changing a test score will cause a change in judo performance.

Direct hikite and tsurite peak-power tests can complement the general panel when the monitoring question concerns kuzushi-related pulling. Morote-seoi-nage should be assessed directly and interpreted with technical observation because its multi-phase output was not represented by a uniform general-test profile. Coaches should therefore review absolute output, FFM-scaled output, and action quality together rather than using one general test as a proxy for all judo-specific actions.

Limitations and future directions

First, the cross-sectional design precludes causal inference and external prediction. The regression models explain sample variance only; longitudinal studies are required to determine whether changes in general physical qualities are associated with subsequent changes in judo-specific performance.

Second, the sample comprised elite athletes from one Chinese national team. The findings may not generalize to other countries, competitive levels, age groups, or training systems, and the data-driven selection of best-fitting two-predictor models may yield optimistic estimates that require external validation.

Third, InBody 770 FFM is an estimate rather than a criterion measurement. In young adults, reported FFM total error and standard error of estimate were approximately 2.4 and 2.2 kg, respectively [22]. In a large physically active sample tested under uncontrolled conditions, InBody 770 FFM exceeded DXA estimates by 3.4 ± 2.8 kg in men and 2.0 ± 2.2 kg in women, despite high retest reliability [23]. Our protocol standardized fasting and pretest exercise but did not record a specific water-intake or voiding protocol. Residual hydration and method error may therefore have propagated into FFM-scaled variables, and available error estimates are not specific to elite judoka.

Fourth, the 20% body-mass load was selected from a morote-seoi-nage pilot and then applied to the isolated actions. Coach observation supported valid execution, and the common relative load standardized testing across body sizes, but 20% may not be the optimal power load for hikite or tsurite. Future work should establish action- and athlete-specific load-power profiles.

Fifth, the morote-seoi-nage assessment quantified whole-action peak power but not phase-specific kinetics, kinematics, or muscle activity. Future studies should combine power testing with motion capture, force measurements, electromyography, or phase-specific analysis.

Sixth, the global FDR procedure controlled multiplicity for the primary correlations, but the multivariable models were selected within the same sample and were not cross-validated. Their coefficients and adjusted R² values should therefore be treated as explanatory and sample-specific.

Seventh, tokui-waza status was self-reported and the groups differed in sex distribution. Even after sex adjustment, the exploratory analysis cannot isolate technique-specific practice, tactical preference, or proficiency and did not retain FDR-significant coefficients.

Eighth, the theoretical exponent was fixed rather than estimated for each variable. Repeating the principal scaled correlations with 0.75 under the same 93-test FDR family did not change any significance classification, but other denominators, segmental lean mass, or sample-derived exponents could yield different quantitative estimates.

These limitations favor replication using criterion body-composition methods, externally validated model specifications, and longitudinal or intervention designs.

Conclusions

Associations between general physical performance and judo-specific peak power in elite judoka depended on the action and on whether performance was expressed in absolute or FFM-scaled terms. Absolute associations were broad, whereas scaled associations were more selective: multiple associations remained for hikite and tsurite, and scaled CMJ peak force was the only retained correlate of scaled morote-seoi-nage peak power. The 0.75 sensitivity analysis, evaluated with the same 93-test FDR family, did not change these conclusions.

Absolute and scaled outcomes therefore provide complementary monitoring information. Squat and force-platform jump measures can form a parsimonious general testing panel, but they should be combined with direct judo-specific testing and technical observation. The findings are associative and sample-specific and should not be used to infer causal training effects or future performance.

Abbreviations

1RM

One-repetition maximum

CI

Confidence interval

CMJ

Countermovement jump

CV

Coefficient of variation

FFM

Fat-free mass

ICC

Intraclass correlation coefficient

IJF

International Judo Federation

IMTP

Isometric mid-thigh pull

MSN

Morote-seoi-nage

PP

Peak power

SD

Standard deviation

SEM

Standard error of measurement

SJ

Squat jump

VIF

Variance inflation factor

Authors’ contributions

AH contributed to conceptualization, study design, writing - original draft, and writing - review and editing. PW contributed to study design and data collection. EF contributed to writing - review and editing, manuscript revision, and guidance on statistical analysis methods. DW contributed to study design and data collection. LH, RJ, and YM contributed to data collection and statistical data organization. XZ contributed to supervision, study design guidance, and writing - review and editing. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets generated and analysed during the current study are not publicly available because they contain confidential performance information from active elite national-team athletes. Aggregated summary data are included in this article and its supplementary information files. Additional data may be available from the corresponding author on reasonable request, subject to institutional and participant-confidentiality restrictions.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committee of Shanghai University of Sport (approval no. 102772025RT256) and was conducted in accordance with the Declaration of Helsinki. All athletes were informed of the study procedures, potential risks, and benefits before participation and provided written informed consent.

Consent for publication

Written informed consent for publication of the identifiable action-demonstration images was obtained from the individual depicted in Fig. 2.

Competing interests

The authors declare no competing interests.