A cross-sectional online questionnaire was distributed to Swedish skiers, coaches, and ski technicians at all levels between October 13, 2021, and January 31, 2022. The questionnaire link was disseminated through local ski clubs, regional skiing networks, and public sport-related social media channels. Participation was voluntary and anonymous, and no incentives were offered. Respondents were eligible if they were active skiers (classic and/or skate) and at least 15 years of age. No additional exclusion criteria were applied. In accordance with the Swedish Ethical Review Act, this study did not require ethical approval as it involved only questionnaire data collection with no processing of personal information, no physical intervention, and no methods designed to physically or psychologically affect the participants. Participants provided informed consent via the survey platform and could withdraw at any time prior to submitting the questionnaire by closing the survey window.

A total of 249 participants completed the questionnaire. We categorised participants as experts (n = 20, 8%) if they met the following criteria:

Over 500 ski preparations per year.

The questionnaire addressed various aspects of ski preparation and glide perception across eight distinct snow conditions. The snow conditions were classified according to SWIX’s snow classification system (as of October 13, 2021), presented with icons in Table 1. Note that this system has been refined since the questionnaire administration. The list of questions and answer options is detailed in Table 2.

Participant characteristics and questionnaire responses were summarised and stratified by expert group. Descriptive statistics were presented as frequencies and percentages for categorical variables, and medians with interquartile ranges for continuous variables. To quantify the level of consensus among participants, we calculated the agreement coefficient A [39], for each question and snow type combination using the agrmt package in R, [40].

van der Eijk’s agreement coefficient A is designed explicitly for ordered rating data. It ranges from −1 (complete polarisation: responses concentrated at opposite ends) to +1 (complete agreement: responses concentrated in a single category), with 0 indicating a uniform distribution across categories. This measure addresses limitations of traditional agreement indices by accounting for the number of response categories and avoiding the assumption of uniform distributions [39]. Agreement was calculated separately for the overall sample, the expert group, and the non-expert group.

To compare responses between expert and non-expert groups, we calculated Cliff’s δ, a non-parametric effect size measure appropriate for ordinal data [37]. Cliff’s δ ranges from −1 to +1, where values near 0 indicate no difference between groups, and values approaching ±1 indicate complete separation between groups. Effect sizes were interpreted using cut-offs based on Vargha and Delaney [38], converted to Cliff’s δ using the relationship δ=2A−1. The resulting thresholds were: negligible (|δ|<0.11), small (|δ|≥0.11), medium (|δ|≥0.28), and large (|δ|≥0.43). These cutoffs correspond to Vargha-Delaney A values of 0.56, 0.64, and 0.71, respectively.

To address unbalanced sample sizes and within-participant dependencies, we applied a stratified bootstrap procedure. Resampling was performed at the participant level, with 10000 bootstrap iterations. Resampling was stratified by expert group to maintain the original proportion of experts and non-experts in each bootstrap sample. For each bootstrap sample, we calculated agreement coefficients (overall, expert, and non-expert) and Cliff’s δ for all question-snow type combinations. Ninety-five per cent confidence intervals were calculated as the 2.5th and 97.5th percentiles of the bootstrap distribution. To assess the statistical significance of group differences, we determined the proportion of bootstrap samples in which Cliff’s δ exceeded the threshold for a negligible effect size (|δ|>0.11). Differences were considered statistically meaningful when more than 95% of bootstrap samples showed |δ|>0.11 in a consistent direction, representing a minimal importance difference approach.

For question-level and snow type-level analyses, we averaged the metrics across the relevant dimensions within each bootstrap sample before calculating confidence intervals, thereby preserving the correlation structure in the data. A fixed random seed (4987) was used to ensure reproducibility of all bootstrap procedures. All statistical analyses were performed in R version 4.4.0 [41].

A total of 249 participants completed the questionnaire. The sample consisted predominantly of middle-aged males (40–59 years, n = 149, 60%). Regarding activity levels, the majority of respondents reported low frequencies of ski preparation (<100 per year) and moderate annual skiing distances (100–1,000 km). The expert group comprised 20 participants (8%), who were distinguished by high preparation volume and greater annual skiing distance. Detailed demographic characteristics stratified by group are summarised in Table 3.

Figure 1 provides an overview of the response distributions across all questions (Q1–Q10) and snow types (ST1–ST8), for experts (orange) and non-experts (black). Detailed numerical summaries are available in Supplementary Tables S1–S10.

Agreement for the full sample (both experts and non-experts) varied substantially across questions (Q1–Q10) and snow types (ST1–ST8), with stone-grind selection and perceived ski speed generating the highest consensus, whereas positioning and glide-determinant questions revealed considerable divergence in opinion. Figure 2 displays agreement coefficients by question, showing the highest values for stone grind structures (Q9) and perceived ski speed on different snow types (Q1, Q2). The lowest agreement was seen for double-poling positioning (Q5), glide determinants (Q7, Q8), and the paired glide test comparing skier mass (Q9). Figure 3 presents agreement by snow type, revealing that frozen wet snow (ST4) elicited the highest consensus, whereas new snow (ST1) and moist fine-grained snow (ST5) showed the lowest. The specific question–snow type combinations with the highest (A>0.85) and lowest (A≤0.00) agreement are summarised in Table 4. Complete agreement values for all combinations, including subgroup analyses, are provided in Supplementary Table S11.

Expert and non-expert respondents differed most notably in their views on glide determinants. Experts more frequently identified camber as the primary glide determinant, both for skate skiing (Q7; 48% vs. 38%, δ=−0.41, 95% CI [−0.56,−0.26]) and classic skiing (Q8; 61% vs. 47%, δ=−0.31, 95% CI [−0.50,−0.11]). Experts also more often preferred skiing in shadowed snow on sunny days for best glide (Q6; 73% vs. 69%, δ=−0.28, 95% CI [−0.38,−0.18]) and double-poling in the tracks (47% vs. 44%, δ=−0.23, 95% CI [−0.31,−0.14]). Full results are provided in Supplementary Tables S11–S12.

The agreement analysis showed that consensus was highest for structure selection and for perceived ski speed across snow types, whereas it was lowest for glide determinants, paired glide tests involving skier mass, and positioning during double poling. Moreover, frozen wet snow (ST4) elicited the highest agreement, while new snow (ST1) and moist fine-grained snow (ST5) showed the lowest levels of consensus.

These patterns are consistent with known differences in the physical stability of the underlying snow regimes. Frozen, coarse-grained and refrozen wet snow typically exhibits comparatively stable grain morphology and liquid-water content over the course of a race, creating a tribological environment in which established structural solutions tend to perform reliably across skiers and setups [18, 20–22, 29]. In contrast, new snow is thermodynamically less stable and undergoes relatively rapid metamorphism, including crystal rounding and changes in surface roughness, particularly near the melting point [29, 32]. Moist fine-grained snow similarly occupies an intermediate regime in which small variations in temperature, contamination, or traffic can alter water-film formation and contact mechanics in ways that are difficult to predict [11, 19].

From this perspective, high agreement in frozen wet snow likely reflects that technicians and skiers repeatedly encounter similar, physically stable conditions in which the same structural and waxing strategies tend to work. Conversely, the low agreement in new and moist fine-grained snow suggests that practitioners are confronted with more variable and less tractable conditions, which naturally foster a wider diversity of heuristics and preparation philosophies. This interpretation aligns with broader work on decision-making under uncertainty, in which less predictable environments are associated with greater reliance on idiosyncratic rules-of-thumb and higher between-person variability [33, 34].

The most profound divergence between experts and non-experts concerned the prioritisation of glide determinants. Experts consistently ranked mechanical properties, particularly the ski camber, above chemical preparation (waxes and additives). This aligns with tribological evidence showing that stiffness profiles govern the apparent contact area, pressure distribution, and load partitioning, and thereby dominate snow deformation, water-film generation, and the effective coefficient of friction [5, 20–23, 26, 27]. Non-experts, by contrast, placed less weight on camber, likely because waxing is visible, well-marketed, and easily adapted before skiing, whereas camber is a fixed mechanical property that requires specialised tools and access to many ski pairs to assess properly. In essence, the ski’s mechanical properties set the boundary conditions within which micro-scale interventions, such as waxing and hand structuring, can operate, with load distribution determining the real area of contact and thus the regime in which chemical and textural adjustments are effective [11].

The lower emphasis on camber among non-experts may reflect a relative overvaluation of chemical preparation relative to the mechanical properties of the ski. Wax choice and application are highly visible, frequently marketed, and can be adjusted shortly before competition, whereas camber selection is constrained by the available number of skis and requires more specialised measurement and testing. For recreational skiers working with a small number of ski pairs, it is therefore plausible that waxing appears as the dominant lever to influence glide, even though the underlying mechanics indicate that camber and load distribution are primary drivers of frictional behaviour.

This “expert gap” in the perceived hierarchy of glide determinants has practical implications. First, it suggests that education aimed at non-experts could profitably emphasise the foundational role of ski mechanics, e.g., by illustrating how camber mismatch can negate the benefits of otherwise optimal wax and structure choices. Second, it underscores the value of systematic ski-fleet management and camber profiling at higher performance levels, where technicians routinely handle many pairs of skis and can exploit mechanical optimisation more fully. Finally, recognising that experts and non-experts may talk past each other when using the same vocabulary (e.g., “fast skis”) but prioritising different underlying factors could improve communication between elite service teams and the broader skiing community.

This study has several limitations. First, the sample relied on voluntary participation and may over-represent skiers with above-average interest in ski preparation, potentially inflating consensus levels. Second, the expert subgroup was necessarily small (n = 20) to ensure validity as elite ski-technicians constitute a limited population. The small group size ensures validity but does limit the precision of estimates. Third, all responses were self-reported and represent subjective perceptions rather than direct physical measurements of glide performance. Fourth, the questionnaire format did not allow participants to seek clarification on ambiguous terminology, which may have introduced response variability. Interviews could have reduced such variability and potentially strengthened observed consensus levels. In addition, several specific sub-topics, such as snow contamination and temperature-dependent structural effects, were omitted to maintain a manageable questionnaire length.

While the reliance on perceptual data is a limitation, gliding properties in competitive settings are still evaluated through field testing and invariably rely on the skier’s subjective judgement [12]. Our findings should therefore be interpreted within the context of a sport in which perceptual assessment remains integral to ski preparation.

Future research should more directly bridge perception and physics by targeting the topics with lowest agreement (e.g., glide determinants, skier-mass effects in paired glide tests, and double-poling positioning) and combining survey responses with controlled friction measurements and field validation. Longitudinal studies may clarify how practitioners acquire and refine what, in effect, amounts to a form of tribology-based intuition, and whether perceptual consensus changes with accumulated experience. To reduce interpretation bias inherent to questionnaire items, qualitative follow-up (e.g., cognitive interviews/think-aloud protocols) should be used to refine terminology and determine whether response dispersion reflects true disagreement or heterogeneous interpretation. Finally, integrating objective on-snow measurements (e.g., GNSS/IMU-based speed and technique metrics, RTK-based friction estimation, and instrumented skis/poles and tribometer measurements) can provide scalable feedback that complements subjective glide testing and enhances technician training and evidence-based preparation practices for non-experts.