The task of discriminating between self and mutual assessment strategies is crucial to the study of animal contests. However, executing it can, sometimes, prove challenging. Despite a workable framework, some studies have obtained results that were inconsistent with self or mutual assessment. One possible cause could be the wrong use of traits to represent RHP, in relation to contest duration. By evaluating which contest model best explain fighting in Perisesarma eumolpe, a common mangrove crab species, we examined if the wrong usage of traits could indeed produce spurious results. Meanwhile, we also determined if contest behaviours quantified as contest intensity could serve as an additional mean to estimate RHP. Using multi-model inferences, mixed effects logistic regressions and linear mixed effects models, we analysed 78 fights from 27 staged contest trials, between randomly paired males. Contest intensity, body and weapon size were found to predict contest outcome. However, contest intensity was the only trait that best explained variation in contest duration. The effect of contest intensity on contest duration suggests that contest assessment in P. eumolpe were most consistent with cumulative assessment model. When effects of body and weapon size on contest duration were examined, results obtained were incompatible with known predictions. Our results thus suggest that studies with non-predictive results may have used traits that are poor predictors of contest duration, despite their ability to predict contest outcome.