Reaction Timing Errors Averaged 32.4 Milliseconds in 312 Combat Sports Athletes

Three hundred and twelve combat sports athletes tapped a screen to intercept a moving target, 60 times each. Their average timing error was 32.44 milliseconds, and only about 7% of attempts landed inside the window the researchers called accurate.

Combat sports athletes were profiled on a screen-based interception task in a cross-sectional study published on October 18, 2026 in Journal of Functional Morphology and Kinesiology. The authors built a single composite score from the results and then trained a machine learning model to reproduce it.

The sample was 312 athletes aged 10 to 25, with a mean age of 17.5 years and a standard deviation of 4.33. About 10% were female. Five combat sports were represented in roughly equal numbers: taekwondo, karate, Greco-Roman wrestling, freestyle wrestling and judo, with athletes ranging from beginners to the highly qualified.

Testing ran from late April to July 2026 in Kharkiv, Ukraine, across five venues including the Kharkiv State Academy of Physical Culture and four youth sports schools and clubs. Sessions took place in the second half of the day. The study was approved by the Bioethics Commission of the National University of Ukraine on Physical Education and Sport under Protocol No. 4, dated April 6, 2026.

The headline numbers are modest and specific. Across the group, the mean absolute timing deviation was 32.44 milliseconds with a standard deviation of 10.29, and the within-athlete variability of that deviation averaged 21.88 milliseconds. Nearly half of all responses came too early, a pattern that held across the five combat sports in the sample.

What the Combat Sports Test Actually Measures

The task is not simple reaction time. Athletes used an iPad of the ninth generation with a 10.2-inch screen, running an app called Reaction RMO Pro. An object travels across the display toward a target position, and the athlete responds at the moment it should arrive.

What the app records is the gap between the response and the instant the object reaches that target, rather than the delay from a stimulus appearing. That makes it an anticipation measure, which is arguably closer to what combat sports actually demand than a lab buzzer-and-button test.

Three stages built up the difficulty: a slowly moving object, a fast-moving object, and a fast-moving object with interfering visual stimuli. Each stage carried 20 attempts, so every athlete logged 60 scored trials. A practice trial came first and was excluded from analysis. A response within minus 3 to plus 3 milliseconds counted as accurate, anything earlier was premature, anything later delayed.

The Numbers Behind the Average

The split between early and late responses is the most striking part of the descriptive data. Premature responses made up 48.18% of attempts with a standard deviation of 10.18, delayed responses 44.66% with a standard deviation of 10.45, and accurate responses just 7.16% with a standard deviation of 4.40.

In other words, these athletes were far more likely to jump early than to wait too long, though the margin is slim. A 6-millisecond accuracy window is unforgiving by design, and the low hit rate says more about the width of that window than about the standard of the combat sports athletes tested.

Variability matters as much as the average here. An athlete who is 30 milliseconds off every single time behaves differently in sparring from one who alternates between 5 and 60 milliseconds, and the study treats both halves of that picture as information worth keeping for combat sports profiling.

How the Composite Index Was Built

Rather than report two separate columns, the authors standardized the mean deviation and its variability into z-scores and combined them into one composite index. Athletes were then sorted into three levels by tertile cut-offs of minus 0.56 and 1.08, giving 104 athletes in each of the High, Medium and Low groups.

Those cut-offs were stress-tested with bootstrap resampling across 10,000 iterations. The lower boundary sat in a range of roughly minus 0.77 to minus 0.35, and the upper boundary between 0.94 and 1.19. Sensitivity checks agreed with the original classification between 94.9% and 98.1% of the time, with kappa values from 0.923 to 0.971.

The labels describe position within this group of combat sports athletes, not a universal standard, and the authors are explicit that the levels are relative. The same approach has run into trouble elsewhere in sport science: our report on how inconsistently countermovement jump fatigue testing is reported found 117 studies pulling in different directions on exactly this kind of thresholding question.

What the Machine Learning Model Added

The authors trained a boosted-tree classifier in Create ML, using nine predictors: stage-specific means and standard deviations plus the premature, delayed and accurate percentages. The model ran to a maximum depth of 3 across 100 iterations, with no discipline label fed in, so it worked from timing behavior alone rather than from which of the combat sports an athlete practiced.

They split the data by stratified sampling into 264 observations for training and validation and a held-out test set of 48 athletes, 16 in each level. On the training and validation data Create ML reported 100% accuracy, which is an internal figure and should not be read as performance on new athletes.

The number that counts is the held-out result: 91.67%, or 44 of 48 correct, with a 95% confidence interval from 80.0 to 97.7. Per-class F1 scores were 0.94 for Low, 0.93 for High and 0.88 for Medium. All four errors landed on an adjacent level, two High cases slipping to Medium and two Medium cases to Low, which is the least damaging kind of mistake a grading model can make.

Validating a test against expert judgment is a familiar step. An automated ventilatory threshold algorithm checked against expert readers cleared a similar bar earlier this year, and ran into the same question of what to do when the human reference itself disagrees.

Why Combat Sports Coaches Might Care

Timing is the currency of striking and grappling exchanges, and most gyms assess it by eye. A 60-trial tablet test that produces a single number is cheap, repeatable and does not require a sports science laboratory, which is a real practical advantage for combat sports programs working on small budgets.

The catch is that this study never connected its levels to anything that happens in a bout. There is no competition result, no coach rating, no sparring metric in the analysis. Until that link is tested, a High or Low label is a description of screen behavior among combat sports athletes and nothing more.

Fatigue is the obvious next variable. Work on serve accuracy under rising workload showed precision holding at 60% of a players workload and falling at 80%, and a timing test done fresh in the afternoon may say little about the same athlete in round three.

How to Read a Reaction Score Without Overreading It

For an athlete or coach handed a number from a test like this, the useful posture is curiosity rather than conclusion. A single session on a tablet, in the afternoon, with no repeat measurement, is a snapshot of one afternoon in one gym. The authors say as much, and combat sports testing has a long history of numbers that travel further than their evidence.

General and safe ways to treat such a score:

• Treat it as a baseline to repeat, not a verdict; without a second session there is no way to know how stable it is.
• Compare an athlete only with athletes tested on the same device, since touch and display latency were not controlled across hardware.
• Watch variability as closely as the average, because consistency is what the composite index rewards.
• Do not use a level label to select, cut or stream young athletes; nothing in this study links the labels to competitive outcomes.

That last point matters most in youth combat sports, where early selection decisions carry real costs. Sports medicine researchers have argued that early talent selection may miss future elite athletes altogether, and a novel index is exactly the kind of tool that invites premature sorting.

Combat sports training also has plenty of better-established levers. Our coverage of cluster training in combat sport athletes found it matched traditional sets for neuromuscular gains in 24 athletes, which is the sort of question a combat sports coach can act on today.