Scientists used artificial intelligence to figure out which carbohydrate-protein sports drinks work best for different athletes. They studied 231 rowing sessions and found that what works great for one person might not work as well for another. By analyzing body weight, muscle power, and eating habits, their computer program could predict which supplement would help each athlete perform better. This personalized approach could help athletes choose the right nutrition strategy instead of guessing what might work.

The Quick Take

  • What they studied: Can a computer program learn to predict which athletes will benefit most from carbohydrate-protein sports drinks by looking at their body type, fitness level, and diet?
  • Who participated: 231 rowing training sessions were analyzed. The study looked at athletes’ baseline fitness, body measurements, and what they ate to build the prediction model.
  • Key finding: The computer program successfully predicted how far athletes would row based on their personal characteristics and supplement choice, with about 53% accuracy. Body weight, explosive power (how fast muscles can work), and diet were the strongest clues for predicting who would benefit most.
  • What it means for you: Instead of all athletes using the same sports drink, this approach suggests personalized recommendations based on your body type and fitness level may work better. However, this is still research—talk to a coach or sports nutritionist before making big changes to your supplement routine.

The Research Details

Researchers collected data from 231 rowing workouts and recorded 46 different pieces of information about each athlete, including their weight, muscle power, and what they ate. They used a special computer technique called WGAN-GP (a type of artificial intelligence) to create fake but realistic training data to help their program learn better from the limited real data they had. This is like teaching a computer to recognize faces by showing it real photos plus computer-generated ones that look realistic.

They then tested several different computer programs (XGBoost, SVR, and MLP) to see which one could best predict how far each athlete would row based on their personal information and supplement choice. The best-performing program was then turned into a recommendation tool that could suggest which supplement would work best for each individual athlete.

This approach is important because real-world sports data is often limited and expensive to collect. By using artificial intelligence to create additional realistic training examples, the researchers could build a more reliable prediction system than they could with just the 231 real sessions alone.

Most sports nutrition research treats all athletes the same, but people’s bodies respond differently to supplements. This study shows that computers can learn to spot patterns in who benefits most from carbohydrate-protein drinks. This personalized approach could help athletes make smarter choices about their nutrition instead of following generic advice.

The study was published in Scientific Reports, a well-respected scientific journal. The researchers used multiple computer programs and compared them to find the best one. They also used a smart method to handle the problem of having limited training data. However, the study only looked at rowers, so results might be different for other sports. The 53% accuracy means the predictions are better than random guessing but not perfect—there’s still a lot of individual variation the computer can’t explain.

What the Results Show

The XGBoost computer program, when combined with the artificial data created by WGAN-GP, was the most accurate at predicting rowing performance. It achieved 53% accuracy, meaning it explained about half of the differences in how far athletes rowed. This is considered strong for predicting human performance, which is naturally very variable.

The computer program identified that 21 specific factors out of the original 46 were most important for predicting performance. The three strongest predictors were: (1) body weight, (2) explosive power (how quickly muscles can generate force), and (3) nutritional factors related to carbohydrate and protein intake.

The study shows that the same carbohydrate-protein supplement doesn’t work equally well for everyone. Athletes with different body types, fitness levels, and current diets showed different responses to the same supplement. This variation is why personalized recommendations could be more effective than one-size-fits-all advice.

The research also found that the artificial data created by the WGAN-GP technique was realistic enough to significantly improve the computer program’s learning. Without this artificial data, the predictions were less accurate. This suggests that when real data is limited, this artificial data generation method could be useful for other sports science questions. The study also showed that combining information about athletes’ bodies, fitness, and diet was more powerful than looking at any single factor alone.

Previous research has shown that carbohydrate-protein supplements generally help endurance athletes, but most studies treated all athletes the same. This research builds on that knowledge by showing that individual differences matter a lot. It aligns with growing evidence in sports science that personalized nutrition strategies may be more effective than generic recommendations. However, most personalized nutrition research has relied on expensive testing or small sample sizes, so this computer-based approach offers a new, scalable way to make predictions.

The study only looked at rowers, so the results might not apply to other endurance sports like running or cycling. The 53% accuracy, while good for predicting human performance, means the computer program can’t explain about half of the variation in results—individual factors we don’t know about still matter a lot. The study didn’t test whether athletes actually performed better when they followed the personalized recommendations in real training. Additionally, the research was based on historical data, so it might not work as well for athletes with very different characteristics than those in the original 231 sessions.

The Bottom Line

If you’re an endurance athlete interested in supplements, this research suggests that personalized recommendations based on your body type, fitness level, and current diet may be more effective than generic advice. However, confidence in this recommendation is moderate—the computer program’s predictions aren’t perfect. Before making changes, consider working with a sports nutritionist or coach who can evaluate your individual situation. The personalized approach is promising but should complement, not replace, professional guidance.

Endurance athletes (rowers, runners, cyclists) who use or are considering carbohydrate-protein supplements should find this interesting. Coaches and sports nutritionists might use this approach to make better recommendations. People new to endurance sports might benefit from personalized guidance rather than generic supplement advice. This research is less relevant for casual exercisers or strength athletes, as it specifically studied endurance performance.

If you were to follow a personalized supplement recommendation based on this research, you might notice performance differences within a few weeks of consistent training with the right supplement for your body type. However, improvements in endurance performance typically take 4-8 weeks to become noticeable. Don’t expect immediate results—consistent training combined with appropriate nutrition is what drives long-term improvements.

Want to Apply This Research?

  • Track your body weight, explosive power measurements (like vertical jump height or sprint time), current carbohydrate and protein intake, and endurance performance metrics (like distance covered or time to fatigue) weekly. This data would feed into a personalized recommendation engine similar to the one in this study.
  • Based on your tracked metrics, the app could suggest adjusting your carbohydrate-protein supplement ratio or timing. For example: ‘Based on your body weight and current power output, try increasing carbs by 10g per serving’ or ‘Your nutrition profile suggests a 3:1 carb-to-protein ratio may work best for you.’ Users could then test this recommendation during training.
  • Create a simple log that tracks: (1) supplement type and amount used, (2) workout distance/duration/intensity, (3) how you felt during the workout (energy level, fatigue), and (4) performance outcome. Review this monthly to see if personalized recommendations are actually improving your results compared to your previous generic approach.

This research presents a promising computer-based approach to personalized supplement recommendations, but it is not a substitute for professional medical or nutritional advice. The study was conducted on rowers and may not apply to other sports or populations. Before making significant changes to your supplement routine, especially if you have any health conditions, take medications, or have dietary restrictions, consult with a qualified sports nutritionist, registered dietitian, or your healthcare provider. The personalized recommendations from this framework should be used as a starting point for discussion with professionals, not as definitive medical advice. Individual results may vary significantly, and what works for one athlete may not work for another.