2,000m Ergometer Time Estimator

[cavanhagan]

Hey everyone,

I started this research into predicting 2,000m time late in 2019 and the entire research article is due to be finished in the coming months. I created a calculator that will predict your 2,000m ergometer time based on 5 key predictive variables, and it is 82.2% accurate. It also estimates your single scull time based on the data but this isn’t very useful due it it’s limitations:

https://cavanhagan.com/rowing-calculator

I hope you can all try it out and let me know how close it is to your actual 2,000m time. Then if you wish you can click the link on the page to submit your own data so I can update the model in the future with more data.

I believe this calculator is a good supplement to talent identification and team selection, and should help validate goal setting into and creating tailored training programmes. It is just interesting to play around with to see what changes result in the biggest changes to your 2,000m time. Please do note the limitations in the model, however. When I have time I will update the page further with information, along with the entire article when completed.

Thanks,

Cavan

[Dangerscouse]

I've just inputted my current 5k PB, and a best guess for a 100m (I've never tried it) and it did give a result of my current 2k PB, so it looks accurate to me.

[cavanhagan]

I've just inputted my current 5k PB, and a best guess for a 100m (I've never tried it) and it did give a result of my current 2k PB, so it looks accurate to me.

Thanks for the reply! I’m glad that it was spot on! You really got to try a 100m they’re great fun

[MartinSH4321]

I got a 6:19 2k estimation which is a lot faster than I can do. IMO a sprint result does not work good for a 2k estimation, esp. a 100m, as it's totally anaerobic, for strong people the estimation is too fast, for very aerobic people too slow. Also, technique is very different for sprints.

[cavanhagan]

I got a 6:19 2k estimation which is a lot faster than I can do. IMO a sprint result does not work good for a 2k estimation, esp. a 100m, as it's totally anaerobic, for strong people the estimation is too fast, for very aerobic people too slow. Also, technique is very different for sprints.

Well the evidence I found shows that 100m time was the second strongest indicator of 2,000m time next to the 5,000m of course. The ratio lines up pretty closely with the evidence showing that 2,000m is 67-79% aerobic glycolysis and 21-33% anaerobic. That was the point of the 100m test, to measure your anaerobic power. 5,000m seemed like the most commonly known long piece- but could have been 6k, 30 minute or 10k I just chose what I thought most people would know.

Looking at your stats though you definitely are much more of an anaerobic athlete. As I said one of the main purposes of this calculator is to help you set goals and make better training programs. You’ll find that taking 1 second off your 100m- which is very difficult to do of course- only reduces your 2k by 5 seconds. This require you to turn your 1:09 split into a 1:04 split (-7.2% change). Though, according to this data, reducing your 5k by 25 seconds would achieve the same 5 seconds. This would require you to reduce your 5k split from 1:47.3 down to 1:44.8 (-2.3% change). So if your goal was to improve your 2k time I think this makes its pretty clear what the better option is. Obviously you already recognised this, but I think this helps to identify and quantify this. And you can estimate that if you get your 5k down to below 17:30 you can get your 2k down to around 6:38.

Hope this helps anyway, sorry it wasn’t terribly accurate for you but sometimes the extremes push the linear model’s limits and reduces its accuracy. And on that note, submitting your own data through the link on the page will help make it more representative of you, if you’d like to do that. Will be a while before I update the model though.

[MartinSH4321]

I got a 6:19 2k estimation which is a lot faster than I can do. IMO a sprint result does not work good for a 2k estimation, esp. a 100m, as it's totally anaerobic, for strong people the estimation is too fast, for very aerobic people too slow. Also, technique is very different for sprints.

Well the evidence I found shows that 100m time was the second strongest indicator of 2,000m time next to the 5,000m of course. The ratio lines up pretty closely with the evidence showing that 2,000m is 67-79% aerobic glycolysis and 21-33% anaerobic. That was the point of the 100m test, to measure your anaerobic power. 5,000m seemed like the most commonly known long piece- but could have been 6k, 30 minute or 10k I just chose what I thought most people would know.

Looking at your stats though you definitely are much more of an anaerobic athlete. As I said one of the main purposes of this calculator is to help you set goals and make better training programs. You’ll find that taking 1 second off your 100m- which is very difficult to do of course- only reduces your 2k by 5 seconds. This require you to turn your 1:09 split into a 1:04 split (-7.2% change). Though, according to this data, reducing your 5k by 25 seconds would achieve the same 5 seconds. This would require you to reduce your 5k split from 1:47.3 down to 1:44.8 (-2.3% change). So if your goal was to improve your 2k time I think this makes its pretty clear what the better option is. Obviously you already recognised this, but I think this helps to identify and quantify this. And you can estimate that if you get your 5k down to below 17:30 you can get your 2k down to around 6:38.

Hope this helps anyway, sorry it wasn’t terribly accurate for you but sometimes the extremes push the linear model’s limits and reduces its accuracy. And on that note, submitting your own data through the link on the page will help make it more representative of you, if you’d like to do that. Will be a while before I update the model though.

Thanks for your answer!
This evidence is very interesting, but, for me, two things can be problematic.
First, according to your page your sample consists of 111 club rowers. The difference in body composition between club rowers propably is much lower than between "normal" people like here in the forum. You won't find many strength athletes, overweights etc at a rowing club I guess.
Second, the 100m is an "exotic" distance, many rowers here in the forum avoid it, there is a much higher injury risk. And preparing the erg for a 100m (like weights on the front, elevated feet) and using the right technique can significantly improve the result, I would estimate around 1 second faster.
Without knowing the data, I would estimate that a 1k or 500m result would work better for "normal" rowers.

There's a predictor that some ergers here have used and most find it quite accurate, maybe you already know it:
https://www.freespiritsrowing.com/forum ... -predictor

[cavanhagan]

Yea you’re absolutely right, the sample is biased, and is very heavily experienced rowers. And yes you’re right about 100m, but for a power measurement it needs to be short and have minimal aerobic input. Even a 500m has quite a big aerobic component. I could certainly ask for more distances within the questionnaire from now on.

That calculator seems accurate to me. It’s seems it based on Paul’s Law but I can’t confirm that. Thanks for sharing!

[Nomath]

I had a quick try on 100m, which I have never done before. A high drag factor (150 instead of the ususal 120) proved beneficial.
Using the 100m result as input, I got 8:19 predicted ; my seasons best is 8:23. Not bad.

However, when I just use the power law: t(2K)/t(5K) = (2000/5000)^1.07 , I get t(2K) = 8:10.
i.e. without age, weight, arm span and 100m time as inputs.
The power 1.07 was found by a regression for some 20 persons of various ages and undisclosed weights who listed their season's best on 0.5K - 2K - 5K - 10K in the Concept2 2020 rankings.

How did you measure accuracy?
82% accurate sounds good, but for rowing pace (not wattage!) I think it is very poor. The power law prediction above is 97% of the realized time. My guess is that if you start from the 5k times, the power law will get above 90% accuracy.

[Nomath]

(double)

[MPx]

Pretty much spot on for me.

[cavanhagan]

I had a quick try on 100m, which I have never done before. A high drag factor (150 instead of the ususal 120) proved beneficial.
Using the 100m result as input, I got 8:19 predicted ; my seasons best is 8:23. Not bad.

However, when I just use the power law: t(2K)/t(5K) = (2000/5000)^1.07 , I get t(2K) = 8:10.
i.e. without age, weight, arm span and 100m time as inputs.
The power 1.07 was found by a regression for some 20 persons of various ages and undisclosed weights who listed their PB on 0.5K - 2K - 5K - 10K in the Concept2 2020 rankings.

How did you measure accuracy?
82% accurate sounds good, but for rowing pace (not wattage!) I think it is very poor. The power law prediction above is 97% of the realized time. My guess is that if you start from the 5k times, the power law will get above 90% accuracy.

Hey,

I’m glad it turned out pretty well for you! Accuracy is the R^2 value for the dataset. It means the model explains 82.2% of the variation of 2,000m time for the participants. So I guess you could say that 82.2% of the time it will get your score right. It’s a pretty strong accuracy, there are quite a few papers doing the same thing as this except they use some complicated tests. This one uses data you will already know or will be easy for you to measure. This model lies within the same accuracy range as those papers. Just to confirm, it does mean that your result can range 17.8%. Planning to make an even more sophisticated model but need to collect more data for it.

[Tony Cook]

PB 6:31. Calculator = 6:29.
Had a play with the numbers and if I can grow each arm 3cm then I’d shave another 2 seconds off

[cavanhagan]

Sounds pretty close to me! I hear the people in medieval times had a good machine for making your arms longer. I think they called it the rack

[Citroen]

https://cavanhagan.com/rowing-calculator

I hope you can all try it out and let me know how close it is to your actual 2,000m time. Then if you wish you can click the link on the page to submit your own data so I can update the model in the future with more data.

I tried it with my ancient PB times and it came out dead on.
The problem I had was seeing the result, pale text on a white field isn't ideal.

Make the text of the result larger. Push out the results in a BOLD font.

[Ernits]

My result was within 1 second of my 2K PB, seems accurate indeed!