PM5 averages: how “deep” is the data

[nick rockliff]

The averaging for everything is done in the same naïve way.

Grab an instantaneous reading at the end of a split. Sum the readings and divide by the number of splits. That's the same for rate, calories, HR, everything.

That's how it's worked since the PM2. Yes it is a crap design. It's not changed despite the PMs getting ever more RAM, ever more flash ROM and ever faster processors.

A rolling average would be a better idea.

So what am I seeing on the PM5 if its not a rolling average?

[Citroen]

So what am I seeing on the PM5 if its not a rolling average?

It's not a true rolling average by any stretch of mathematics. As soon as you cross the end of the next split it changes by always applying the same naïve average. If you want better stats and better readings use Ergdata on your phone.

[nick rockliff]

So what am I seeing on the PM5 if its not a rolling average?

It's not a true rolling average by any stretch of mathematics. As soon as you cross the end of the next split it changes by always applying the same naïve average. If you want better stats and better readings use Ergdata on your phone.

Are you sure Dougie? Looks to me as if it updates after every stroke not at the end of each split? I do use Ergdata but only for stroke count.

It doesn't really bother me because I've been looking at the PM as long as you have. I take notice of four readings as I row.

1. Stroke rate
2. Pace
3. Rolling average session pace
4. Rolling average split pace

Less so predicted time or distance.

I've never seen 3 update only at the end of the split. 4 will update at the end of each split.

[Citroen]

It may have been improved with recent PM5 firmware updates. It's now exactly one year since I sat on an ergo.

When I'm rowing I care about one value; the metres left or the time left and nothing else.

[Carl Watts]

It may have been improved with recent PM5 firmware updates. It's now exactly one year since I sat on an ergo.

When I'm rowing I care about one value; the metres left or the time left and nothing else.

Yeah people getting lost in the data I think, even the PM2 had the core of what you needed in reality, it just didn't have the wants like USB internet connectivity and the extra bits for the wireless, logbook etc. that people expect these days that are "Nice" to have.

[jackarabit]

I would not call the calculation of predicted finish “naïve.” It demands iterative updating every stroke, something I’m not capable of doing by very fast movements of a pencil stub! Is ‘naive’ a synonym for simple as in the binary: simple/complex? I’m off to discover the quality of naïvete in mathematical operations

[Carl Watts]

I would not call the calculation of predicted finish “naïve.” It demands iterative updating every stroke, something I’m not capable of doing by very fast movements of a pencil stub! Is ‘naive’ a synonym for simple as in the binary: simple/complex? I’m off to discover the quality of naïvete in mathematical operations

I agree if there is anything to complain about its the same format that has never been improved in the form of the results in the memory. I'm lead to believe it was limited due to the amount of memory in the PM3 and the very limited memory of the LogCard, which it was because you couldn't even get a years worth of results on it but then the PM5 came along with pretty much unlimited memory and the format remained unchanged. There certainly could be options on the number of recorded splits now and the average spm for the whole row in the top line could be fixed with an accurate measurement to 0.1spm.

[Tsnor]

In any case, the actual data gets exported nicely into a spreadsheet, including rest strokes taken during intervals with rest. People who want to poke the data can download it into spread sheets and play with it.

I'd prefer a heart rate average that is not biased up if you push at the end of each measured interval or down if you push the first few strokes at the start of the next interval. That said, changing it now without renaming data fields would be bad -- data compares with prior workouts would now be wrong.

(aside keeping an accurate average heart rate does not take a lot of memory. That can't/shouldn't be the reason the Pm5 does what it does. One counter field updated each time the heart beats is all it takes to get an average - say 4 bytes of memory per interval. The PM5 already knows the elapsed time. And dropped data is no worse than with the current approach. Similar approach works with the other data values.)