Pool swim

I understand it. The point is that it should be possible to use the compass data to determine if you have changed swimming direction in the pool, and if you have changed swimming style. It would require analysis, just like it is done with gyroscope data or accelerometer data today.

ovekvam said:

it should be possible to use the compass data to determine if you have changed swimming direction in the pool,

I don’t understand  how a compass can be used to determine a change in the direction of a swimmer’s movement when the compass is attached to the wrist and in one stroke the compass will show at least two opposite directions. From a compass point of view, a turn is the same as a change of direction - how does this differ from a stroke?

A separate question is the rate at which the course is determined using the compass, the hand does not move in a straight line, should the readings be averaged, over what interval?

Find some kind of pattern based on the compass readings on your wrist, and the deviation from it?  What watch power and compass accuracy will be required for this search and discovery? Is this even real?

How the definition of style will help here is unclear to me, after a turn, the style can also change, it may not change, or it can change without turning

You can fantasize, but in my opinion you shouldn’t overestimate yourself and underestimate the specialists from Garmin.

3878047 said:

I think the algorithm could be easily improved

 (learn, stupid ones), well, well.

Well, I am a robotics engineer, and it does not sound all that complicated or computing intensive to me. Remember the watch knows how the arm is moving and rotating from the gyro and acceleration data. After all, most of the counting is based on the glide after push off. 

It seems to me that, from a compass point of view, it is impossible to distinguish the S trajectory of a freestyle stroke from an flip turns. What do you think?

ovekvam said:

After all, most of the counting is based on the glide after push off. 

I don’t understand, if glide after push is detected, then there is no need to detect turns. It's already been discovered

That is Garmin's explanation about how they count lengths. They look for the push off and glide. 
But it would also be possible to discover that the compass is opposite direction in the swim pattern from gyro and accelerometer, and count lengths by that.

I don't think anyone is trying to detect the actual turn.

ovekvam said:

That is Garmin's explanation about how they count lengths.

I know this, this is the existing algorithm, and I like it

4684381 said:

The existing algorithm, in my opinion, is very smart.

I have doubts about the need to change this algorithm by using a compass to improve it, I am discussing this issue.

The counting is spot on for me, and I am not a very good swimmer. It also seems to work well for good swimmers, but some people complain it is not able to count correctly. Not even close. Maybe they have a strange swimming style.

3878047 said:

where it is absolutely clear that there cannot be 2 turns in such a very short timeframe

This statement is not clear to me, what does a very short time interval mean?

Even if we assume some kind of definition, what should we do if there are several such intervals in a row, the end of which of them corresponds to the turn?

In my opinion, for the conditions you wrote about, it’s better to just use drill log.

For the opposite case I feel like sometimes when my watch misses a turn it catches up eventually.  My theory is it has seen me make n turns about every 20 seconds it is obvious the one it missed should have been there given some other independent data like accelerometer - assuming it doesn't use GPS and couldn't in any case reliably tell one end from the other.  I'm probably just be kidding myself because I usually don't keep a good mental count of laps - that's why I have the watch.  TBH I'm not swimming and logging my swims for performance / training so mostly I look at total time and heart rate and laps only count toward the occasional challenge that's based on distance.