@frugaldoc:
Who knows to what extent the programmers and functional designers of these devices are on the same page. I was exerting myself quite strenuously, running up stairs, to achieve max heart rate, but generally my exertion has been quite moderate. I feel like all my metrics are In alignment with my own sense of being more like a chubby teenage girl than most women my age.
Ego's Journal
Re: Ego's Journal
You certainly have the max heart rate of a teenager. 
-
jacob
- Site Admin
- Posts: 17132
- Joined: Fri Jun 28, 2013 8:38 pm
- Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 77
- Contact:
Re: Ego's Journal
This should keep you entertained: https://towardsdatascience.com/when-you ... 7aa5bf452/
As a rule wrist monitors should be cinched down tight enough to leave a mark during exercise for maximum accuracy.
As a rule wrist monitors should be cinched down tight enough to leave a mark during exercise for maximum accuracy.
Re: Ego's Journal
I think the zones on a Garmin, by default, are calculated by a relatively simple formula. Something like 220-age. IIRC you can input a user determined maximum heart rate and tell it to use a heart rate reserve method instead. Or you can manually set them based upon your testing.
The Garmin I had could not intelligently determine the zones. I don't know if the fancier watches can. I set zone 2 manually, since that's what I cared about.
In my experience, the wrist based HR tracking does not handle intervals well. The recommendation is to use a chest strap for accurate readings during such work. My guess is that's why you're seeing such variance in the max HR during a workout. The 169 on your chart looks like an extreme spike. It could be a measurement error. Though I'd expect you're getting higher than 155. That seems low given age and training status.
Here's a grab from my ride this morning. I think the peak HR is likely, but the jump into zone 5 around minute 24 strikes me as a bad reading. The drop around minute 8 as well. While I did walk my bike for a minute, shedding 40 bpm seems unlikely:
I am using a Samsung watch. I feel like it has more measurement error than my Garmin.
Re: Ego's Journal
Interesting! I crank mine down tight. Mrs. Ego, not so much.
Actually, the 155 and 169 were captured while I was using a Garmin chest strap with the Fenix 6, but I agree that the 169 appears to be an aberration. Since I've been practicing endurance sports for many years, one of the things I watch for is signs of AFib. The Garmin does not have an ECG, but very high spikes are a telltale of AFib. I don't think 169 would be considered tachycardia for me because it returned to baseline right away, though I am not sure. Today I ran a slow, easy 10k all in zone 2 and my max HR was 146 and average 126. I felt like I could have had a full conversation the entire run. Next cycle class is Tuesday.Scott 2 wrote: ↑Sat May 24, 2025 11:40 amThe recommendation is to use a chest strap for accurate readings during such work. My guess is that's why you're seeing such variance in the max HR during a workout. The 169 on your chart looks like an extreme spike. It could be a measurement error. Though I'd expect you're getting higher than 155. That seems low given age and training status.
In other news, HRV is inching back up.
Re: Ego's Journal
I wouldn't call it tachycardia if you are not at rest. Such definitions are based on HRs taken with ECGs at rest. However, if you are seeing elevations that don't go down when you cease activity then you could have a tachyarrhythmia. Afib with RVR usually doesn't get overly high (110-160) but SVT can get quite high.
Re: Ego's Journal
Ah, good to know. Thank you @frugaldoc. A cycling friend had to curtail high intensity workouts because his watch detected abnormal rhythms, he went to the doctor and they somehow determined he suffered from AFib. I don't believe I have that problem.
Re: Ego's Journal
If you are having symptoms: dizziness, lightheadedness, chest pain, or shortness of breath then I would have it looked into otherwise I wouldn't worry too much about it. There is evidence of increased incidence of Afib in long-term endurance athletes once they reach middle age and beyond, but the positive benefits of said exercise outweigh this increased risk.
I am curious why your friend had to curtail his high intensity workouts. Perhaps he was put on a beta blocker and just has a decreased work capacity with the lower heart rate. I have never specifically advised someone with Afib to reduce the intensity of their workouts but granted the older patients I get usually don't exercise at all.
I am curious why your friend had to curtail his high intensity workouts. Perhaps he was put on a beta blocker and just has a decreased work capacity with the lower heart rate. I have never specifically advised someone with Afib to reduce the intensity of their workouts but granted the older patients I get usually don't exercise at all.
Re: Ego's Journal
I have zero symptoms, so not something I worry about very much. Thank you for lending your wisdom. Very much appreciated! I don't know why my friend is no longer doing high intensity work. I will ask him. He told me that he set his watch to warn if his HR goes above a particular level. Every so often he would just fall off the back of the peloton when his watch buzzed. He didn't like getting dropped, so he bought one of the $15k electric assist road bikes to stay with the group. That purchase prompted the whole conversation.
-
jacob
- Site Admin
- Posts: 17132
- Joined: Fri Jun 28, 2013 8:38 pm
- Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 77
- Contact:
Re: Ego's Journal
It's maybe helpful to know how those watches actually work and fail or develop spurious readings. The technology is pretty simple and uses the fact that the absorption spectrum for hemoglobin changes when it is carrying oxygen compared to when it doesn't. Imagine looking at a river that varying in terms of how bright or dark it is as it flows along. Now imagine that it's not flowing steadily but moves along according to the heartbeat. If you can measure and analyze the rhythm of that brightness variation, you got a way to indirectly measure the heart beat. What can possibly go wrong?
The first instruments used red light (where the hemoglobin is sensitive to the difference) and an infrared (where hemoglobin is not sensitive) reference light. You can measure "straight through" the finger with the emitter on one side and the receiver on the other side. Typically such an instrument would clamp down on the finger to eliminate all sources of external light or variation. You can put both on the same side and look at the reflection and now you're allowing for some light scatter as well as a change in the measuring spot because the instruments are no longer fixed.
One problem with red light is that it only works on people with translucent skin (white people aren't really white or pink for that matter, they're semi-transparent---shine a flashlight through your finger and you'll see what I mean). However, there's also a part of the green light spectrum that works for hemoglobin AND skin pigmentation, so gadget-builders switched to green light. This however means losing the IR reference beam which could previously be used to cancel some of these errors out.
You can eliminate some measurement error by fixing the instrument tight to the skin to eliminate movement (xyz as well as angle---imagine the kind of errors if the observer next to the river wasn't standing still but also wasn't aware that he was moving around and measuring different parts of the river or sometimes randomly turning the telescope towards the sky) and external light sources. Hence the recommendation for the tight band!
So now you have raw sensor data. This is turned into BPM or frequency by some kind of filtering algorithm. Filtering algorithms follow a trilemma in that they're either simple, fast, accurate (pick any two, but only two). If I had to bet money of what some gadget maker would put in a tiny watch chip, I'd go with simple&fast. Such algorithms tend to produce garbage data. They are INaccurate.
Garbage data is then cleaned up during "post-processing" (again, simple, fast, accurate, pick any two, but only..). Basically, if something "doesn't make sense", it's discarded. For example, if the watch measures 0 BPM for 45 seconds, it's almost surely not because the heart stopped but because either the instrument lost its "lock" (the watch was dangling too far from the wrist) or the algorithm lost its bearings and needs to sync up again. Likewise if the algorithmic rate suddenly shot up, it could be because the watch was jingled and instead of getting 3 flashes of light corresponding to three pulses of saturated blood, it got 5: 3 from the blood and 2 from a spurious light source like e.g. the bright sun. Boom! The heart rate just went up by 66% for that measurement interval. The watch has no way of knowing whether those 5 ticks came from the BPM shooting up OR from the wrist suddenly jerking the watch around admitting spurious light.
Then typically, you'd send the clean (washed?) data into another algorithm that would make sense of it based on some model of what it's measuring. This is easy when it comes to predicting the position of a falling rock. If you momentarily lose track of it, you have a pretty good idea of where to look to find it again, because gravity. Heart rate is rather more tricky. Insofar the actual input deviates from that model, typically the model just tries to make the best of it, sometimes returning a wrong result. For example, it might know that BPM doesn't instantly go from 150 to 200 or back again. But what if it got convinced that it was 200 (for example, due to the light noise mentioned above) ... then it might take a while before it accepts dropping back to 150BPM where it actually is... because of the "not instant" rule. Many assumptions will have to be made. This is non-trivial and depends on what the goal is. For example, is it better to avoid false negatives or false positives? Responsiveness is often a trade-off for confidence. Confidence is a trade-off for cheap and convenient.
These are issues that every sensor-maker struggles with when it comes to precision measurements, whether it's measuring ticker-data in finance, flashes of light in a subterranean neutrino detector, or in the fire control radar of a modern fighter plane. Most users don't really get to see or deal with all the assumptions made behind the scene to actually produce a "picture".
(Obviously, this also explains why you don't get the same picture if you're wearing two different monitors at the same time.)
All that to say is maybe not put too much faith into these overly simplified gadgets and the pictures they make. You have the navigational equivalent of a sextant on your wrist, not a GPS system or a complete map. For sure, a sextant is better than nothing and quite useful in skilled hands. It's still a navigational aid and not a complete substitute.
The first instruments used red light (where the hemoglobin is sensitive to the difference) and an infrared (where hemoglobin is not sensitive) reference light. You can measure "straight through" the finger with the emitter on one side and the receiver on the other side. Typically such an instrument would clamp down on the finger to eliminate all sources of external light or variation. You can put both on the same side and look at the reflection and now you're allowing for some light scatter as well as a change in the measuring spot because the instruments are no longer fixed.
One problem with red light is that it only works on people with translucent skin (white people aren't really white or pink for that matter, they're semi-transparent---shine a flashlight through your finger and you'll see what I mean). However, there's also a part of the green light spectrum that works for hemoglobin AND skin pigmentation, so gadget-builders switched to green light. This however means losing the IR reference beam which could previously be used to cancel some of these errors out.
You can eliminate some measurement error by fixing the instrument tight to the skin to eliminate movement (xyz as well as angle---imagine the kind of errors if the observer next to the river wasn't standing still but also wasn't aware that he was moving around and measuring different parts of the river or sometimes randomly turning the telescope towards the sky) and external light sources. Hence the recommendation for the tight band!
So now you have raw sensor data. This is turned into BPM or frequency by some kind of filtering algorithm. Filtering algorithms follow a trilemma in that they're either simple, fast, accurate (pick any two, but only two). If I had to bet money of what some gadget maker would put in a tiny watch chip, I'd go with simple&fast. Such algorithms tend to produce garbage data. They are INaccurate.
Garbage data is then cleaned up during "post-processing" (again, simple, fast, accurate, pick any two, but only..). Basically, if something "doesn't make sense", it's discarded. For example, if the watch measures 0 BPM for 45 seconds, it's almost surely not because the heart stopped but because either the instrument lost its "lock" (the watch was dangling too far from the wrist) or the algorithm lost its bearings and needs to sync up again. Likewise if the algorithmic rate suddenly shot up, it could be because the watch was jingled and instead of getting 3 flashes of light corresponding to three pulses of saturated blood, it got 5: 3 from the blood and 2 from a spurious light source like e.g. the bright sun. Boom! The heart rate just went up by 66% for that measurement interval. The watch has no way of knowing whether those 5 ticks came from the BPM shooting up OR from the wrist suddenly jerking the watch around admitting spurious light.
Then typically, you'd send the clean (washed?) data into another algorithm that would make sense of it based on some model of what it's measuring. This is easy when it comes to predicting the position of a falling rock. If you momentarily lose track of it, you have a pretty good idea of where to look to find it again, because gravity. Heart rate is rather more tricky. Insofar the actual input deviates from that model, typically the model just tries to make the best of it, sometimes returning a wrong result. For example, it might know that BPM doesn't instantly go from 150 to 200 or back again. But what if it got convinced that it was 200 (for example, due to the light noise mentioned above) ... then it might take a while before it accepts dropping back to 150BPM where it actually is... because of the "not instant" rule. Many assumptions will have to be made. This is non-trivial and depends on what the goal is. For example, is it better to avoid false negatives or false positives? Responsiveness is often a trade-off for confidence. Confidence is a trade-off for cheap and convenient.
These are issues that every sensor-maker struggles with when it comes to precision measurements, whether it's measuring ticker-data in finance, flashes of light in a subterranean neutrino detector, or in the fire control radar of a modern fighter plane. Most users don't really get to see or deal with all the assumptions made behind the scene to actually produce a "picture".
(Obviously, this also explains why you don't get the same picture if you're wearing two different monitors at the same time.)
All that to say is maybe not put too much faith into these overly simplified gadgets and the pictures they make. You have the navigational equivalent of a sextant on your wrist, not a GPS system or a complete map. For sure, a sextant is better than nothing and quite useful in skilled hands. It's still a navigational aid and not a complete substitute.
Re: Ego's Journal
@Jacob, as you know, for years I was very much against measuring this kind of stuff. Back then I believed that people tended to use them as a way to define the upper limit rather than a baseline and I still think that is true. In other words, they provide limits that are not actually true limits. I agree that the gadgets (at least mine) are generally providing piss poor data. It bothers me that I get data that shows poor performance (or sleep) when I felt like I went hard or visa versa. For now, I will continue to use it (it was free, after all) but I am going to go by feel rather than the "data" it is providing.
In other news, this week marked our first bi-annual apartment inspections since returning. The inspections involve our maintenance man entering every apartment and looking for any obvious signs of problems like leaks, mold, hoarding or anything else that needs to be fixed. The tenants always freak out when we post the notices that we will be entering their apartments. Usually, in the days leading up to the inspections the trash gets filled to the brim with interesting stuff.
This time around the bags full of clothing filled a large Amazon bin. I will take it to the swap meet and sell it for $2 a piece.
On top of one of the bins there was a little box with a tangle of trinkets that looked like costume jewelry. I took it back to the apartment and when I looked at it though a loop I saw the telltale 14K marked on one of the pieces. Today, I took it to the swap meet and handed it over to my friend who buys gold. Generally he pays 85% of spot. He dropped it on his scale, measured the weigh at 10.2g, rubbed them against his testing stone, put a little acid on the stone to confirm it was all gold, then peeled off six hundreds and asked me for $28 change. Insane.
In other news, this week marked our first bi-annual apartment inspections since returning. The inspections involve our maintenance man entering every apartment and looking for any obvious signs of problems like leaks, mold, hoarding or anything else that needs to be fixed. The tenants always freak out when we post the notices that we will be entering their apartments. Usually, in the days leading up to the inspections the trash gets filled to the brim with interesting stuff.
This time around the bags full of clothing filled a large Amazon bin. I will take it to the swap meet and sell it for $2 a piece.
On top of one of the bins there was a little box with a tangle of trinkets that looked like costume jewelry. I took it back to the apartment and when I looked at it though a loop I saw the telltale 14K marked on one of the pieces. Today, I took it to the swap meet and handed it over to my friend who buys gold. Generally he pays 85% of spot. He dropped it on his scale, measured the weigh at 10.2g, rubbed them against his testing stone, put a little acid on the stone to confirm it was all gold, then peeled off six hundreds and asked me for $28 change. Insane.
Re: Ego's Journal
Ego wrote: ↑Fri May 30, 2025 2:24 pm
This time around the bags full of clothing filled a large Amazon bin. I will take it to the swap meet and sell it for $2 a piece.
On top of one of the bins there was a little box with a tangle of trinkets that looked like costume jewelry. I took it back to the apartment and when I looked at it though a loop I saw the telltale 14K marked on one of the pieces. Today, I took it to the swap meet and handed it over to my friend who buys gold. Generally he pays 85% of spot. He dropped it on his scale, measured the weigh at 10.2g, rubbed them against his testing stone, put a little acid on the stone to confirm it was all gold, then peeled off six hundreds and asked me for $28 change. Insane.
