I’ve already described which apps and wearables I was using but last few months my QS stack were updated. But let me explain logic I’m using to choose a good tracking device.
Collecting data is not free and we pay in terms of:
Time spent to maintain protocols. Purchase, setup, prepare, sync/export, fix some stuff etc
Comfort. Wearing something introduce discomfort which should be in acceptable range
Time to process data, analyse and derive insights. Low data quality requires more time and advanced analytical skills
But thats not as easy as it may look:
Read studies to find out most practical / valid design. No reason to repeat errors made by pioneers which can be avoided by reading few studies.
Having clue about desired effect size translates to minimum acceptable accuracy for tracking device.
Choosing right device is a trade-off between price and comfort. Trade as less accuracy as possible.
And a few cautions:
Best devices provide unmodified raw data with full resolution in a similar units / range as gold-standard.
Device must be validated by independent research otherwise its accuracy is unknown and no reason buy it. Big companies spending a lot of money buying reviewers / health influencers to advertise inaccurate/unvalidated devices as being best and validated by research
Avoid high-processed aggregated data if possible. Sadly, sometimes device is good, but it is not possible to get full resolution data.
Device software must have raw data export / raw data web api
Sleep is #1 priority for health and longevity and might be most serious tracking in my QS stack. I’ve used several non-EEG wearables and all of them were not accurate enough to measure effect sizes of interest. I’m healthy and have ~90 minutes of N3 per night and the effect sizes for most powerful interventions do not exceed 20-30 minute range. So error should be less than 10 minutes or so to see the effect of that power. This is 10/90=11% error and is already in a sleep specialist personal / ML approach error range. That means device should be almost in a gold standard range to have good chances to detect even strong effect.
Dreem 2 is my first EEG device, but company went bankrupt and I’ve switched to Hypnodyne ZMax which is more comfortable than Dreem 2 and provides raw data. Both devices are in a gold-standard accuracy range or just a bit lower.
But my current king of sleep is OpenBCI:
Fully open-source hardware, firmware and software. Session can be streamed to pc and/or written to SD card and doesnt depend on manufacturer at all.
8-16 channels for EEG/EMG/EOG with high sampling rate (500-1000Hz) can be used with preferred electrodes. I use gold-cup electodes and Ten20 conductive paste. This combo provides gold-standard EEG data for sleep stage classification.
DIY headband for fast montage (less than 5 minutes) and automation. The weakest point is non-standard montage
YASA can be used for simple auto-scoring. New / other algorithms can be applied to past data.
With this approach I assume that my data will be in gold-standard range for next 50+ years (Lindy effect) and its unlikely to become obsolete. That means i can make comparisons during period of my whole life and do not lose data continuity even when i change to another gold-standard device. I’ve travelled few times with this setup - so far so good.
Cardiovascular system seems to be the weakest point of human health leading to biggest cause of death on a planet right now. Heart rate during sleep can reveal sickness, nightly HRV seems to be even more sensitive to sickness and morning standing HRV can reveal ANS status and training adaptation. Daily HRV can be used to reveal stress levels.
A lot of devices can track heart rate during sleep, but not all can track it during daytime or exercise:
Oura 3 for sleep resting HR
Fitbit Charge 6 for resting HR during daytime and sleep
Shimmer3 ECG 24/7 in day/night/exercise. Sometimes data is bad during exercise.
Polar H10 for exercise. Always good data during exercise.
I also sleep on Eight sleep but not using its metrics at all (data doesnt looks well compared to other devices).
HRV is harder to measure due to sensitivty to movements artifacts and arrithmia. PPG can measure it with acceptable accuracy only during stationary periods and even there it might be not accurate if one have frequent arrithmia.
Oura 3 and Fitbit Charge 6 for average nightly HRV. Full resolution data doesnt look accurate enough
Polar H10 for exercise. Sometimes data is bad during exercise.
Movesense MD. Using only for morning standing 1-minute HRV test. One can do same with Polar H10 if 512Hz raw ECG data arent needed.
Shimmer3 ECG 24/7. Sometimes data is bad during exercise.
Shimmer3 ECG is used to get gold-standard raw ECG data in 512Hz for 24/7 monitoring making my cardiovascular system tracking approach as solid as my sleep tracking.
Withings BPM for occasional measurements
[Now testing] Aktiia for 24/7 measurements
Body and skin temperature can be used for sickness monitoring and to reveal circadian rhythm
Calera research. Accurate enough and fast (1-2 days) circadian rhythm detection. Now i dont need my iButtons anymore. Using for occasional circadion rhythm assessent and to verify my circadian night is aligned with bedtime.
Kinsa Smart. Accurate and pretty fast oral temp tracking.
Withings BodyScan. It also provide a lot of metrics but i dont use most of them. Weighting takes ~1-1.5 minute and i combine it with morning standing HRV test.
Resistance training. Polar Vantage V + H10. Each exercise type with set/rep/weight/rpe is recorded manually for Training Load calculation. Overall session RPE also recorded
Running. Polar Vantage V + H10 + Stryd. Overall session RPE
Cycling. Polar Vantage V + Garmin Edge 1040 + Garmin Rally. Overall session RPE
I do meditate occasionally and tried different styles: NSDR, Mindfulness, Mantra, Dantian breathing. Now I’m building approach to analyse brain activity during meditation. I’ve used Muse 2 and Muse S for that but unsatisfied with signal qualtity.
Right now i’m switched to OpenBCI and have tried 16 channel setup with 1000Hz sampling rate:
EEG Cap with Sigma Gel. Signal quality is pretty good. Montage time is ~7 minutes and ~2 minutes to check everything and start a session. After session it takes 3-4 minutes to clean everything and put device into charger.
Ultracortex with passive Dry Comb electrodes. Took some time to setup first time but now it takes 1-2 minutes to start a session and nothing need to be done at the end. Not sure how long the electrodes will last. Signal quality is worse than with wet EEG cap, but better than with Muse. Easy to catch powerline noise. Due to prep time this is a preferred setup for now
I’ve ordered Thinkpulse electrodes. Since they active it may improve powerline noise resistance and allow for prolonged use due to electrodes material.
It is hard to travel with Ultracortex, this is where i might use ZMax / Muse for convenience but do not expect good data.
Mood patterns. Recently i’ve discovered thing called random experience sampling and trying it out in addition to my weekly extensive questionnaires. CSV export
PhoneTrack. Tracks location to Nextcloud which is accesable for Nextcloud database
SmarterTime. Tracks phone apps activity. CSV export
Interval Timer. For morning HRV test
Telegram Bot. I send messages to bot which being imported into database as events with timestamps. Also all my questionnaires being answered in bot chat, telegram has nice UI with buttons for different aswers. Answers saved to database.