In my original sleep analysis, I highlighted my REM sleep and sleep efficiency as two areas to improve. Unfortunately, when it comes to improving either, an internet search returns a deluge of articles with generic advice and no data backing up their claims.
To try to get to an answer of what actually works, I’ll run a handful of sleep experiments over the coming months. I decided to start with cold showers because (a) I like them (b) they’re easy and cheap and (c) there is some evidence that they have health benefits.
It turns out that quarantine is the best time we’ll ever get to conduct self-experiments because almost every day is the same, limiting the number of variables between the control and intervention periods.
The Experiment
My experiment was dead simple: 3 weeks of normal sleep, followed by 3 weeks with a cold shower every night within 90 minutes of bedtime. The showers, which were as cold as my shower would go, lasted between ~3 to ~10 minutes, but I didn’t time them every night, and didn’t use the duration in this analysis.
I tracked my sleep with the Oura ring, which you can read more about in my last sleep post. I’m a fan of the ring because it’s non-invasive, holds a decent charge, and provides quick feedback on the main things I care about.
One night I forgot to take a shower, and another I forgot to charge my Oura, so I ended up with 19 nights of sleep data.
As far as I could tell, I had similar exercise, diet, stress, caffeine intake, and schedule during the Cold Shower and non-Cold Shower periods, but didn’t track most of this.
Cold Showers
Representative of the first 15 seconds in a Cold Shower
There’s a masochistic element to cold showers. The first 15 seconds suck, but they’re unbelievably invigorating and have an immediate positive impact on my mood, at any time of day.
There is some science behind cold showers. If you’re interested in the details, this is a good meta-analysis of various studies. Among the potential effects: short-term increases to noradrenaline and dopamine, increased well-being and decreased anxiety, antidepressant effects, and immunity boosting.
Whether any or all of these purported benefits are true, it’s hard to deny the subjectively great feeling when hopping out of a cold shower or plunge.
Results
My total sleep was impacted the most, up 30 minutes during the Cold Shower experiment. My Deep Sleep remained unchanged, while I saw a 13% boost to REM Sleep for an extra 10 minutes of REM sleep a night. While my REM Sleep:Total Sleep ratio was largely unchanged (18% → 19%), I’m happy about any increase in overall REM.
Sleep Efficiency
In addition to REM, I’m most interested in increasing my sleep efficiency so I can spend the same amount of time in bed, while getting more sleep. I was happy to see a bump from 90% to ~93% during this experiment.
Recovery
I’ve spent less time focused on recovery metrics like Resting Heart Rate and HRV, but it was encouraging to see the positive impact cold showers had on both of these metrics. My RHR was almost a beat and a half lower, and my HRV increased about 10%.
Conclusion
I didn’t track my energy levels or mood, but my subjective review is that my “top end” energy was about the same, with a more consistent, steady energy throughout the day. Those mid-day energy troughs were eliminated, which was a big win.
This is decidedly not evidence that cold showers are a sleep panacea. I’m always wary of n=1 experiments, but the results were uniformly positive, so this is something I’ll definitely continue.
As for the mechanism of action, my hunch is that there are equal parts placebo, and a stress-reducing, mood-enhancing effect, leading to better sleep.
I highly recommend people give this a shot for at least a few consecutive days. Once you get past the initial burn, you’ll settle into what soon becomes an unexpected endorphin high that lasts longer than you’d expect.
With this in the books, my next experiment will be not eating for at least 2 hours before bed. This is a commonly espoused sleep tactic, so we’ll see what the impact is. I’ve noticed that very large meals before bed increase my heart rate, so I’d expect that to come down, but I’m not sure how or if this will affect other metrics. Stay tuned.
In Part 1, I outlined why our sleep environment is so important and what I’ve done to optimize it, explained why I track my sleep, and analyzed my sleep trends over the last year and a half, as measured by the Oura ring.
In Part 2, I’ll analyze my sleep data from the last year, which has been particularly interesting because I’ve had such a differing sleep schedules and a lot of travel.
I work as an analyst for the San Francisco Giants, and for 6 weeks during Spring Training in Arizona, I’m on a very early-to-bed, early-to-rise schedule. Once the 6-month season starts, the majority of games finish after 10pm, which leads to a much later schedule. This is followed by a 4 month off-season where I’m able to shift back to my natural circadian rhythm.
These graphs show the drastic changes in sleep schedule between the season (April-September) and the offseason.
The later bedtime and wake times in-season are in stark contrast with Spring Training (Feb-March) and my natural circadian rhythm, reflected in March-May data, when Covid-19 sent everyone home from Spring Training.
The most drastic sleep changes come in Spring Training, when overall sleep and REM sleep plummet, but deep sleep actually increases. My guess as to why deep sleep increases is that the days are long and intense, and your body tries to recover as much as possible.
Travel
While people with a later circadian rhythm don’t need to adjust their sleep schedules as much during the season, everyone deals with frequent travel and inconsistent sleep hours, which isn’t ideal for optimized sleep.
One downside West Coast teams have is increased travel. The Giants had the 5th most travel of any team in baseball last year, with 41,934 miles covered during the season.
Two scenarios are particularly tough on sleep quality. First, when the last game of a series is a night game. On a night getaway game at Dodger Stadium, by the time we eat, leave the clubhouse, drive to LAX, fly to SFO, take the bus to Oracle park, and finally go home, it’s around 3am.
The other scenario that impacted sleep the most were any East Coast trips, which usually got us to the hotel around 3am-4am. All of those trips were followed by an off-day.
Luckily those scenarios didn’t happen often–we had 4 night getaway days, and 4 East Coast trips. Unfortunately I’m missing sleep data from a lot of these nights.
The last collective bargaining agreement smartly mandated more day games on travel days, limiting the nights teams would arrive to a new city at 3am and play the next day.
As you’d expect, day games were incredibly helpful for getting more sleep that night–I averaged a full extra hour of sleep on nights after a day game.
Conversely, on nights before day games, I had significantly lower sleep quality. Not surprising, given the quick turnaround of a night game, and an early morning the following day.
For sleep quality, nothing beat home. On nights I slept in my own bed, I had more total sleep, more REM sleep, and more deep sleep. Given the sleep environment I’ve set up there, this isn’t surprising. One thing to note: I tended to go to bed earlier at home, which likely impacted these numbers, independent of the environment.
When looking at road games by time zone, it isn’t surprising to see better sleep in cities on PDT compared to CDT or EDT. Much has been written on the negative effects of time zone changes. The shorter travel definitely helps, and is reflected in all metrics.
It’s also interesting to see how bad my sleep was in Denver–of all the cities we played in, it was where I had by far the worst sleep.
Subjectively, I always felt like I slept poorly there, likely because of the altitude and dry air. It was also one of the hotels that didn’t get particularly cold AC, so I ended up bringing a fan for our last trip there.
—
What I Learned
Having gone through this exercise, I have a few ideas of what I’d do differently in Year 2 to manage the quirks of an MLB season, and keep my sleep quality as high as possible.
I’d take my sleep environment on the road more seriously. I remember two nights last year where I didn’t turn the AC up to full blast until I was ready to go to bed, only to find it wasn’t functioning properly. They got fixed the next day, but led to a few nights of poorer sleep. Moving forward, I’ll crank the AC up immediately when I arrive to confirm it works, and allow time for fixing if not. I’ll also travel with a fan for scenarios where the AC just doesn’t get that cold, or doesn’t make enough white noise.
Especially on East Coast trips, I’d get more sunlight early in the day, and possibly travel with a full spectrum light (I have one, but haven’t really used it yet). Sunlight helps regulate our circadian rhythm, so instead of staying in the hotel the morning after arriving in an East Coast city, I’d get outside and get some sun on my face and let my body know it’s time to go.
I’d take travel naps more seriously. Knowing my sleep on the road is going to be impacted, I’d make better use of buses and plane rides to sleep, which I didn’t do often enough.
I’d supplement more aggressively on days I know it will be harder to get good sleep. I’m cautious of taking melatonin or other sleep-inducing compounds, but I think they could be a very beneficial aid during nights when I know it’ll be harder to sleep well, like an East Coast trip, or night getaway day.
I’d take more saunas. I haven’t logged the data well, but anecdotally, I sleep better on nights I take a sauna, which I did for a lot of the offseason, and part of Spring Training.
Experiments to Run
I haven’t formally experimented enough to see how different interventions impact my sleep. I’ve relied more on anecdotal observations, but I’d like to run the following experiments and see what, if anything, happens:
Cold showers. I take cold showers often, but at random intervals, and have never tracked them.
Full-spectrum light in the morning. I’m up every morning before the sun rises, so I wonder if getting exposed to this light early in the morning will impact sleep.
Monitor drinking. Anecdotally, 2 drinks or less has little to no impact on sleep, but 3 does. I’d like to test this.
Float tank. I floated 3 times last year, and felt great after each session. I can’t remember how I slept, but I’m interested to see if there’s an impact.
Epsom salt baths. Magnesium, one of the minerals in epsom salt, is a well-documented sleep aid (and a hypothesis for why the float tank could impact sleep).
Limit screens before bed. I’m good, not great at this. I wonder what would happen if I extended this to 60 minutes before bed.
Don’t eat for at least 2 hours before bed. A late meal skyrockets my resting heart rate, but I’ve never tested this over a long period.
Hopefully I’ll have something interesting to report back.
A paradox of the modern world is that we undervalue and deprioritize things we know to make us healthier, happier and more productive. Sleep is chief among them.
The research is unambiguous: Good Sleep is a powerful performance enhancer. With Good Sleep, we’re stronger, we’re smarter, and we look better. Everything we want to accomplish–physical goals, cognitive goals, relationship goals–becomes easier.
Going through the world with Bad Sleep is running a marathon with a weight vest on. Yet not sleeping has become a badge of honor. Imagine athletes bragging about their poor preparation and lack of training. Would we ever imagine they were in the best position to perform well? The same goes for sleep, for athletes and non-athletes alike.
The fact that we’re unconscious for a third of our lives provides a hint to the importance of sleep. It’s hard to think of a worse evolutionary disadvantage than conking out for a third of each day, prone to attack, and unable to find food. In spite of that cost, humans evolved to require a huge amount of sleep, something we fail to appreciate as a society.
Sleep Environment
Environment plays a prime role in sleep quality. You wouldn’t try to study at a construction site, and you wouldn’t try to PR your 10K in 10 degree weather, so why sleep in an environment not conducive to optimal sleep?
I can only speak for myself, but I think the optimal sleep environment differs for everyone. Over the years, I’ve experimented with different set-ups and gadgets, and have landed on a core “Sleep Stack” to create the optimal sleeping environment. I sleep best in a room that’s very cold, pitch black, with white noise.
Cold
If I could choose one characteristic of an optimal sleep environment, it would be temperature somewhere between 62-65 degrees. Sleeping in a warm room is a slow drip IV of discomfort, so I avoid it at all costs.
Central AC is rare in San Francisco, so the tools of choice are a bedroom AC unit and Chili Pad. The Chili Pad goes under the sheet, and circulates cold water, keeping your sleep surface as icy as you want. The first night you sleep with the Chili Pad, you wonder how you ever slept without it.
Dark
Our circadian rhythms follow the pattern of the sun. We’re awake when it’s light, we sleep when it’s dark. I’ve found that any light shining through in the morning impacts my ability to sleep optimally.
I’ve tried numerous black out curtains and have slept in loads of hotels with heavy-duty curtains, but nothing has blocked light as well (and at as low of a cost) as overlapping black towels.
Let There Be No Light
If you can get past your significant other tolerating the eye-sore, this is highly recommended.
When I’m on the road and don’t have black towels blocking all the light, I use this sleep mask, which works well.
White Noise
For some reason, sleeping in dead silence is uncomfortable and my mind seems to wander more. To create some white noise, I blast a fan, with the air pointing in the opposite direction.
Blanket
My most recent acquisition is the Gravity Blanket, a 35 lb blanket that’s insanely cozy and meant to make you feel peaceful and comforted. I’ve only had it for a few weeks, but I’ve enjoyed it so far.
Sleep Tracking
I’ve been tracking my sleep on and off since 2013, originally with the (now defunct) Zeo and for the last year and a half, with the Oura ring.
There are a lot of devices that track sleep, but I enjoy the Oura ring because I don’t love wearing watches (to be fair, I don’t love wearing rings either, but this is the lesser of two evils). I’m less interested in non-wearable sleep trackers, because I think heart rate and heart rate variability are really important metrics to track, which non-wearables don’t get, or don’t get as accurately.
Tracking sleep nightly has several benefits for me. First, it’s a forcing function to do the things that lead to better sleep, knowing I’ll see the results in the morning. Second, it builds up a dataset that’s useful in seeing trends over longer time-horizons.
ROI
These things aren’t cheap. The combined cost of these items is over $1,000, but I look at them as investments, not expenses. If you could guarantee me a 10% increase in the length and quality of my sleep for the next year, the return on investment in the form of increased health, happiness, and productivity would be higher than any other investment I could make.
Analysis
I’ve collected 464 nights of sleep data from the Oura ring. The nights I’ve missed have been a combination of either forgetting to put it on, or a depleted battery. The missed nights seem to be distributed randomly, and shouldn’t skew the dataset.
Oura records a lot of metrics (full list here), but the ones I track most closely are total sleep, deep sleep, REM sleep, sleep efficiency, bedtime, heart rate, heart variability, and respiratory rate (breaths per minute).
Over the last year and a half, these are my nightly averages:
Takeaway 1: Get more overall REM sleep, and increase my REM:Total Sleep ratio
My REM:Total Sleep ratio is currently 19%. Studies suggest 20%-25% to be optimal, so getting an extra 25 minutes of REM sleep a night would get me to a 25% REM:Total ratio, assuming the same amount of total sleep. More on this below.
Takeaway 2: I can get more total sleep without spending more time in bed
Increasing sleep efficiency (total sleep / total time in bed) from 88% to 93% would give me an extra “free” 20 minutes of sleep a night. That’s an extra 2h 20m a week, or 121h 40m in a year. That’s 5 full days worth of sleep a year by tossing and turning less, and spending 0 additional minutes in bed. Seems worth pursuing.
Correlations
I’ll start by looking at correlations across the entire dataset. The chart below highlights the correlation coefficients between different metrics. The size and intensity of the colors reflect the intensity of the correlation. Blue marks high positive correlation, where an increase in one metric is correlated with an increase in the other; red marks high negative correlation, where an increase in one metric is correlated with a decrease in the other.
Obligatory correlation ≠ causation note here. A lot of these correlations are simply results of other behaviors. For example, take the correlation between Bedtime and Heart Rate. I don’t believe going to bed later inherently increases my heart rate. But on nights when I go out and have drinks, I tend to go to bed later, and the drinking increases heart rate.
Takeaway 1: REM sleep is highly correlated with Total Sleep
Shorter nights of sleep disproportionately impact REM. This is an important point. Because more REM occurs later in the sleep cycle, a night with 80% of normal sleep doesn’t impact all phases of sleep by 80%. It disproportionately crushes REM, which impacts our ability to concentrate, skill acquisition, problem-solving, and other things most of us need to function in the modern world.
Takeaway 2: Total Sleep is highly negatively correlated with Bedtime
This confirms what I’ve always felt: if I go to bed later than usual, I still wake up around the same time and have less overall sleep. Because my circadian rhythm is so early, I try hard to go to bed early every night, knowing I won’t get a full night of sleep otherwise. In my ideal world, my sleep schedule would be something like 9pm-5am. Early mornings with no one awake yet are special times, and when I do my best thinking.
Takeaway 3: Heart Rate Variability is highly negatively correlated with Heart Rate
You can read more about the importance of Heart Rate Variability here, but it serves as a great proxy for stress levels, recovery, and readiness to perform. I don’t think Heart Rate directly impacts Heart Rate Variability, rather the things that lead to an increased heart rate–drinking, late workouts, stress–can crush Heart Rate Variability and our ability to perform at peak performance.
Takeaway 4: REM is positively correlated with Breaths Per Minute, and negatively correlated with HRV
I don’t know what to make of this. An increase in BPM is negatively correlated with sleep efficiency, deep sleep, and heart rate–all bad outcomes. Not surprising, given that the things that reliably make my BPM increase are drinking and eating a big meal close to bedtime.
So why the positive correlation with REM? Could it be that a late meal increases my BPM, but is stress-reducing which lets me get more REM?
I’m also not sure why REM negatively correlates with HRV. I’d assume the things that lead to high HRV–good workouts, good sleep, lower stress–would also increase REM. Something to explore more.
I’ll stop here for now, but in Part 2, I’ll dive into how the schedule and travel of an MLB season affected my sleep, what I learned from the data, what I’d do differently, and what sleep experiments I’ll be running this year.
