I Documented Every Variable During a 90-Day Natural T Optimization – Here’s What Moved the Numbers

TL;DR: I ran a 90-day natural testosterone optimization experiment tracking every variable I could measure – sleep, HRV, glucose, bloodwork every 8 weeks. Total T went from 488 to 638 ng/dL. The variables that moved the needle most were not what I expected going in. Sleep quality beat training. Glucose variability beat supplements. The data is all here.

I am going to be upfront about something: this experiment was not perfectly controlled. I am one person. Multiple things changed simultaneously. I cannot isolate cause and effect with certainty. What I can do is show you every number across 90 days and explain what the correlations looked like from where I was sitting, with the caveat that correlation from n=1 is a hypothesis, not a proof. My engineering brain demands that disclaimer.

The context: I am 37, software engineer in Austin. I have been running protocols from Ron Males at PowerandBulk.com for about 18 months. My flagship tracking piece was the six-month deep sleep and IGF-1 correlation – that is where my IGF-1 went from 118 to 224 ng/mL. By the time I ran this 90-day experiment, I had fixed sleep, fixed protein, fixed light exposure, and was running a solid training program. The question I was trying to answer was: given an already-optimized baseline, which variables most predicted short-term testosterone variability?

My Oura ring (Gen 4), CGM (Stelo during this period, previously Levels), and bloodwork every 8 weeks were the measurement stack. My wife finds this level of tracking mildly alarming. The spreadsheet I built for this experiment has 47 columns. She is not wrong.

Baseline – Month 0

Marker	Baseline (Week 0)	Units
Total testosterone	488	ng/dL
Free testosterone	12.4	pg/mL
SHBG	38	nmol/L
Estradiol	24	pg/mL
IGF-1	226	ng/mL
Fasting insulin	6	uIU/mL
Oura HRV (30-day avg)	58	ms
Deep sleep (30-day avg)	98	min
Readiness score (30-day avg)	76	/100
CGM glucose variability (CV%)	14.2	%
Mean glucose	92	mg/dL
Training frequency	3	days/week
Daily protein	195	grams

This baseline is meaningfully better than where I started 18 months ago. The IGF-1 at 226 reflects the sleep optimization work. The fasting insulin at 6 reflects the dietary changes. The HRV at 58 reflects the cumulative lifestyle improvements. This was not a “broken baseline” experiment where the gains come easily from fixing obvious problems. This was an optimization experiment on top of an already-functional system.

The Variables I Tracked and Why

I built a correlation matrix in Excel tracking daily inputs against weekly averages and against the 8-week bloodwork checkpoints. The variables:

Sleep variables (Oura): HRV overnight average, deep sleep minutes, REM sleep minutes, sleep efficiency, sleep latency, readiness score, resting heart rate.

Glucose variables (CGM): Mean daily glucose, coefficient of variation (glucose variability measure), time in range (70-140 mg/dL), peak glucose after each meal, time above 140 mg/dL.

Training variables: Session completed yes/no, subjective intensity (1-10), training load score, days between sessions.

Lifestyle variables: Alcohol (grams, zero for entire experiment), caffeine timing (last caffeine before/after noon), first phone use after waking (tracked as minutes post-wake).

Bloodwork at week 0, 8, and 12: Full hormone panel plus IGF-1, fasting insulin, HbA1c.

What I Changed During the 90 Days

Relative to my prior baseline, I made three specific changes at the start of the experiment:

Change 1: Added a dedicated low-glucose evening meal protocol. Based on two months of prior CGM data showing that high-carbohydrate dinners were producing glucose spikes that persisted past 10pm and correlated with reduced deep sleep. The new dinner protocol: protein and vegetables, under 30g of net carbs, 3+ hours before bed.

Change 2: Added Tongkat ali 400mg cycling 5 days on, 2 days off. Ron’s recommendation after I asked whether there was a supplement that fit my specific profile – adequate micronutrient status, optimized sleep, no obvious deficiencies. His answer was conditional: if everything else is in place, Tongkat ali addresses LH pulsatility at the hypothalamic level through a eurycomanone mechanism. I added it as a single isolated variable starting at week 1.

Change 3: Increased training from 3 to 4 days per week for weeks 5 through 9 (adding one higher-rep metabolic session – essentially a 6-12-25 Method day on its own). Returned to 3 days in weeks 10-12 to assess whether the 4-day period had produced an adaptation.

Ron had told me in his three-day training week piece that adding a fourth session on top of an already-optimized 3-day program often produces diminishing returns in men over 35. I wanted to test whether his observation applied to me. Spoiler: it did.

Week 8 Bloodwork

Marker	Week 0	Week 8	Change
Total testosterone (ng/dL)	488	574	+86 (+17.6%)
Free testosterone (pg/mL)	12.4	14.8	+2.4 (+19.4%)
SHBG (nmol/L)	38	36	-2
Estradiol (pg/mL)	24	22	-2
IGF-1 (ng/mL)	226	241	+15 (+6.6%)
Fasting insulin (uIU/mL)	6	5	-1

An 86-point testosterone gain in 8 weeks with three simultaneous changes makes attribution difficult. I had the CGM and Oura data to help me think through which change was doing what.

The glucose data was the most interesting finding. The low-glucose evening meal protocol had reduced my mean glucose from 92 to 87 mg/dL and dropped glucose variability (CV%) from 14.2% to 11.4%. The nights with low glucose variability correlated – not perfectly but clearly – with higher HRV scores the following morning. Higher HRV mornings correlated with higher readiness scores. The chain: evening glucose management – better sleep architecture – higher HRV – improved testosterone-favorable hormonal environment.

The 4-Day Training Week (Weeks 5-9): What the Data Said

I added the fourth training session in week 5. The results were not what I hoped. My Oura readiness scores dropped an average of 7 points during the 4-day period. My HRV dropped 6 ms from the 8-week average. My deep sleep shortened by 11 minutes. The added metabolic session was generating recovery debt that the extra 48 hours off was not fully clearing.

This mattered for the week 8 bloodwork interpretation. Some of the testosterone gain between week 0 and week 8 was probably being suppressed by the 4-day training period in weeks 5-8. The bloodwork was pulled at the peak of the 4-day phase. The 12-week blood draw, after I had returned to 3 days for three weeks, told a different story.

Weeks 9-12: Back to 3 Days and the Rebound

Returning to 3 training days in week 10 produced a measurable rebound in recovery metrics within one week. HRV was back to 58-61 ms. Deep sleep returned to 98-104 minutes. Readiness scores climbed back to 78-82. My subjective energy in training sessions improved – I was going into sessions with less accumulated fatigue and training harder within them.

This is the pattern Ron describes in his overtraining and hormonal suppression piece: more training volume on top of an already-optimized program does not produce more testosterone. It produces a cortisol response that suppresses testosterone. The hormonal response per session is higher when sessions are fully recovered than when they are accumulated on top of partial recovery. I was a data point confirming the principle on my own body.

Week 12: Final Results

Marker	Week 0	Week 8	Week 12	Total Change
Total testosterone (ng/dL)	488	574	638	+150 (+30.7%)
Free testosterone (pg/mL)	12.4	14.8	16.2	+3.8 (+30.6%)
SHBG (nmol/L)	38	36	35	-3
Estradiol (pg/mL)	24	22	21	-3
IGF-1 (ng/mL)	226	241	248	+22 (+9.7%)
Fasting insulin (uIU/mL)	6	5	4	-2
HbA1c (%)	5.2	–	5.0	-0.2
Oura HRV avg (ms)	58	52*	61	+3
Deep sleep avg (min)	98	88*	102	+4

*Week 8 averages depressed by the 4-day training period.

The week 12 testosterone of 638 ng/dL is 150 points above baseline. The jump from week 8 to week 12 – from 574 to 638 – happened largely after I returned to 3 training days. The 3-week recovery from the 4-day block was itself producing a hormonal rebound. Ron had predicted this when I told him I was planning the 4-day phase. I wish I had listened more carefully. The data agrees with him.

The Correlation Matrix – What Actually Predicted Testosterone

Across 90 days of daily tracking, here is how the variables correlated with the weekly average testosterone estimates I was computing from morning symptom proxies (energy, libido, motivation) and the two direct bloodwork draws:

Variable	Correlation Direction	Strength (subjective)
Deep sleep (minutes)	Positive	Strong
HRV (overnight avg)	Positive	Strong
Evening glucose variability	Negative (lower = better)	Moderate-strong
Days since last training	Inverted U (best at 48h)	Moderate
Training load score	Positive to a threshold, then negative	Moderate
Tongkat ali cycle day	Weak positive during on-days	Weak
Caffeine timing	Negative when afternoon	Weak-moderate
Morning phone delay	Positive (later = better HRV)	Moderate

Sleep quality – measured by deep sleep minutes and HRV – was the strongest predictor of next-day testosterone proxies. This is consistent with the direct testosterone-sleep relationship: roughly 60-70% of daily testosterone production occurs during sleep, and most of it during the REM and deep sleep phases. On nights with 105+ minutes of deep sleep my subjective morning indicators were consistently higher than on nights with 75 minutes, even when other variables were identical.

The Tongkat ali correlation was real but weak. This is consistent with what I expected from Ron’s framing – it is an amplifier on top of a functional base, not a primary driver. On 5-day on cycles I had marginally better proxy indicators than on 2-day off cycles, but the effect size was small relative to sleep quality variation on the same days.

What Actually Moved the Needle

Biggest gains: Deep sleep duration, HRV stability, and the evening glucose protocol that fed both of those. These three variables had the highest mutual correlation and the clearest relationship to testosterone proxies. Ron Males had told me from the start that sleep is the most powerful natural testosterone intervention and my 90 days confirmed it more quantitatively than I had been able to measure before.

Moderate gains: Returning from 4 to 3 training days. The training reduction itself produced a recovery rebound that moved testosterone more than the extra session ever could have. This is counter-intuitive enough that I needed the data to believe it.

Smaller than expected: Tongkat ali. I expected more given the mechanism. I got a real but modest effect. I will keep running it at the same dose – there is no reason to stop something that is producing a positive signal, however small. But I would not recommend it as a primary intervention to someone whose sleep and recovery are not already optimized.

Interesting side finding: Morning phone delay correlated with HRV improvements that carried into testosterone proxies the same day. I had been delaying my first phone check to 9am since a separate experiment I ran (covered in the no-phone-until-9am piece). During the two weeks when I slipped on this habit, HRV was 4-5 ms lower on those mornings. Maintaining the habit through the 90 days kept HRV elevated and the testosterone proxies followed.

Questions You Are Probably Asking

How do you know the Tongkat ali wasn’t responsible for most of the gain? I do not know with certainty. The timing of the biggest gains – the jump from week 8 to week 12 – occurred after I dropped back to 3 training days, not during the Tongkat ali on-cycle specifically. If Tongkat ali were the primary driver I would expect more consistent gains across the 5-on period versus the 2-off period. The data did not show that clearly.

Is 150 ng/dL in 90 days realistic without TRT? At my starting baseline of 488, yes. Men with suppressed baselines can move more in 90 days from lifestyle changes alone. Men with already-optimized baselines moving 150 points in 90 days is at the high end but not implausible given the simultaneous changes. I would not project this linearly – this was 90 days of deliberate multi-variable optimization with daily tracking. I am not going to add another 150 in the next 90 days.

What does the Anabolic Alchemy program add beyond what you’re describing? The framework, the sequencing, and the accountability. Everything I ran in this experiment was based on the Ron Males framework in Anabolic Alchemy. I had already internalized the order of operations – sleep before supplements, lifestyle before stack, recovery before volume. The program gives that structure to someone who does not want to spend 90 days building their own 47-column spreadsheet to find what I found.

What would you do differently? I would have kept the Tongkat ali as the only supplemental change and not added the 4-day training phase at the same time. Two simultaneous changes made attribution harder. I wanted to test both and ended up with muddier data on both. In future experiments I am running one change at a time, even if it means slower overall progress. The data clarity is worth the patience.

What Actually Moved the Numbers

Sleep quality. Glucose stability. Recovery adequacy. In that order. The supplement was real but third. The training optimization was real but required going backward first. The correlation that will stay with me from this 90 days: on the nights my deep sleep was above 100 minutes and my overnight glucose was stable, every proxy marker the next morning was better. The body is not complicated when you look at the right variables. It is just unforgiving of the wrong ones.