Every fighter has the same problem during taper week: you feel like you should be doing more. You've spent weeks building fitness, the competition is right there, and sitting on the couch feels wrong. Your coach says rest. Your brain says one more hard session.
I went with my brain. The data had something to say about that.
The Setup
Some context: I'd just come back from a 9-day illness layoff, 7 days out from the WKU English Championships. Fight Coach AI had tracked the whole illness arc (I wrote about that here) and built a careful return-to-training plan. Short reactivation sessions in Zone 4, testing the engine without redlining it. The plan was working:
| Date | Duration | Avg HR | Max HR | Zone 5 | Zone 4 | Purpose |
|---|---|---|---|---|---|---|
| Feb 15 | 40 min | 158 | 194 | 17.4% | 41.6% | Reactivation |
| Feb 16 | 27 min | 169 | 193 | 35.6% | 35.0% | Sharpening |
| Feb 17 | 23 min | 149 | 190 | 10.7% | 31.2% | Light technical |
The intensity was stepping down. Sessions were getting shorter. The max HR proved the top end was still accessible post-illness (194 on the first day back). Sleep was recovering. By the 17th, the coaching briefing was clear: the hard work is done, now we taper. Wednesday should be light technical work only. Thursday and Friday, rest.
The Deviation
On Thursday the 19th — three days before competition — I went to the gym and hit the bag. Hard.
| Date | Duration | Avg HR | Max HR | Zone 5 | Zone 4 |
|---|---|---|---|---|---|
| Feb 19 | 47 min | 165 | 201 | 32.2% | 39.6% |
That's a 201 max heart rate — the highest in my last 30 days of training. A third of the session spent in Zone 5. Nearly 50 minutes of work when I should have been resting. It wasn't the plan. It wasn't smart. But it felt necessary in the moment, which is exactly the problem.
The AI coach didn't sugarcoat it:
"You deviated from the taper plan. Today's session was significantly harder than Wednesday's agreed light technical work. You hit 201 max HR — your highest in the last 30 days — across 11 intervals with inconsistent recovery dips (108–173bpm shows poor regulation). This is the opposite of what we discussed: arrive rested, not accumulated fatigue."
It went further, connecting the session to my broader recovery state:
"The concerning part isn't the intensity itself; it's the timing and the pattern. You're 3 days out. Your HRV is still suppressed at 59ms, RHR remains elevated at 50, and SpO2 lows are still dipping into the 80s. Your body hasn't fully cleared the post-illness markers yet. A maximal session today burns recovery capacity you need for Sunday."
This is the part a human coach might not catch. A coach who sees you once or twice a week might notice you went hard and say "probably shouldn't have done that." Fight Coach AI knew my exact HRV trajectory, knew my SpO2 had been dipping, knew my recovery heart rates between intervals were erratic (108 to 173bpm — that range is a red flag), and could calculate exactly how much recovery capacity I'd just burned three days out from a fight.
The Cost in the Data
The sleep data that night told the story:
| Night | Sleep Score | Deep Sleep | HRV | Resting HR | Body Battery |
|---|---|---|---|---|---|
| Feb 19 (before session) | 88 | 73 min | 59 | 50 | +66 |
| Feb 20 (after session) | 69 | 43 min | 47 | 54 | +40 |
The night before my hard session, I had my best sleep of the week — 88 score, 73 minutes of deep sleep, HRV at 59. I went into that session in the best recovery state I'd had since returning from illness. And I spent it.
The night after: sleep score crashed 19 points. Deep sleep dropped by 30 minutes. HRV fell from 59 to 47. Resting HR climbed 4bpm. Body battery recovery halved.
I'd taken my best recovery day and converted it into my worst.
The Adaptation
Here's what separates a data-driven coach from a training plan stuck to the fridge. The plan said rest Thursday and Friday. I broke that on Thursday. A static plan would just say "rest Friday" and move on. Fight Coach AI adapted to what actually happened:
"Tomorrow: complete rest. No training. No 'light pads.' No temptation. You've done enough work this week. Friday same — full rest. Your job between now and Sunday is sleep, hydration, and showing up less fatigued than you are right now."
Then it set a specific monitoring target:
"Check your HRV and RHR when you wake Saturday morning. If HRV is trending back up toward 62–63 and RHR dropping toward 49, you're on track. If both are flat or worse, that tells us something about your readiness that matters more than any pad work."
This is what persistent context gives you. The coach remembered what my baseline HRV should be (67–69), knew where it currently was (47 after the hard session), and set a realistic recovery target for 48 hours out (62–63). It wasn't guessing — it was projecting from weeks of data.
Two full rest days later, the morning-of numbers were encouraging:
| Night | Sleep Score | HRV | Resting HR | Body Battery |
|---|---|---|---|---|
| Feb 21 (Sat) | 80 | 56 | 51 | +75 |
| Feb 22 (Competition morning) | 83 | 57 | 49 | +68 |
HRV climbed back from 47 to 57. Not my baseline, but enough. Resting HR dropped back to 49. Body battery was the highest it had been in days. The Saturday coaching briefing reflected the data:
"You're in good shape to compete tomorrow. Thursday's session proved the gas tank is there, even coming off illness. Two full rest days means you'll show up tomorrow rested enough to compete safely, even if you're not at peak condition."
Not peak condition. But ready. That distinction matters when you're 41 and coming off illness two weeks before a fight.
Competition Day
The WKU session data told its own story:
| Date | Duration | Avg HR | Max HR | Zone 5 | Zone 4 | Zone 3 |
|---|---|---|---|---|---|---|
| Feb 22 | 47 min | 147 | 195 | 5.2% | 18.8% | 39.7% |
Only 5% in Zone 5, compared to 32% in Thursday's rogue session. Much more time in Zone 3. Lower average HR. This wasn't a redline effort — it was a paced, controlled competition performance. The top end was there when I needed it (195 max), but I wasn't burning through it recklessly like I had on Thursday.
What I Actually Learned
The Thursday session wasn't a disaster. The AI coach even acknowledged it afterward — it proved my conditioning was still there post-illness, which removed the doubt going into Sunday. But the cost of removing that doubt was measurable: 19 points off my sleep score, HRV crashed, two days of recovery burned that I couldn't afford.
The right move would have been to trust the data from earlier in the week. The Feb 15 reactivation session had already shown I could hit 194 max HR. That was the proof. I didn't need to prove it again at 201 three days before a fight.
This is the hardest part of data-driven training: the data will tell you what you need to know, but you have to actually listen. The AI coach gave me the right plan. I overrode it because taper anxiety is real and rest days feel wrong when a fight is coming.
Next time — and there is a next time, ICO is 13 days away — I'll trust the taper. The data will hold me to it.