The Future of AI and Training: From Data to Personalized Plans—A Coach Breaks Down What AI Coaching Platforms Can Really Do

Coach’s Opening: The Day the Algorithm Told My Athlete to “Push More,” I Hit Pause

Let me start with a scenario so real it couldn’t be more real. Last year, one of my athletes, Xiao Chen, was preparing for Challenge Taiwan. He’d been training with a well-known AI coaching app for three months, and his numbers looked incredible—the fitness curve kept climbing, and the app cheerfully told him every day, “You’re in great shape today, recommended: threshold intervals.” Then, six weeks before race day, on what was supposed to be an easy ride, his heart rate ran a full 15 bpm higher than usual, while his pace dropped significantly. The app didn’t notice a thing. It just followed the plan and scheduled another high-intensity workout for the next day.

I glanced at his sleep log, asked a couple of questions about how much overtime he’d been pulling at work, and felt out his resting heart rate from that morning. Right then and there, I scrapped the entire workout the app had scheduled, replacing it with a full rest day plus one very easy active recovery session. Three days later, his numbers came back on their own. If we’d pushed through that week following the algorithm, I’m almost certain he would have walked straight into overtraining.

I’ve coached triathletes for 15 years—from helping people finish their first 51.5 to sending athletes to qualify for the Kona World Championship. Through it all, the advances in AI and wearable tech have genuinely made me both excited and wary. What I love is that it’s put data analysis, which used to be something only pro teams could afford, into the watch of every age-group athlete. What I’m wary of is that too many people treat “the app says” as gospel, forgetting that data is just a shadow of the body, not the body itself.

In this article, I want to use a coach’s perspective to walk you through AI coaching platforms from start to finish: how they actually work, what machine learning can really do for training plans, and what it will never learn to do—the things only a human coach can fill in. By the end, I hope you won’t be more blindly trusting of AI, but better at using it.

Foundational Concepts: What Is an AI Coaching Platform Actually Calculating?

To understand AI coaching, you first have to understand the raw materials it consumes. At their core, all these platforms do the same thing—quantify your training into a few key metrics, use those metrics to estimate your fatigue and fitness levels, and then work backward to decide what you should do tomorrow.

Three Underlying Metrics You Absolutely Need to Know

The first is FTP (Functional Threshold Power). It’s defined as “the highest power a rider can maintain in a quasi-steady state without fatiguing for approximately one hour,” measured in watts (W). Since actually doing a full one-hour all-out test is brutally painful, in practice it’s usually estimated by taking the average power from a 20-minute all-out test and multiplying by 0.95. FTP matters because it sits right around your lactate threshold and serves as the baseline for dividing up all your power zones.

The second is TSS (Training Stress Score). This is a concept introduced by TrainingPeaks: it uses “one hour at FTP intensity equals 100 points” as its baseline, while factoring in both “intensity” and “duration.” In other words, an easy two-hour ride and a hard one-hour ride might produce similar TSS scores, but their meaning for your body is completely different—and that’s one of the traps I’ll get to later.

The third is actually a set of interrelated concepts, and a table makes it clearest:

Metric	Full English Name	Plain-Language Explanation	Time Window
CTL	Chronic Training Load	“Fitness”—the foundation you’ve built up over the long term	Weighted average of TSS over roughly the past 42 days
ATL	Acute Training Load	“Fatigue”—the recent load you’ve accumulated but not yet shed	Weighted average of TSS over roughly the past 7 days
TSB	Training Stress Balance	“Form/Freshness”—= yesterday’s CTL minus yesterday’s ATL	Current moment

The relationship among these three is the core logic behind all AI training plans. CTL represents the fitness foundation you’ve stacked up over time, and it should climb slowly; ATL reflects the fatigue from the last few days, arriving fast and leaving fast; TSB is the difference between the two, telling you whether your body is in a “fresh” or “wrecked” state today. TSB is usually negative during training blocks (because you’re constantly accumulating fatigue) and only turns positive as you approach race day or during a taper—which is also what “tapering” looks like in the data.

The First Thing an AI Platform Does: Turn You Into a Curve

Every time you upload a ride, swim, or run, the platform converts it into TSS, feeds it into the CTL/ATL/TSB model, and plots your “fitness–fatigue–form” curve. That pretty upward line you see in the app is essentially CTL; that “today’s status” that changes daily is a variation of TSB.

Sounds very scientific, right? It is. But here’s the key—everything I’ve described above isn’t actually “AI” yet; it’s a sports physiology mathematical model that matured back in the 2000s (rooted in Banister’s impulse–response concept). The real machine learning sits on top of this layer, handling the “prediction” and “personalized recommendation” side. So let’s move into that layer now.

What Machine Learning Actually Does for Training Plans

A lot of people think an AI coach is just “a computer that builds your training plan,” but it’s not that simple. I’d break down the role machine learning plays here into three levels, from shallow to deep:

Level One: Pattern Recognition—Finding How “People Like You” Got Stronger

The most basic application is that the platform has training data from hundreds of thousands, even millions, of users. When a new user comes in, the algorithm compares: among the people with a similar FTP, similar body weight, similar weekly training hours, and similar target race—what combination of workouts did that group use to build up? This is essentially a recommendation system, the same logic streaming platforms use to recommend shows to you.

The value of this level is “a reasonable starting point.” It won’t hand a 9-to-5 worker with only 6 hours a week to train the same volume as a pro athlete, and it does that well. I often say that a complete beginner who knows nothing about training will usually get a far better starting plan from AI than from their own haphazard DIY approach—that’s the power of data; it helps you avoid the most obvious landmines. But keep in mind, “people like you” doesn’t mean “you.” That group includes people with iron stomachs, people who sleep perfectly, people who don’t have jobs. Their average doesn’t necessarily fit your real-life circumstances.

Level 2: Adaptive Adjustment — Dynamically Modifying Your Plan Based on Your Completion Rate

More advanced platforms incorporate “adaptive” features. They monitor whether you complete your workouts, the degree of power-to-heart-rate decoupling during completion, and whether your power drops in the latter part of intervals, to determine whether a given plan is too easy, just right, or too hard for you — then dynamically adjust the intensity of your next session.

Here’s a real example I observed. An athlete performed 5 sets of 4-minute VO2max intervals with a target power of 320 watts:

Set	Target Power	Actual Avg Power	Avg Heart Rate	AI Assessment
Set 1	320 W	322 W	168 bpm	Normal
Set 2	320 W	319 W	172 bpm	Normal
Set 3	320 W	315 W	176 bpm	Slightly Hard
Set 4	320 W	305 W	178 bpm	Significant Drop-off
Set 5	320 W	291 W	179 bpm	Exhausted

Seeing power collapse in sets 4 and 5 while heart rate hits the ceiling, a smart adaptive system would determine “this target is too high” and next time lower the target to 310 watts, or reduce the set count to 4. This is where machine learning truly starts to add value — it turns “whether you can actually handle the training” from a coach’s subjective judgment into a quantifiable, real-time feedback signal.

Level 3: Status Prediction — Attempting to Predict Your Fatigue and Performance

The most cutting-edge — and most easily overhyped — level is “prediction.” Some platforms claim to combine your HRV (heart rate variability), resting heart rate, sleep, and even body temperature to predict your “readiness” for the day, and further forecast your performance curve over the coming weeks.

I have to be honest about this level: the direction is right, but reliability varies greatly from person to person, and it’s particularly sensitive to noise in amateur athletes. Metrics like HRV are heavily influenced by sleep, alcohol, caffeine, psychological stress, and measurement posture. A father whose child cried three times last night will inevitably have poor HRV the next morning — but that has nothing to do with his training fatigue. The algorithm sees a number; it doesn’t see those three crying episodes.

This naturally leads to the next section — the pitfalls AI will inevitably stumble into, and which you need to learn to handle yourself.

Practical Approach: Treat AI as a Dashboard, Not Autopilot

Now that we’ve covered the theory, here’s what you can actually use. When I work with athletes, I position AI platforms as “an excellent dashboard and training log” rather than “autopilot.” Here’s my practical framework.

How I “Re-process” an AI-Generated Training Plan

Suppose a platform generates the following week for an advanced triathlete (a typical bike + run crossover week). Here’s the original version, and I’ve added a “Coach’s Notes” column so you can see the difference:

Day	AI Original Plan	Coach’s Notes & Adjustments
Mon	Rest	Keep, but change to 20 min active recovery + foam rolling
Tue	Bike 90 min incl. 3×10 min threshold	If slept less than 6 hours the night before, change threshold to 3×8 min
Wed	Run 60 min easy	Taiwan summer afternoons are too hot; move to early morning or indoor treadmill
Thu	Bike 2 hr endurance ride	Keep; add nutrition rehearsal (eat every 45 min)
Fri	Swim technique session	Keep; AI is weakest at assessing swimming, technique needs human oversight
Sat	Bike 100 min incl. 5×3 min VO2max	Decide based on Thursday’s fatigue; if too tired, cut to 4 sets
Sun	Run 90 min long run	Add brick (bike-to-run transition) simulation in the pre-race period

See the key point? The skeleton AI lays out is broadly correct, but whether to follow each day as written, and how much to do, depends on a host of variables it can’t see: how much sleep you got last night, today’s temperature and humidity, work stress, and subjective bodily sensations. I’m not overturning its logic — I’m adding the variable of “the human in the moment” on top of it.

Local Variables for Taiwanese Athletes That AI Almost Never Accounts For

This is especially important for us triathletes in Taiwan, and deserves its own section:

• Climate and Humidity: Taiwan summers routinely exceed 32°C with 80% humidity. The same threshold workout carries vastly different physiological costs if you do it on a July afternoon in Taipei versus in dry, cold Europe. AI’s TSS completely ignores heat and humidity — you have to adjust intensity downward yourself. Classic events like Taroko and Sun Moon Lake Come!Bikeday each have their own course and climate quirks; looking at power numbers alone isn’t enough to plan training.
• Course Characteristics: A course like Wuling — relentlessly uphill with elevation gain often exceeding 3,000 meters — demands a completely different energy system profile than a flat time trial. AI’s “500 TSS per week” sounds the same for both, but one is evenly distributed effort while the other is sustained high-intensity climbing; the preparation approaches are entirely different.
• Eating-Out and Nutrition Environment: Taiwan’s convenience stores and eating-out culture are so convenient that many athletes end up with chaotic nutrition. AI platforms barely touch your diet, yet fueling strategy affects long-distance triathlon performance as much as any interval session.

The Four “Human Verification” Signals I Always Check

Every morning (or before each training session), I have athletes self-check these four things. If two of them show red flags, the day’s plan gets scaled back:

1. Resting Heart Rate: More than 5–7 bpm above your personal baseline usually indicates incomplete recovery.
2. Subjective Sleep Quality: Two consecutive nights of poor sleep deserve more attention than any HRV number.
3. Training Motivation: A strong reluctance or resistance to training is often your body sending a signal first.
4. Muscle Soreness and Joint Sensation: Sharp, localized pain (not ordinary soreness) must be taken seriously — this is the first line of defense against injury.

Common Mistakes and Corrections: The Five Traps I See Amateur Athletes Fall Into Most

Mistake 1: Treating CTL Growth as the Only Goal

Many people become obsessed with that fitness curve and feel compelled to watch it rise every week, so they pile up TSS frantically. The problem is, when CTL rises too fast, ATL inevitably spikes in tandem, TSB stays deeply negative for extended periods, and overtraining and injury are waiting right there. The fix is simple: CTL should generally increase by no more than about 5–8 points per week (varies by individual). Better to go slower and steadier. Foundations are built layer by layer, not in a single day.

Mistake 2: All Training Happens in the “Gray Zone”

This is the most common and most regrettable mistake I see. Many people ride every session at a moderate intensity — “a bit breathless, but not really breathless” — with both power and heart rate stuck in the middle. The result: easy days aren’t easy enough, and hard days aren’t hard enough — you miss out on both ends. Research and professional team practice have long supported the broad direction of “polarized training” — large volumes of low-intensity aerobic work, paired with a small amount of genuinely hard high-intensity work, while minimizing the middle ground. If you execute AI-generated plans with “discounts” across the board, everything easily collapses into the gray zone. The fix: on easy days, be genuinely easy enough to hold a conversation while riding; on hard days, hurt enough to get the job done.

Mistake 3: FTP Set Too High, Throwing Off All Training Zones

FTP is the foundation of all power zones. If you set your FTP higher than reality to save face, then your “threshold workouts” are actually being done at “VO2max intensity,” and your “endurance rides” are actually “sweet spot” — the entire intensity distribution shifts upward, and you end up with nothing but fatigue and no results. The fix: honestly retest your FTP every 4 to 6 weeks; using a 20-minute test multiplied by 0.95 is the most practical approach.

Mistake 4: Completely Neglecting Swim and Run Technique, Only Stacking Volume

AI platforms are almost blind when it comes to “technique.” They can calculate how far you swam or what pace you ran, but they can’t tell if your catch is correct in the water or if you’re overstriding on landing. In triathlon, swimming especially relies on technique. Simply piling on volume with poor technique only entrenches bad habits deeper. Fix: Technique sessions must have someone watching you, or at the very least, record yourself for self-review.

Mistake 5: Ignoring the Specificity of Brick (Transition) Training

The hardest part of triathlon isn’t any single discipline—it’s that “run right after getting off the bike” feeling where your legs feel like lead. AI scheduling often treats cycling and running as two separate events, forgetting the importance of brick training. Fix: Schedule at least one brick session per week during the build phase to get your body accustomed to the pain of transitions.

Actionable Advice for Athletes of Different Levels

The AI coach tool works completely differently depending on where you are in your journey. I’ll break down concrete steps for three levels.

Beginners / Newbies Who Just Finished a Race (Goal: Safely Finish a 51.5 or First Olympic Distance)

• AI’s greatest value: Helping you build consistency and tracking habits. At this stage, the most important thing isn’t how precise your plan is, but whether you’re consistently moving. Let the app remind you and log your workouts—that alone puts you ahead of most people who quit after a few weeks.
• Don’t obsess over numbers like FTP or TSS from the start. Your body is still highly responsive to training; just training consistently will yield rapid progress.
• Focus on: Building solid fundamentals in all three disciplines (especially swimming), learning to structure your training week, and avoiding injury. More people are sidelined by injury at this stage than by undertraining.
• Action: Use AI to generate a conservative 8–12 week plan, but execute every high-intensity session at 80% effort initially, and observe how your body responds.

Intermediate / Mid-Level Athletes Chasing a PB (Have Finished Several Races, Aiming to Improve at 226 or 113)

• AI’s greatest value: Quantifying your fatigue management. At this stage, you’re already training enough. The key differentiator shifts from “how much you train” to “how you recover and taper.” The CTL/ATL/TSB framework becomes genuinely meaningful for you now.
• Learn to read TSB: Bringing your TSB from deeply negative back to near or slightly above zero before race day is the core operation of the taper. Master this, and your race-day form will feel like night and day.
• Start prioritizing specificity and nutrition rehearsal: In long-distance racing, your fueling strategy is just as important as your fitness. Use long rides and runs as testing grounds to find a nutrition rhythm your gut can handle.
• Action: Let AI build the framework, but spend 10 minutes each week reviewing your fatigue signals yourself, and proactively decide which days to add or cut.

Advanced / Athletes Chasing a Kona Slot or Personal Limits

• AI’s role downgrades to a supporting dashboard. At this level, individual variability is so large that generic algorithms struggle to account for it. You need highly customized coaching, which is exactly where a human coach becomes irreplaceable.
• Use data for hypothesis testing: Use the platform’s historical data to analyze retrospectively—under which training plan combinations did your FTP grow fastest? Which taper length produced your best race performance? Treat AI as your database, not your decision-maker.
• Beware of false precision in data: The higher you go, the more a 1–2% difference decides the outcome. Judgments at this level often rely on a coach’s eye—one that has seen hundreds of athletes and can read your state from a single video clip or conversation.
• Action: Find a human coach or build your own analytical framework, using AI data as supporting evidence rather than the final verdict.

Periodization: Where AI Most Often “Plans Beautifully but Points in the Wrong Direction”

I want to address periodization separately because it’s where I see AI platforms most frequently get “the numbers right but the direction wrong.” The core concept of periodization is simple—training isn’t equally hard every day; it’s divided into phases where you first build the foundation, then stack specificity, and finally taper for peak performance. But many AI platforms’ periodization logic applies a generic template and works backward from your target date. It doesn’t truly understand “what you, as an individual, are lacking right now.”

Let me use a typical 20-week 113 half-ironman (commonly called 70.3) preparation cycle to show you the key differences between phases:

Phase	Weeks	Primary Goal	Training Focus	Common AI Blind Spots
Base Phase	Weeks 1–8	Build aerobic foundation, work on technique	High volume low intensity, swim technique, strength	AI often inserts high intensity too early; foundation isn’t solid
Build Phase	Weeks 9–14	Raise threshold and specific endurance	Threshold intervals, long rides/runs, bricks	AI is weak at recognizing brick specificity
Specific Phase	Weeks 15–18	Simulate race intensity and fueling	Race-pace rehearsal, fueling tests	AI largely ignores your gut and nutrition
Taper Phase	Weeks 19–20	Flush fatigue, bring form back	Cut volume, keep intensity, TSB back to positive	AI tapers are often either too aggressive or insufficient

The base phase is where amateur athletes (and some AIs) mess up most. Everyone wants to feel that “I really trained” sensation from intensity sessions, so they skip the solid low-intensity accumulation. But the aerobic foundation is like a house’s foundation—you can’t see it, yet it determines how tall the building can go. I often tell my athletes: the base phase is boring, slow, and seems unproductive, but those eight weeks of low-intensity accumulation are what determine whether you can handle the high intensity later. If an AI template has you doing VO2max sessions in week three, I usually pull it straight back.

The taper phase is the other extreme. Tapering means “cut volume, not intensity”—reduce total training load by 30–50%, but retain a few short, sharp high-intensity stimuli to keep the body “sharp” while flushing fatigue. The goal is to bring TSB from deeply negative back to near or slightly above zero. I’ve seen AI schedule a taper that’s “almost complete rest,” and the athlete shows up on race day with legs that feel fresh but also flat, unable to find race rhythm. This delicate balance is the part of periodization that most requires a “feel” for it.

A Complete Case Study: From Algorithm Red Flags to a PB in Three Weeks

Let me give you one more complete, real-world example to ground all the concepts above. Athlete May, a 38-year-old working mom, was preparing for the 51.5 Olympic distance at the Puyuma Ironman, aiming for a podium spot in her age group. She had been training autonomously with an AI platform for over six months. When she came to me, she was stuck at a plateau—her CTL was steadily rising on paper, but her actual race results were getting worse, not better.

I pulled up three months of her data, and the problem was immediately obvious: her training was almost entirely in the gray zone (Mistake 2 mentioned earlier), and because the AI saw she was “completing all her workouts,” it judged her to be in good shape and kept adding volume. Her ATL was chronically hugging her CTL, and her TSB never once returned to positive—she had never truly rested, and she had never truly trained hard.

The adjustments I made were simple, just three things:

1. Recalibrate intensity distribution: I changed 70% of her weekly sessions to genuinely low intensity (heart rate held in the aerobic zone, able to hold a conversation while riding), and the remaining 30% had to hurt properly. The middle ground was eliminated.
2. Force in recovery: Every third week, I inserted a “recovery week,” cutting volume by 40% to give TSB a chance to return to positive. This was something the AI had never done for her.
3. Add race-pace bricks: Four weeks before the race, she did one “ride immediately followed by run” transition session per week to get used to that heavy-leg feeling.

The first week she was very uncomfortable, constantly thinking “riding this easy on a recovery day feels like slacking off.” By the third week, she messaged me herself: “Coach, in today’s threshold session, my power was higher than before, but my heart rate was actually lower.” That was the signal that her aerobic foundation was kicking in and fatigue was receding. She ended up taking third in her age group at Puyuma. The same AI platform, the same data—but with a different way of reading it, the results were worlds apart. That’s what I want to emphasize—the tool isn’t the problem; how you use it is what matters.

FAQ

Q1: Is an AI coaching app worth paying for?
For the vast majority of amateur athletes, yes—but you need the right mindset. Its greatest value lies in “recording, quantifying, reminding, and providing a reasonable starting point,” and it does these things very well. If you expect it to read your current state like a human coach and pick up on the fatigue in your voice, you’ll be disappointed. Treat it as a tool, not a teacher.

Q2: So do I still need a human coach?
It depends on your stage and goals. For finishing your first triathlon, AI plus a bit of self-discipline is usually enough. If you’re serious about setting a PB, tackling long-distance events, or qualifying for the World Championships, the judgment, corrections, and psychological support a good coach provides are something no algorithm can offer. The two aren’t mutually exclusive—the best combination is often “AI handles the recording and quantifying, while the coach handles interpretation and decision-making.”

Q3: Should I check HRV every day?
You can, but look at the “trend,” not the “single-day reading.” Single-day HRV is too noisy—a cup of coffee, a bad night’s sleep, or an incorrect measurement posture can all make it swing wildly. A downward trend over 5–7 consecutive days is what’s actually worth paying attention to. Don’t let one ugly daily number ruin a workout you were supposed to do, and don’t let a pretty number trick you into pushing too hard.

Q4: I can’t finish the workouts AI schedules for me. Am I just too weak?
Usually not. More common causes are an FTP setting that’s too high, or the algorithm not accounting for your current fatigue and Taiwan’s hot, humid environment. First, honestly retest your FTP, then factor in your subjective feelings. Not finishing a workout isn’t a sin—finishing it by injuring yourself is.

Q5: I don’t have a power meter, only a heart rate monitor. Can I still use AI platforms?
Yes, but you need to understand the limitations. Heart rate is an “outcome metric”—it reflects your body’s response to intensity and can drift due to temperature, dehydration, caffeine, and sleep. Power is an “input metric”—what you put out is what you get, making it more straightforward. Without a power meter, dial in your heart rate zones and lean more on subjective perception (RPE, Rate of Perceived Exertion) to guide you—you can still train very well. In Taiwan’s hot, humid environment, your heart rate will often be higher than it would be in cool weather at the same intensity. In that case, don’t stare at the numbers; trust your perceived effort. For running, many platforms use pace combined with GAP (Grade Adjusted Pace) to provide similar quantification.

Q6: Do AI platforms differ much from each other? How should I choose?
They differ quite a bit, but for amateur athletes, “whether you’ll keep using it” matters more than “how advanced its algorithm is.” Choosing a platform with an interface you like, one you’re willing to open every day, and one that syncs smoothly with your watch is more valuable than picking the flashiest option you’ll stop looking at after three weeks. In practice, I’d suggest: first confirm it can export your raw data (don’t get locked into a single platform), check whether the adaptive features actually adjust based on your performance, and see how detailed its analysis is for your main sport (swimming/cycling/running).

Q7: Will AI completely replace coaches in the future?
My view is this: AI will replace the “calculation and record-keeping parts of a coach’s job,” but it won’t replace “the coach as a person.” Training has never been just about numbers—it involves motivation, fear, life stress, life stages, and the psychological battles during competition. For example: an athlete suddenly feels inexplicable fear about an upcoming race three days before it. On paper, his condition is perfect, but that fear will ruin his race. What he needs at that moment isn’t an adjusted workout plan—it’s a sincere conversation, a “I’ll walk you through this.” These are things that can only be passed between people. The more powerful the tools, the more valuable the coaches and athletes who know how to use them.

Conclusion: Let Data Serve You, Don’t Be Held Hostage by It

Let’s return to Xiao Chen from the beginning. He went on to successfully complete Challenge Taiwan, and even set a new PB. He told me afterward that what he was most grateful for wasn’t the beautiful data, but that “someone saw the me who was about to break, beyond the numbers.”

That sentence, in fact, is the whole point of this article. AI and machine learning have pushed the “quantification” and “personalized starting points” of training to unprecedented heights—this is a good thing, a tremendous blessing for amateur athletes of this era. But data is always just a shadow cast by your body—it will miss those three middle-of-the-night cries, that stifling Taipei afternoon so humid you could barely breathe, that hidden fatigue you never voiced.

So my advice remains the same: Embrace data, but don’t worship it. Use AI to help you record, quantify, and see trends; use your own judgment (or that of a coach you trust) to handle the things algorithms can never see. When the day comes that you can glance at the app’s curves, feel your own pulse, and calmly say, “I’m changing today’s session”—that’s what truly using AI right looks like.

The future is already here, and it arrived faster than we imagined. May you be the one who masters the tools, not the one mastered by them. Treat the algorithm as your assistant coach—one that calculates fast and accurately but never reads your mood—while you yourself remain the head coach who makes the real decisions. See you on the racecourse.

This article is for educational purposes and does not replace individual assessment by a physician, physical therapist, or nutritionist.

References

• Definition of Functional Threshold Power (FTP) and the 20-minute test conversion: BikeRadar — What is FTP and why does it matter for cyclists?
• The relationship between FTP, lactate threshold, and power zones: TrainerRoad Blog — What FTP Really Means to Cyclists
• Definitions and time windows for TSS, CTL, ATL, and TSB: A Coach’s Guide to ATL, CTL & TSB — TrainingPeaks
• How Form (TSB) is calculated (CTL minus ATL): Form (TSB) — TrainingPeaks Help Center

Related Reading

• The Future of Sports Technology: AI, Sensors, and Personalization—Opportunities and Pitfalls Through a Coach’s Eyes
• The AI Training Revolution: How Artificial Intelligence Creates Personalized Training Plans
• Supercompensation and Training Adaptation Theory: The Physiological Logic Behind Your Workout Plan
• Sleep: The Cheapest Legal Performance Enhancer—How Coaches Help You Steal Back Your Training Gains