Transparency

No black box. Not AI. Here's exactly how your coach thinks.

TrailFree's trainer runs entirely on your phone. The plan it builds comes from coaching rules and plain arithmetic, all laid out here. Read the short version, or dig into every constant and formula and audit any plan it makes.

Part one: in plain language

The shape of a plan

You tell it a goal (say a 100 km ride) and how long you have. It arranges your weeks into the four phases every coach builds from:

• Base builds your aerobic engine with steady, unhurried volume.
• Build keeps the volume and adds intensity (tempo, then intervals).
• Peak is the hardest, most event-specific stretch, where the long session tops out.
• Taper eases right off so you arrive at the event fresh.

Your long session, week by week

The one non-negotiable session each week is the long one. It grows steadily, backs off every fourth week so you can absorb the work, peaks two to three weeks out, then tapers. Below is a real 16-week plan to a 100 km ride, built from a 35 km starting point. Every bar is the engine's own output.

Read across it: no step is more than 10% up on the week before, every fourth week dips for recovery, the biggest training session is 78% of the goal (you never ride the full distance in training), and the final week is the event.

Five rules it never breaks

• Progressive overload, capped. No week's long session climbs more than 10% over the last one.
• Recovery every fourth week. Three weeks on, one week easy. Adaptation happens in the rest, not the effort.
• Peak at 78%. The biggest session is 78% of your goal distance, two to three weeks out. Doing the full distance in training just costs you the event.
• A real taper. The last week or two ease off so you turn up fresh.
• Honest feasibility. If your date is too soon to get there safely, it says so and offers the earliest realistic one. It won't sell you a plan that would hurt you.

It watches what you do

• Auto-matching. Record an outing and it ticks off a planned session on its own, as long as the outing covers about 70% of that session's target.
• Replans without blame. Miss a week and it freezes the past and rebuilds the rest from where you really are, never skipping a step and never jumping more than 10%.
• A gentle heads-up if the date has grown tight for the distance. A nudge, never a scolding.

Form: fresh, in rhythm, or tired

If you've been recording a while, the trainer compares your last week's load against your longer-term rhythm and tells you one honest thing: Fresh (rested, good day for the hard session), In rhythm (steady, carry on), or Carrying fatigue (the last week was big, an easy day is training too). It stays quiet until it has enough history to be honest.

Heart-rate zones, if you want them

Pair a Bluetooth strap and set your max heart rate, and every session gains a plain zone briefing alongside the effort cue. No strap, no problem: the words alone still guide you.

Strength, and what you're training for

Strength is coached differently from endurance. Instead of one long build, it starts with a question: what are you training for? Pick a focus and the routine shelf reshapes around it, so a runner isn't handed a cyclist's plan:

• Everyday, Cycling, Running & hiking, Lean & toned, Muscle building, or Long-term health.

Follow a routine and it hands you each move in order, compounds first, with sets, reps, a rest timer, and a plain form cue for every exercise. The next time round it reads your last log and prescribes the next step: a little more weight, a rep more, or a few more seconds, only once you've earned it. Fall short and it holds you where you are. Log the workout and it ticks off your plan's weekly strength session.

What it deliberately won't do: prescribe near-maximal single-rep lifts or explosive plyometrics. Those need a human watching your form, so they're left out on purpose.

How it changes for your sport

The endurance plan speaks in distance, so the same principled build works whether your goal is a ride, a run, a hike or a long walk. It auto-detects which you're doing when you record, and a recorded outing only completes a session of the plan's own type, so a run counts toward a run plan, a ride toward a ride plan.

Being honest about the limits: it doesn't set pace or power targets, and it doesn't model sport-specific physiology beyond distance, time and climbing. That's a deliberate choice, the simple, robust model a coach hands someone without a lab, rather than a number that pretends to a precision it doesn't have. Where your sport really changes the work is in the gym, through the focus you pick above.

Weekly goals and the coach

Separate from a plan, you can set one recurring weekly goal, and choose a coaching voice (or silence) to nudge you toward it. Six goals, split into effort and habit:

• Activity: distance, outings, time, or climbing.
• Lifestyle: steps, or days out.

Part two: the rules engine

For the curious

Everything above comes from the constants and formulas below. They live in one small, pure, unit-tested engine. A coach could re-tune any number here without touching the structure. Distances are in metres internally and rounded to the nearest 500 m on screen.

Every constant

Name	Value	What it does
Max weekly growth	+10%	The most a long session may rise from one progression week to the next.
Peak fraction	78%	The peak long session, as a fraction of goal distance.
Min start fraction	15%	Floor for the starting long session, if your history is thin.
Recovery fraction	62%	A recovery week's long session, versus the week before.
Taper long fraction	50%	A taper week's long session, versus the week before.
Missed-week floor	60%	When replanning, the anchor never drops below 60% of the last planned long.
Behind fraction	70%	You're flagged "behind" if the best remaining long is under 70% of goal.
Match fraction	70%	An outing completes a session if it covers at least 70% of its target.
Taper length	1-2 wk	Plans of 12+ weeks get a two-week taper; shorter plans get one.
Min plan length	4 wk	The shortest plan the engine will build.
Ability window	8 wk	How far back it looks at your history to gauge a starting point.
Target rounding	500 m	Every distance target is rounded to the nearest 500 m.

How a plan is generated

1. Starting point. It reads your last 8 weeks of the chosen activity to find your longest recent outing, then floors it so a plan always has somewhere to grow from:

start = max( your longest recent outing, 15% of goal )

2. Feasibility. It works out the fewest weeks in which that start can safely reach the peak, and refuses shorter dates (offering the earliest realistic one instead):

peak = 78% of goal progression_weeks = ceil( ln(peak / start) / ln(1.10) ) minimum_weeks = max( 4, ceil(progression_weeks × 4/3) + 2 )

The × 4/3 makes room for the recovery weeks; the + 2 for the taper.

3. Phase split. One or two weeks are reserved for the taper; the rest are split 40% Base, 40% Build, and the remainder Peak (each at least one week).

4. The long-session ramp. Across the non-recovery weeks it grows geometrically toward the peak, but never faster than the cap:

steps = number of progression weeks ratio r = min( (peak / start) ^ (1 / (steps - 1)), 1.10 ) long(week n) = start × r ^ n, capped at the peak

5. Recovery weeks. Every fourth working week (except the last, and never inside the taper) drops to 62% of the week before, then the climb resumes.

6. The rest of the week. Weekly volume is a multiple of the long session, and the leftover is spread across the other sessions by effort:

Sessions / week	Weekly volume	Quality work
3 or fewer	1.8 × long	Tempo (Base), Intervals (Build+)
4	2.2 × long	Tempo (Base), Intervals (Build+)
5	2.6 × long	Both Tempo and Intervals

Support sessions are weighted Tempo 1.2, Intervals 1.0, Easy 0.8. Recovery and taper weeks stay all-easy. A weekly strength session is added through Base and Build if you asked for one.

7. Taper and event week. Taper weeks drop the long to 50% of the week before; the final week's "long session" is the event itself, with a couple of very short easy spins around it.

How it adapts to reality

• Matching is recomputed from scratch on every sync, so it self-heals if you delete, rename or reclassify an outing. Within each Monday week, of the plan's activity type, the largest outstanding session is matched to the largest outing that covers at least 70% of it; each outing counts once. Strength sessions complete from any logged workout that week.
• Replanning freezes every past week and rebuilds the rest, anchored on what you achieved, clamped so the next step never collapses or jumps:

anchor = clamp( your recent longest, 60% of last planned long … last planned long )

• Behind is flagged only when the best remaining long can't reach 70% of the goal. It changes the tone, never the effort.

The form model

This is the calm version of the fitness-and-fatigue model sports scientists call CTL/ATL (the app never uses the acronyms). Two exponentially-weighted averages of your daily training hours, over a 90-day window:

chronic load = EWMA of daily hours, 42-day time constant (your fitness) acute load = EWMA of daily hours, 7-day time constant (recent fatigue) alpha = 1 - e^(-1 / days) Carrying fatigue when acute > 1.30 × chronic Fresh when acute < 0.60 × chronic In rhythm otherwise

Both averages are seeded at your window's mean so a short history doesn't fake a fitness ramp, and it says nothing at all until you have at least 4 outings across 2 weeks.

Weekly goals and coach states

Goal	Kind	Presets
Distance	Activity	20 / 30 / 50 km
Outings	Activity	2 / 3 / 5
Time	Activity	60 / 120 / 180 min
Climbing	Activity	300 / 500 / 1000 m
Steps	Lifestyle	30k / 50k / 70k
Days out	Lifestyle	2 / 3 / 5

The coach's tone follows where the week stands: met (100%), close (70% or more), underway (some progress), or a fresh start (nothing yet). The words change with your chosen voice; the thresholds don't.

Strength: routines and progression

The gym side is a catalog of 13 coach-written routines and 22 exercises in four groups (legs, core, upper body, mobility), spread across the six focuses so each has at least two. Curated and offline, with no cloud database and no videos; the exercise keys are permanent so your records never break. Each routine is an ordered list of blocks, compounds first:

block = ( exercise, sets, reps, rest seconds, form cue )

plus a written progression rule for when it gets easy (reps double as seconds for timed holds). Deliberately excluded, and documented as such in the code: near-maximal single-rep lifts and plyometrics, which shouldn't be prescribed without a human watching form.

Adaptive prescription. Each session, every block is prescribed from your last log for that exercise, the way a trainer would:

• New to it gets the routine's base sets and reps.
• Hit every target rep and it progresses: +2.5 kg (with reps reset to the bottom of the range) on loaded lifts, +1 rep on bodyweight moves, or +5 s on timed holds.
• Fell short and it holds today's target where you got to. No shame, no doubling down.

The science behind it

These rules encode mainstream endurance and strength-training science. Here is where each principle comes from, in plain terms, and where the evidence is still argued, because an honest coach tells you both. This is a starting bibliography: the engine follows the principles.

• Periodization (Base, Build, Peak, Taper). Structuring a season into progressive phases, with adaptation consolidated in planned recovery, is the foundational framework of training design. Bompa & Buckley, Periodization: Theory and Methodology of Training (Human Kinetics). Issurin, "New horizons for the methodology and physiology of training periodization," Sports Medicine, 2010.
• Progressive overload, capped near 10% a week. Increasing load gradually is bedrock; the specific "10% rule" is a widely-taught safety heuristic. We use it as a ceiling, not a promise, because the evidence for that exact figure is mixed. Gabbett, "The training-injury prevention paradox," British Journal of Sports Medicine, 2016. Contested: Buist et al., "No effect of a graded training program on the number of running-related injuries in novice runners," American Journal of Sports Medicine, 2008.
• Recovery week every fourth (3:1 loading). Fitness is built during recovery, not the session; scheduled back-off weeks are standard periodization practice. Bompa & Buckley (above). Zatsiorsky & Kraemer, Science and Practice of Strength Training (Human Kinetics).
• A peak long session, then a taper. Reducing volume before the event while keeping a little intensity reliably improves performance in meta-analysis. The peak long being a fraction of goal distance is endurance-coaching convention rather than one lab number. Bosquet et al., "Effects of tapering on performance: a meta-analysis," Medicine & Science in Sports & Exercise, 2007. Mujika & Padilla, "Scientific bases for precompetition tapering strategies," MSSE, 2003.
• Form: fitness versus fatigue. Modelling readiness as a slow fitness signal balanced against a fast fatigue signal goes back to the impulse-response model; the acute-to-chronic load ratio is a popular modern version. Banister, impulse-response modelling of training, 1975. Gabbett, 2016 (above). Contested: Impellizzeri et al., "Acute:Chronic Workload Ratio: conceptual issues and fundamental pitfalls," International Journal of Sports Physiology and Performance, 2020. We use it only to say one gentle thing, never to prescribe.
• Easy-hard intensity and heart-rate zones. Most endurance work should be genuinely easy, with a smaller dose of quality; zones as a percentage of max heart rate are the simple field version of that idea. Seiler, "What is best practice for training intensity and duration distribution in endurance athletes?" IJSPP, 2010.
• Strength progression: earn the reps, then add load. Progressing volume and intensity over time, adding weight once you own the top of the rep range, is the consensus resistance-training model. American College of Sports Medicine position stand (Ratamess et al.), "Progression Models in Resistance Training for Healthy Adults," MSSE, 2009.

None of this needs a subscription or a lab. It's the model a good coach hands someone with a watch and a bit of consistency, made reproducible so you can see every step.

Back to training