What's in the almanac

TSB (Training Stress Balance) = CTL − ATL. Positive numbers mean fresh; negative means tired. Above +5 you're race-ready; below −10 you're cooking. Use it to time hard sessions and tapers — chase the trend over 2–3 weeks, not a fixed number.

The PMC chart reads in trends, not snapshots. Watch CTL (orange) rising — that's fitness building over weeks. ATL (white) reacts to last week's load. The dashed projection from your active plan shows where your form lands on race day if you stick to the prescription.

The Recent Traces card shows your last six runs with feel ratings and notes. Use it to spot patterns: three "tired" days in a row before a key workout means recovery first.

The generator looks at your last 28 days of training to set a realistic baseline (it won't prescribe 70 mi/wk if you've been running 30) and ramps from there. Build → peak → taper. Peak weeks include back-to-back long runs (the only proven way to handle the second half of a 100), with a race-effort kicker so you know what 100-mile pace feels like before the gun.

Tap any week to expand it. Editing one week's mileage tells the planner you want different load — the rest re-balances forward, not backward. Use the calendar view to spot collisions with travel/work; use the list view to fine-tune week-to-week structure. Most plan failures are scheduling failures, not training-load failures.

Generate it once, then edit. Most plans need 2-3 tweaks before they fit your life — a key trip, a planned tune-up race, a known weak spot. The Coach Hub's "Plan Critique" section flags structural risks once you've generated.

A coach is $300/month and can't read 90 days of your notes in 3 seconds. The AI catches what you can't — "always tank Tuesday tempos after Sunday long runs," ramp speed too aggressive vs. your baseline, taper not actually working.

Four sections: Today's tactical brief (read before your warm-up), the weekly summary, the 4-week block read (regenerates monthly), and the notes-pattern scan that reads your last 90 days of training notes for recurring themes. Plus a plan-critique section that flags structural risks once you've generated a new plan.

The more honest your training notes, the sharper the AI gets. Even one-word entries help — "tired", "great", "stomach off". Write them right after the run while it's fresh.

The all-time fitness curve is the long answer to "am I building?" — fitness compounds slowly and a year of consistent training shows up as a slope you can see with your eyes, not just calculate.

Period comparison (30/90/180/365 days) is the cleanest answer to "am I getting better?" Direct head-to-head vs. the prior equivalent window — no smoothing, no moving averages, just totals and rates side-by-side. The delta arrow flips for pace and HR (lower is better there).

Vertical CTL/ATL tracks climbing load separately from mileage. Two equal-mileage weeks with a 5,000 ft vs. 500 ft difference produce very different fatigue signatures — the standard PMC misses that. Match the curve to the course you're targeting (heavy vert for steep ultras like Hardrock, opposite for road-flat fast 50Ks).

PR finder scans every per-second slice of every run. That 50K from August had a 5K split inside it that was your year's best, you just never saw it. Plus polarized HR check (80/15/5 distribution) and a pace-by-grade scatter so you can read your real climbing economy.

Per-mile splits, the full elevation profile sampled per-second from your watch, best efforts (1K through 10mi) extracted from the stream cache, HR trace, and your own feel rating + notes. Inline-edit the notes; tap any feel emoji to adjust.

GAP (grade-adjusted pace) is the read on your "actual effort" — what flat-ground pace would have produced the same heart rate. Useful for comparing trail runs of wildly different terrain to your road tempo benchmarks.

TRIMP is what the PMC uses; bigger TRIMP = bigger CTL/ATL bumps the next day. Filter the activities list by date range, type, distance, HR, or notes presence.

Add at least one race. Mark one as "primary" — that's your A-race. The dashboard countdown, the plan generator, and the race-day planner all anchor to your primary race.

You can have multiple races on the calendar (B-races as tune-ups, C-races as workouts in disguise). Only the primary drives the plan; the others sit there for context. Drop a GPX and the race detail page comes alive — course profile, projected splits, pacing strategy, fueling.

Generic VDOT tables assume a road marathon. Trail kills you with grade you can't run, altitude, heat, and hours your legs have never seen. Summit Line projects from your real grade-bucketed pace, not a textbook. The difference between "12-hour finish" and "14:38, MI 65 cutoff" is whether you DNF.

Set your race start time near the header. All wall-clock displays anchor to it: aid-station cutoffs show as both "+14:00 elapsed" and "22:00 local"; pacing projection layers show projected arrival times.

The Course Briefing AI reads the course objectively, then projects how you'll handle it given your training history. Hard-points list flags the segments where you'll need a deliberate plan ("walk the steep mile from 22.8–24.0; drop your pace 90s/mi for the night descent"). Read it twice in race week.

The auto numbers are a starting point — the per-hour band scales with goal time, body weight, and heat. Pick a profile (Light / Standard / Heavy / Custom) to fill the per-hour fields in one click, then tune any value.

Stage your actual fueling items in the editor — gels, drinks, real food. Set "every N min" for items you'll consume on a cadence (e.g. 1 gel every 30 min), and Summit Line auto-fills the count from your goal time. The card sums what you've planned vs. the target so you can see if you're under-fueled before race day.

325 g of carbs across the day is attainable, not automatic. Build the gut for it now — 90 g/h on long runs for at least four weeks before the race.

Your crew shows up at MI 62 expecting to pace you — what do they need to know? Drop bag at MI 41 needs which gels? Pack weight at MI 0 vs MI 65? Caffeine timing for the H22 sleep-dep window? All of it on one screen, all editable race week.

Drop bags route to specific aid stations; crew members get assigned aid stations and tasks; pacers get from→to leg assignments. Kit items roll up to a pack-weight number you can actually see. Mandatory-gear flag tracks what must go vs. what's optional.

Only shows for races over 24 hours where you'll cross at least one circadian low. The card maps your alertness from the start time you set on this race — caffeine markers flag windows where a 100mg top-up keeps you sharp.

Use it to plan crew handoffs and pacer entry around your low points (typically 02:00–05:00 local). A pacer hitting your circadian trough at MI 65 does more for the finish than one at MI 30.

One-click connect for Garmin, Coros, and Suunto. The first sync pulls your last 12 months of activities; after that, your watch service keeps it fresh in near real-time. Every chart on the site builds from the same canonical metrics store derived from these activities.

Manual re-sync available in Settings if anything goes stale. Disconnect at any time — your derived analytics stay; new activities just stop flowing in.

Body weight scales the fueling card's per-hour fluid + sodium prescription to you. Birthdate sets your maxHR estimate (used by HR-zone calculations when you don't have lab data). Timezone groups activities into the right calendar day for weekly aggregates and the dashboard countdown.

You can leave fields blank — the math degrades to a field-average athlete. But filling them in makes every projection on the site sharper.

Every upcoming race in chronological order. The primary (A-race) is highlighted in brass-gold; the rest are tune-up Bs and Cs. Tap any race chip to open its detail page.

The list view is best for chronology; switch to the month grid for the visual shape of the season.