climbing // training log
← active plan

Mesocycle 4 — Depth by repetition: the same three problems, every Monday

Span: 03-08-2026 → 29-08-2026 (4 weeks: 3 load + 1 deload) Sessions: 16 (Mon/Wed/Fri/Sat) Theme: cycle 3 had the right idea and the wrong execution. It prescribed depth, logged 25 → 9 engagement rows at the 7A band, and produced zero 7A sends (down from 2 in cycle 2). The reason was not load, RPE or finger strength: across the whole cycle 12 distinct project names each appeared on exactly one date. The projects were rotated weekly, not ground across sessions. Gym sets stay up ~8 weeks, so the problems were still there and simply were not returned to. Cycle 4 removes every option to rotate. Success = at least one 7A redpoint, and a rising high-point trend on three named problems.

What changed from cycle 3 (and why):

  • Three NAMED projects, carried in the plan itself. Not "2–3 tracked projects, rotate by closeness-to-send" — the actual names are written into the Monday steps and tapped in as chips when logging. A trend keyed on a name the athlete retypes from memory cannot survive a week; cycle 3 proved that empirically.
  • Ceiling probe CUT. Breadth was the diagnosed failure, so the block that manufactures breadth goes first.
  • Campus CUT. Contact power is not the limiter at +40 kg pull 1RM and 124 % BW on both grips, and it was consistently the first thing skipped anyway.
  • The recovered ~18–20 min goes straight back into the project block: ~5 goes over 3 projects becomes 9 goes over 3 projects, 3 each. This concentrates the existing ≤8–10 hard-attempt cap, it does not raise the load.
  • A problem-count cap, not just an attempt cap. 13-07 hit 33 goes across 12 problems at RPE 8: the attempt cap never fired because the breach was in breadth. Monday is now capped at 3 problems, full stop.
  • Pinch block keeps its slot and finally gets a number. Three weeks of dedicated pinch work in cycle 3 produced nothing falsifiable because hangboard_tests had no pinch channel. It does now: log the block every Wednesday.
  • Weighted pull-ups untouched. Athlete-confirmed, enjoyed, and progressing (+36 → +38.5 → +40). Not a candidate for cuts.

Glossary: open /protocols → glossary for any term (half-crimp, open-hand, repeater, PAP, contact strength, RFD, completion-anchored, etc). Keep on a phone tab during sessions.

Baseline reference (standing anchor 06-07-2026; pull-up retest W1 Mon 03-08)

Only the pull-up is retested this cycle. Half-crimp has read +15 kg on three consecutive tests (08-05, 08-06, 06-07) — a fourth flat read buys no information and costs fresh fingers on a limit day. Repeaters are completion-anchored and never derive from a max, so nothing in the plan depends on a finger retest.

Test Standing anchor (06-07) Retest 03-08 Notes
Half-crimp 20mm 7s max +15 kg · 124 % [id 6] not retested flat ×3 tests; tracking reference only
Open-hand 20mm 7s max +15 kg · 124 % [id 7] not retested level with HC since 06-07
Pull-up 1RM +40 kg (total 102 kg · 164.5 %) [id 4] retest W1 Mon working ≈ 86 % of total = +25.5 kg × 3; ramp = +13 kg
Pull-up max reps @ BW 14 [05-08] carry stale, not load-bearing
Pinch block 7s bodyweight / block only, no added kg [06-07] log every Wed completion-anchored (6 clean 7s reps/hand, last rep hard). NOT a max hang
Self-grade boulder flash 6C / RP 6C+ — —
Bodyweight 62 kg [id 7] confirm W1 Mon —

If the 03-08 retest moves the 1RM: write the new number into this table and recompute every %-derived load to explicit kg the SAME session — node scripts/recompute_loads.mjs --check webapp/cycles/2026-08-cycle-4.md. A max logged in the app does nothing for the plan; the plan derives from this file. The kg written below are correct for a +40 kg anchor.

The three tracked projects

These are commercial gym sets, not Kilter. Confirmed 28-07: sets stay up ~8 weeks, comfortably longer than this cycle, so a problem picked in W1 is still there in W3. That is the whole premise.

Slot Problem Grade Crux move type High-point Sessions tried
A Yellow crimpy 7A crimp 8/9 (13-07) 1
B Un cast shadow 7A static reach 5/9 (20-07) 0
C Power long reach 6C+ static reach 5/10 (03-08) 1
  • All three are named in the plan, not picked on the day. Chosen 29-07 off the 13-07 high-point rows; every Monday project step carries a projects: hint so the chips write problem_name byte-identical and the trend actually joins. Do not retype a name the chip already offers.
  • A is locked and starts W1. Yellow crimpy was at 8 of 9 holds on 13-07 and has not been touched since. It has been up since at least 13-07, so its remaining life is unknown and possibly short. It does not get deferred to W2.
  • All three were picked on distance-to-send, not on crux variety. Originally A and B were both crimp on purpose: crimp was the #1 crux of cycle 3 (9 tags) and is the actual limiter at 7A.
  • B and C were re-picked 03-08, after W1 ran. Technical crimps turned out to be a 6C, not a 7A, and you redpointed it that night, so it is off the list. Crazy hard sloper balancy was replaced by Power long reach (same 6C+ static-reach family, high-point 5 of 10). B is now Un cast shadow — 7A at 5 of 9 on 20-07, the closest to send of the 7As that are left.
  • Consequence, stated plainly: crimp is now covered by A alone, and B + C are both static reach. That is a real change to the cycle's shape, accepted because static reach is cycle 3's #2 crux (6 tags) and because the alternative for B was Súper hard crimp power at 2 of 8 — a trend line, almost certainly no send. If a crimpy 7A you can get off the ground on appears on the wall, it takes B's slot.
  • W1's steps were left exactly as authored (they were trained on 03-08). The swap lands on W2 and W3 only, so Technical crimps and Crazy hard sloper balancy still appear on the W1 Monday and in its prose. That is deliberate: trained days are not rewritten.
  • If one sends: promote it to redpoint, then pick a replacement of the same crux family and carry on. Do not use a send as a reason to rotate the other two.
  • If one stalls 3 sessions with a flat high-point: it is the wrong problem for now. Replace it, and say so in the session notes.
Movement-type tags — one tag per go, filled in at the log step

Hold crimp · pinch · sloper · pocket · undercling · gaston Move dyno · bump · compression · static reach · cross-through · lock-off · mantle · high step · drop knee · sit-start Foot heel hook · toe hook · smear Style/wall coordination · balance · slab · tension

Crimp was the #1 crux tag in both cycle 2 and cycle 3 (9 tags each) and has not moved. Static reach is the new #2. Every hard go gets a tag, but you do NOT stop to log between goes: climb the 3 goes, then the log step opens 3 rows at once and you tag them there. The re-rank at the checkpoint is what sets cycle 5's emphasis.

Priority hierarchy
  1. Depth on three problems — Mon, 3 goes each, high-point logged every go
  2. Crimp under power — the standing #1 crux; project A is a crimp problem by design
  3. Pinch strength — Wed dedicated block, now with a logged benchmark
  4. Finger strength maintenance — Wed repeaters, completion-anchored (NO max hangs)
  5. Maintain — weighted pull-ups + front-lever project + antagonists
  6. Movement skill — Fri Perfect Repeats + foot precision
  7. Light PE — Fri only, easy dose
Hard rules (override everything below)
  • No full crimp — half-crimp + open-hand only, through 09-11-2026 (6-month rule from 09-05-2026)
  • No max hangs — repeaters only, completion-anchored (30mm edge, bodyweight; shrink the edge toward 20mm BEFORE adding any weight)
  • Hangboard + loaded pinch Wed gym only — never on consecutive or rest days
  • Monday is capped at 3 problems. Not "about three". If a fourth looks tempting, that is the cycle-3 failure restarting.
  • Limit-volume cap: ≤ ~9–10 hard attempts on the Monday project block. Two consecutive Mondays at first-block RPE 9 with no high-point progress → next Monday −2 attempts.
  • Rate every session. A blank RPE is not a neutral outcome: it disarms the load-watch guards, which cannot evaluate a missing number. The done screen will ask if you skipped it.
  • Front-lever = first cut if Monday limit quality or Wednesday first-hang quality dips
  • Stop on sharp finger / elbow / shoulder pain (not soreness, not pump)
  • Missed sends 2× in a row OR sleep <6 h × 2 nights → next session −1 level (see the low-readiness line in each session)
  • Sun = full rest, no exceptions
Cut the front-lever / legs block if any of these flip
  • First hang of the day feels heavy → cut front-lever, keep fingers
  • Monday first project go is worse than last week's last go → cut front-lever
  • Elbow or shoulder niggle → cut both, keep climbing only
  • Session running past its stated total → cut the accessory block, never the project block

Sessions

End-of-cycle checkpoint (Sat 29-08-2026 evening)

Run /checkpoint with the cycle selected. This cycle is falsifiable on purpose — three questions, three places to look:

  1. Did a 7A go down? query.mjs sends --since 2026-08-03 --grade 7A. Cycle 3 produced zero. One redpoint is the bar.
  2. Did the high-point trend rise on the three tracked problems? The project-depth table needs ≥2 sessions per problem and a non-zero Δ. Cycle 3's Δ was 0 on every problem because each name appeared once. If any problem still shows one session, the rotation habit came back and that is the finding.
  3. Did the pinch block move? W1 → W2 → W3 at a comparable load, in the "Pinch block (completion)" row. Flat across three weeks = the block does not earn a slot in cycle 5. This is the number cycle 3 could not produce.

Also re-rank the move-type tags. Crimp has been #1 for two cycles running; if it is still #1 after a crimp-led cycle, the problem is not exposure.

Logging discipline

  • Tap the project chip, never retype the name. The chip writes the string the trend groups by. Retyping is what flatlined cycle 3.
  • High-point on every project go, not just the good ones. A go that ends at hold 3 is data.
  • A send is redpoint, not project. The style field is what the pyramid counts.
  • Tag the crux move-type on every hard go (target >75 %).
  • Rate every session. A blank RPE disarms the load-watch guards; they cannot evaluate a missing number, and silence gets read as all-clear.
  • Gym sets log as bouldering; only the Wednesday Kilter block is board. Getting this wrong corrupts every discipline-partitioned chart (see migration 029).

Red flags — pause the cycle, message Claude

  • Sharp finger, pulley, elbow or shoulder pain that persists past a session
  • Two consecutive Mondays at RPE 9 with no high-point progress
  • Sleep <6 h for two nights running
  • Any joint that is still complaining on the next rest day

Rest days

Tue, Thu, Sun. Sunday is a full rest day before Monday — no exceptions. Bands and mobility only on rest days; loaded hangs stay Wednesday-gym-only.