Quantifying training load in climbing: what to actually use

Simple question: how much did your athlete train last week?

Almost no climbing coach can answer that with a number. And of those who have one, a fair share are using a number that does not mean what they think it means.

This is not a failure of the trade. It is that climbing, unlike nearly every sport that reached training science earlier, has no unit of load. Weightlifting has tonnage: sets × reps × kilos, and out comes a number you can add up, compare and plot. Endurance sports have minutes per intensity zone, and a heart rate monitor to verify them. Climbing has no equivalent, and the field solves it in three ways, all of them bad: measure nothing, count hours, or borrow football’s sRPE without checking whether it holds.

This article is about why the problem is real, which proxies exist, and what the minimum is that you can actually sustain week after week without abandoning it in October.

One thing up front: you will not find how much load an athlete should carry here. That depends on their history, age, tissue and season, and it is decided by whoever coaches them. What you will find is how to build the measurement system so that decision gets made looking at data rather than at memories.

Why tonnage does not translate

Three reasons, and none of them is fixed by a better formula.

The weight moved does not determine the stimulus. A 65 kg climber does the same pull-ups they did six months ago, but the difference between today’s session and that one lives in the edge, the grip type and the percentage of maximum they hang at. The variable that matters appears in no kilo count.

The same duration contains incomparable sessions. Three hours in the gym can be continuous volume on big holds with a heart rate of 110, or six attempts on a limit boulder with twenty minutes between them and an hour of talking. Both are “3 h”. One wrecks the fingers and the other barely touches them.

Rest is part of the prescription, not dead time. In endurance, more minutes is more load. In climbing, lengthening the rest between attempts raises the quality of the stimulus and lowers accumulated fatigue. Any time-based metric penalises exactly what the coach wants to happen.

Hence the three usual answers. Measuring nothing leaves the coach deciding on feel, which is why an injury is always a surprise. Counting hours produces a time series that moves, which gives a false sense of control. And borrowing sRPE as-is works better than you would expect, but only for half the problem.

Climbing needs two axes of load, not one

This is the part that decides everything else, and the part nobody says out loud.

A climber accumulates two distinct loads that do not move together:

And now the practical consequence, which is a strong one:

A deload week that cuts total time but keeps the work on small edges is not a deload for the fingers.

This is exactly how people get hurt during the weeks the plan calls easy. The athlete trains fewer hours, feels fresh, and precisely because they feel fresh they pull harder on the little bouldering they do. The systemic axis goes down and the finger axis goes up. If you only measure one, you see a quiet week.

This is not an exotic hypothesis: it is the mechanical explanation for the patterns already described in when to cut the load on finger injuries, where the overload signals appear without any apparent rise in volume. And it is why a well-built deload - the 3:1 progression inside a mesocycle - has to specify what drops on each axis, not just how much the total drops.

Working rule: every time you touch volume, say what happens to the edge. If the answer is “same”, you have not deloaded the fingers.

What each proxy captures and what it drops

There is no good proxy. There are six mediocre ones, and the coach’s job is knowing which one answers which question.

None works alone. And the temptation to build a single index combining them into one tidy number is precisely the mistake: you end up with a figure that rises and falls without anyone knowing why, and that nobody uses to decide anything.

What does work is picking one per axis and never changing it.

sRPE: how it is calculated and where it breaks

sRPE is the borrowed metric that holds up best, so it is worth understanding properly before deciding how far it goes.

The calculation. The athlete rates the whole session from 0 to 10 - not an exercise, the session. That number is multiplied by the duration in minutes, giving a figure in arbitrary units. A 90-minute session at RPE 7 is 630 AU. The week’s load is the sum of its sessions.

Three protocol details that decide whether the data is worth anything:

• You ask afterwards, not during. The convention is to wait around thirty minutes from the end. Asked in the heat of it, the last attempt dominates the answer and the whole session gets scored by its worst minute.
• The question is always the same, literally. “How hard was the session?” is not the same as “how tired are you?”. The second one measures last night’s sleep. Change the wording mid-season and you break the series.
• The scale is explained once and never retouched. A 7 has to mean the same thing in January and in June. That is harder than it sounds: athletes recalibrate on their own as they improve, and March’s 7 ends up being October’s 8 in absolute terms.

There is a derivative that rarely gets mentioned and is cheap: monotony, the week’s daily mean divided by its standard deviation. It translates something the total cannot see: a 3,000 AU week spread over three hard days and four rest days is not the same week as 3,000 AU spread evenly over seven. The second has higher monotony, and sustained high monotony is a pattern worth watching. You do not have to calculate it every week; you just have to stop confusing “same load” with “same week”.

Where it breaks in climbing. On the finger axis, and it breaks badly:

• A 25-minute hangboard session can be the most aggressive tissue stimulus of the week and yields 25 × 8 = 200 AU, less than a long wander around the gym.
• A three-hour projecting session with long rests inflates the time denominator with minutes the athlete spent sitting down.

The partial fix is to use working time, not room time: the minutes of the session the athlete actually spends climbing or hanging. It is an estimate, but a consistent estimate beats an exact measurement of the wrong thing.

The real fix is not asking sRPE for what it cannot give. It is a good number for the systemic axis and a bad number for the fingers. Use it for the first and measure the second separately.

The finger axis: what can actually be counted

There is no borrowed metric here, so it has to be built. The good news is that it does not need to be precise; it needs to be consistent.

What loads finger tissue is the combination of edge depth, grip type, number of fingers involved, intensity relative to maximum, and time under tension. None of those five appears in hours or in RPE.

A weekly count you can sustain without tooling:

Finger exposure = the sum of sets and attempts on demanding grips, weighted like this:

The weights are arbitrary. And that is fine, because the number is not compared against another athlete or against any published reference: it is compared against the same athlete four weeks ago. The only thing the weighting has to do is not change. A crude system applied every single week detects step changes; a refined system applied now and then detects nothing.

On the question everyone asks - how many hangboard sessions per week - it is the wrong question. The frequency of dedicated sessions is not the variable, because hard bouldering loads the fingers too, and in many athletes it loads them more than the board does. Two “finger sessions” plus four days of bouldering on slopers is six exposures, not two. Count exposures, not sessions, and the number stops lying to you.

For this axis to have a hard reference rather than just a tally, you need a strength baseline measured with a fixed protocol: finger strength testing with a dynamometer covers which one and how to read it. Without a baseline, the count tells you exposure went up; with a baseline, it also tells you whether the tissue is accepting it.

The load you are not counting

Everything above measures what is inside the plan. The problem is that a good share of the real load is outside it.

Weekend climbing. A flawless Monday-to-Friday plan plus three days projecting outdoors is not a flawless plan: it is volume nobody is adding up. If trips do not go in the same place as the plan, the weekly calculation is wrong by definition, not by imprecision.

The approach and the full day. Forty minutes uphill with a pack, six hours standing at the base and the walk back down are real systemic load that appears in no climbing log. A weekend outdoors with little climbing can still be a hard week.

Hours spent belaying. A day of routes with a partner is hours of forearm and shoulder tension that nobody counts. It is not the main factor, but it explains why some athletes arrive on Monday worse than their logged moves suggest.

The athlete’s working week. This is the biggest distorter and the most ignored. Someone who works eight hours with their hands - carpentry, construction, kitchens, physiotherapy - arrives at the board with tissue that has already worked. An athlete with three work trips and four hours of sleep is not in the week the plan says. We flagged it when writing about planning: a week with three trips is not a high-load week, it is a forced deload.

None of this needs to be quantified precisely. What it needs is somewhere for the athlete to say it, and for that somewhere to be where the plan lives. When the information arrives by WhatsApp it stops being data: it is not anchored to a date, it cannot be sorted, and in three months nobody can find it. It is the same problem that makes coaching remotely depend entirely on what the athlete reports.

The acute:chronic ratio, honestly

It is the fashionable metric the moment anyone starts talking about load, so it is worth telling the whole story, including the part that does not make it into the courses.

What it is. You divide the load of the last 7 days by the average of the last 4 weeks. If the result is 1.0, this week looks like the ones before it. If it is 1.6, you have added 60% more than the athlete had been absorbing.

Why it got famous. A “safe zone” between 0.8 and 1.3 circulated widely, along with the idea that stepping outside it multiplies injury risk.

Why that version does not hold up. The methodological criticisms have been piling up for years and they are serious, not quibbles:

• Numerator and denominator share data. The current week is part of the four-week average, and dividing two coupled quantities produces correlation even when no real relationship exists. It is a known mathematical artefact.
• The cut-offs are arbitrary. Neither 7 and 28 days nor 0.8 and 1.3 come from physiology. They come from a choice that later hardened into a convention.
• The ratio erases absolute magnitude. 400 AU over 400 and 40 over 40 both give exactly 1.0. One athlete is training and the other is parked, and the metric sees them as identical.
• It is association, not causation. And several attempts to replicate the original findings in other contexts have failed.

How to use it anyway, because you can. As a trend signal and never as a threshold:

1. Watch the direction and the size of the jump, not a line being crossed. A 1.45 this week in an athlete who had been sitting at 1.0 is a question to ask, not an alarm going off.
2. Calculate it per axis, not globally. A systemic ratio of 0.7 sitting alongside a finger ratio of 1.5 is the badly-built deload week from earlier, and globally it comes out at 1.0: perfect and false.
3. It means nothing before four weeks of data. With a new athlete the denominator is noise. For the first two months the metric does not exist.
4. It never decides alone. A high ratio in an athlete reporting zero niggles, sleeping well and completing sessions is different information from the same ratio with two pain flags.

If this sounds like something already said in other words, that is because it is: comparing the current week against the average of the previous four and distrusting jumps above 30-40% is this ratio without the decimal place and without the false precision. That is, honestly, the version most worth using.

The workable minimum: three numbers per athlete per week

All of the above reduces to something you can sustain. Three numbers, one per question:

1. Weekly systemic load. The week’s sRPE total, using working time. Answers “how much have they trained?”.
2. Weekly finger exposure. The weighted count from the previous section. Answers “how much have the fingers worked?”, which is a different question and the one that predicts the expensive problem.
3. Percentage of sessions completed. Answers “was the plan real?”. It is the most underrated of the three: a drop in adherence is almost never motivation, it is that something hurts and the athlete is avoiding the stimulus without saying so.

Plus a binary that is not a number but outweighs all three: pain yes/no, and where.

That is enough to read a week. The four combinations that show up in practice:

That third row is the one you cannot see without splitting the axes, and it is the one that sends people to the physio.

Mistakes that ruin the series

The system almost always fails for the same reasons, and never for lack of sophistication:

• Changing the scale mid-season. Refining your method in March means throwing away January and February. If you are going to change something, start a new series and say so.
• Asking for RPE in the heat of it. Turns a session metric into a metric of the last attempt.
• Measuring some athletes and not others. Partial data invites comparing between athletes, which is exactly what you must not do.
• Comparing between athletes. These numbers are not comparable across people. They are individual series, and they only mean something against the same athlete’s past.
• Not logging outdoor climbing. The most common mistake and the most expensive, because it hits the week with the most load in it.
• Measuring and not using it. If nobody mentions the log on Monday, it stops being filled in on Tuesday. Data that changes no decision is admin.

What decides whether this survives February

None of the above is hard to calculate. All of it is hard to maintain.

Three numbers per athlete per week, across twenty athletes, is sixty weekly calculations. And they are not sixty clean sums: each one needs you to pull up the week’s sessions, pull up the RPE each athlete reported, classify every attempt by grip type, remember to add the weekend outdoors, and compare it against the average of the previous four weeks. Nobody does that by hand two months running. It gets done in January with enthusiasm, half-done in February, and by March it is gone.

Which is why this only works if the data goes in structured from the start. An RPE written in a chat cannot be summed; an attempt described in a notes field cannot be classified. It is exactly the argument in what a tool for climbing coaches has to do: execution mode, grip and session log are not interface details, they are the columns without which these three numbers do not exist. And it is the same boundary where the spreadsheet stops holding up: the information is there, but scattered, and nobody cross-references it.

The conclusion is not that a better metric is needed. It is that a crude metric applied every week beats a perfect metric applied three times. Pick one proxy per axis, never change it, count what happens outside the plan too, and treat any ratio as a question rather than an answer.

Do that, and in six months you will be able to answer how much your athlete trained last week. Which today, almost certainly, you cannot.

---

ClimbrLab calculates this for you: session logging with RPE on the athlete’s phone, prescription by execution mode with grip and edge as their own fields, outdoor trips in the same history as the plan, and dynamometer measurements per hand. Climbing coaching software with early access open.