Coming Back After Time Off: What to Lift in Session One
Every article gives you a percentage to take off the bar. None of them tell you why — or how to know when you're actually back. Your training history answers both, and the second answer isn't the one you'd expect.
You're back in the gym tomorrow. The internet has a number for you: take 10% off. Or take half. Or start at 50–70% and build up. Every article names a number, and almost none of them tell you where the number came from.
Then there's the question that arrives three weeks later, and nobody answers it at all. You finally hit your old working weight again — and it felt terrible. So are you back, or aren't you?
The bar says yes. Something else says no.
That something else is measurable. If you logged anything at all before the break, you're already holding it.
Follow along with your setup
We’ll tailor the connection steps to the way you use Eccentric.
Connect Eccentric to ChatGPT
Add Eccentric as a custom app, then sign in when ChatGPT asks for permission.
- Open ChatGPT Settings, then Apps.
- Turn on developer mode if your plan and workspace make it available.
- Choose Create, enter the Eccentric connection address shown in Eccentric Settings, and scan the available actions.
- Finish the sign-in flow, then enable Eccentric in a new chat.
ChatGPT’s custom-app availability and write access vary by plan and workspace. Check the current ChatGPT instructions before you start.
First, the number for tomorrow
You're in a hurry, so here it is before the reasoning.
Find your last normal week of training — not your best-ever session, not a max you tested once in March, the ordinary week right before things stopped. Take the working weights from that week. Then, for session one:
- Under two weeks off: roughly 85–90% of those loads, and cut a set or two from each movement.
- Two to four weeks: roughly 75–85%, with the same volume cut.
- Four to eight weeks: roughly 65–75%, and expect to spend two or three sessions climbing rather than one.
- Longer than that: treat it as restarting the program, not resuming it. More on that below.
The volume cut matters more than the load cut, and it's the part every "take 10% off" article leaves out. Three sets at 85% will leave you far fresher than five sets at 75%, and the thing you're managing in session one is total work, not peak weight.
Why every article names a different percentage
Because they're all guessing at the same missing number: how much strength you lost.
That's the wrong question. It's why the advice is inconsistent, and it's why following it feels arbitrary.
You probably didn't get weak
Strength is one of the more stubborn adaptations you have. In their review of short-term detraining — defined as under four weeks without an adequate training stimulus — Mujika and Padilla found that while muscle fibre cross-sectional area changes start to reverse, "strength performance declines are limited." The companion review covering layoffs longer than four weeks is where the more meaningful declines show up.
Two weeks off a holiday is not a strength problem. Neither, mostly, is three.
What about much longer? Staron and colleagues followed six women who trained hard for 20 weeks, stopped for seven to eight months, then retrained for six. Maximal strength dropped over the layoff — but not back to where they started, and six weeks of retraining drove rapid re-adaptation. Six people is a small study, worth holding loosely, but it points where the rest of the literature points: what you built doesn't evaporate.
So if you kept the strength, why go light?
The protection is on a different clock than the strength
Lifting a heavy set and recovering from a heavy set are two different adaptations, and they don't decay together.
The relevant one here is the repeated-bout effect: a hard bout of training protects you against the damage and soreness that a similar bout would otherwise cause. McHugh's review describes it plainly — one bout of eccentric exercise measurably protects against muscle damage from later bouts — while being equally plain that after decades of study "the mechanism for this adaptation is poorly understood."
How long does the protection last? Nosaka and colleagues tested exactly that, running two bouts separated by 6, 9, or 12 months. Protection was still detectable at six and nine months. At twelve months, it was gone.
Read that carefully, because it cuts against the version of this story you'll see elsewhere: after two weeks off you have not lost your protection. You've lost some of it, and the longer the gap, the more of it is gone.
Which gives you the actual reason session one is light:
You kept enough strength to lift a load you can no longer comfortably recover from. The strength came back to the gym with you. The tolerance for the work is a step behind it.
That's an asymmetry problem, and the math isn't close. Too light costs you one slightly boring session. Too heavy costs you a week of walking down stairs sideways, a missed session or two, and a return that's now slower than if you'd never rushed it.
Does it matter why you were off?
Yes, and this is the distinction the generic advice flattens completely.
- Holiday or work travel. You were eating, sleeping, and walking around. This is the most forgiving version. The top of each range above is reasonable.
- Illness. You were under-eating, under-sleeping, and systemically stressed, and you may still be carrying some of it. Start at the bottom of the range and add a session to the ramp. A two-week holiday and a two-week flu are not the same two weeks.
- A deliberate deload or planned break. You arrived rested on purpose. Come back nearer the top of the range — that was the point.
- Injury. Different category entirely. Get it cleared by a clinician who has actually looked at you. Nothing in this article is loading advice for a healing structure, and an article can't examine you.
Finding your own numbers instead of a percentage
Everything above is a fallback. The better anchor is in your own training history — Eccentric still knows exactly what you did before you stopped. Start by finding the shape of the gap.
Your assistant can read the dates of your logged sessions, so the gap is a fact rather than a guess — and "the last few sessions before that" matters, because the week before a break is often already unusual. People deload before holidays. People train badly when they're getting ill. You don't want to anchor session one to a week that was itself an outlier.
Then ask for the loads:
That last sentence is the important one. You want the recommendation and its evidence, so you can disagree with it. If it anchored on a session you know was a bad night, you'll spot that immediately — and you're the only one who can.
The question nobody answers: am I back?
Three weeks later you're back at your old weight. This is where lifters get it wrong, and the mistake is subtle enough that it feels like rigor.
Matching the load is not the test. The load is an input. You chose it. You can choose it whether or not you're ready for it, which is precisely the problem — it's the one variable in this comparison that can't tell you anything you didn't already decide.
The real test is: at the same load, what did it cost you?
That's what an effort rating is for. If your squat was 100 kg at RPE 7 before the break and it's 100 kg at RPE 9 now, you are not back. You're doing the same work at a much higher price, and the bar has no way of telling you that.
(If your account is set to RIR rather than RPE, the same comparison runs on your scale — the two describe the same thing from opposite ends: RIR = 10 − RPE.)
This isn't a soft measure. Zourdos and colleagues had experienced and novice squatters rate effort in terms of reps in reserve across loads from 60% to 100% of their max, while measuring bar speed. In experienced lifters, the relationship between average velocity and rating was strongly inverse — r = −0.88 — and still strong in novices at r = −0.77. A trained lifter's sense of how hard a set was tracks something real and physical, closely enough to plan from.
Two practical notes from that same study. Experienced lifters rated more accurately than novices, so this comparison sharpens with training age. And because the relationship was measured on heavy singles and the ends of a working set, the cleanest thing to compare is your top set — not a high-rep back-off, where everyone's ratings get mushier.
Reading it in Eccentric
It's two questions. First, pull up the session before the gap:
Then let it make the judgment it now has the evidence for:
Because Eccentric keeps every set — the load, the reps, the effort rating, and whatever you typed in the notes — those questions have real answers instead of recollections. Your session record already carries the thing your memory is worst at: what actually happened, set by set, including the sets that didn't go to plan.
If you never logged effort before the break
Then this specific comparison is closed to you, and it would be dishonest to pretend otherwise.
Eccentric won't invent it either. If a session trained a movement but no effort rating was recorded, that session is left out of the effort trend and counted separately as a known gap — you get told how many sessions are missing rather than shown a tidier chart than your data supports.
So use what you do have:
Load × reps × sets at comparable form is a blunter instrument, but it isn't nothing — completing 5×5 where you used to complete 5×5, at the same weight, with the same bar path, still means something. And the useful move today is a small one: start logging effort now, so the next gap is readable. Every layoff is followed by another one eventually.
The ramp: three sessions, not three weeks
Once you're back, the return should be short. You're not rebuilding — you're re-establishing tolerance to a load you already own.
For most gaps under a month, three sessions per movement is enough: session one at the reduced load, session two most of the way back, session three at your old working weight. If effort at the old weight lands where it used to, rejoin your normal progression from there. If it doesn't, hold the weight for another session rather than pushing on — you'll pay for it either way, and holding is cheaper.
Have it propose the change, not make it
Changing your plan is a bigger step than answering a question, and Eccentric treats it that way.
When your assistant wants to alter your actual training plan, it doesn't edit it. It puts together the change and sends it to you as a proposal, which you accept, edit, or reject. There's one open proposal at a time, so you're never triaging a queue of half-remembered suggestions.
That's not a promise about how the assistant will behave. It's how the connection is built: your plan is not something it can quietly rewrite.
What coming back too fast looks like in the log
Before it looks like an injury, it usually looks like this:
- Effort ratings climbing at loads that aren't climbing.
- Reps quietly disappearing off the back end of sets — a 5×5 that's really been 5, 5, 5, 4, 3 for two weeks.
- Soreness that's still there when the next session comes around.
- Session notes that have started to read like complaints.
All four are in your training history, and the last one only exists if you write things down.
If your equipment reports bar speed
Only relevant if you train on hardware that measures each rep — most lifters can skip this.
If you have it, you get the same comparison without having rated anything. The Zourdos data above is why: bar speed and perceived effort are two views of one thing. Same movement, same load, slower bar means a higher cost — and the measurement doesn't care whether you were paying attention.
One caveat worth more than it sounds: compare like with like. Speed is only meaningful within the same movement and the same setup, and a handful of sessions is not a trend.
When to restart instead of resume
Somewhere past two months, "picking up where I left off" stops being the right frame — roughly where the detraining research stops describing limited declines and starts describing real ones.
But the tell isn't only the calendar. It's whether the plan still fits the person. A program built around four sessions a week doesn't survive a layoff caused by a new job that took two of those evenings permanently.
Restarting a linear program after a long break isn't a demotion, incidentally. It's the cheapest run of easy progress you'll ever get, and the retraining literature points to you moving through it faster than you did the first time.
What this looks like next time
The uncomfortable thing about this article is that most of it depends on a decision you made months ago: whether you wrote down how hard it felt.
If you did, coming back is a lookup. If you didn't, it's a guess dressed up as a percentage.
So the single most useful thing you can do this week isn't choosing the right load for session one. It's logging the effort rating on your top set, every session, starting now. It takes about four seconds and it's the difference between "I think I'm back" and knowing.
If you're running a specific program, the same thinking applies inside it — our guide to running StrongLifts 5×5 with AI covers what these conversations look like when there's a fixed progression underneath them.
And when you're properly back and the ramp is behind you, the question gets simpler again:
Start using Eccentric with your training →
Sources: Mujika & Padilla, Detraining, Part I and Part II, Sports Medicine (2000) · Staron et al., Strength and skeletal muscle adaptations in heavy-resistance-trained women after detraining and retraining, Journal of Applied Physiology (1991) · McHugh, Recent advances in the understanding of the repeated bout effect, Scandinavian Journal of Medicine & Science in Sports (2003) · Nosaka et al., How long does the protective effect on eccentric exercise-induced muscle damage last?, Medicine & Science in Sports & Exercise (2001) · Zourdos et al., Novel resistance training-specific rating of perceived exertion scale measuring repetitions in reserve, Journal of Strength and Conditioning Research (2016).
Train on purpose.
Eccentric turns what your equipment records into progress you can read. It's free to start.