How we write and where our numbers come from

Who is responsible for these pages

I'm Arkadiusz Chrabąszczewski, the creator of SmartWorkout. I run the knowledge base and approve every Polish text. I have been programming for about 10 years, have lifted for years and keep up with strength training research. Our advice rests on research and on training log data, not on what worked for one person, so every claim about what works carries a citation. More about me and why I built the app is on the author page.

How an article is made

Every article goes through the same 6 steps.

1. We collect the questions people type into search engines about the topic and check what the existing answers leave out.
2. For every claim we record the paper, its exact sentence and its PubMed ID or DOI in a source sheet.
3. We write the text following our style guide and a list of facts about the app.
4. We check every number separately: against the sentence in the paper, the app's code or the result of a data query. Plan fragments are pasted into the app's plan import and compared with the preview.
5. Editing removes filler and keeps the terms straight: a set is a set, tonnage is weight times reps, and volume is the number of sets.
6. The author reads the whole Polish text next to the source sheet and approves it or corrects it.

How we check numbers

The author reads and approves every Polish text, and we check every number in the source or in the app's data. A number we cannot point to in a paper, in the app's code or in a query result is cut, however plausible it sounds.

How we choose and read research

Claims about what works, how big the effect is and how tiring it is rest first on meta-analyses and position stands of scientific organizations, and only then on single studies.

We check every number from a study in the paper itself, in the full text or in its abstract. We take no numbers from other people's summaries or from memory; if we can't find the number in the paper, we leave it out. Every citation leads to PubMed, PMC, a DOI or the official page of a position stand, and the full source list sits at the bottom of every page.

In the "What the research says" section we give the type of study, the number of participants and the duration, and we end with one sentence on what the research does not show yet. Statistics such as effect sizes or confidence intervals appear only there; in the other sections we describe the result in words, with one plain number.

Advice without a citation is our practice

A tip in the steps, examples, exercise tables or common mistakes that has no citation is our practice: a choice that fits the research results but is not itself a research result. That is how we pick, for example, the exercises for examples and the kilograms in tables. Where our choice could easily be mistaken for a research finding, we say so: the "Source" column of the key numbers table then reads "our practice".

We state SmartWorkout rules as rules and always give the reason. We count a whole drop set as 1 set, because there is no rest between the drops; the reasoning and sources are in the article on how to count sets. When our advice goes further than the research or against it, we say so in the same place.

Where the SmartWorkout log data comes from

The "SmartWorkout data" boxes show how people who log their workouts in the app actually train. We compute them with read-only queries, usually over the last 90 days, and run the same set of queries once a year; the period is stated in every box.

The numbers in the boxes therefore describe mostly men from Poland who train in a gym, not an average of everyone who lifts. Where women train clearly differently, as with volume, we give their results separately.

Definitions in the data boxes

Every box uses the same definitions, and its "Sample" line says which group it covers.

In the wiki's advice a working set has one more condition: it ends with at most 4 reps in reserve (which sets count as working sets). We cannot check that in the logs because only a minority logs RIR, so log data counts sets more broadly than our advice does.

Privacy: aggregates only

Only aggregate numbers reach the site: percentages, medians and the ranges that hold the middle half of people. Every table row and every cell with a percentage covers at least 50 people; we do not show the result of a smaller group at all, even when it would be interesting.

We do not publish single workouts or plans, names, account names or exercises that users created themselves. We exclude deleted workouts and our own internal and test accounts. In numbers about plans we take only plans that people build themselves in the plan creator: no ready plans from the library and no coach account templates, because most of them are the same sample plans.

What log data cannot show

Log data describes how people train. Conclusions about cause and effect, meaning what works, come only from research: the sentence "in workouts with supersets the same person needed less time per working set" describes an association and does not mean that supersets by themselves shorten a workout.

• A log is a record. If someone does not mark warm-ups, drop sets or tempo, the data does not have them, and warm-ups logged as normal sets inflate the working sets.
• A minority logs RIR and tempo. These numbers describe the people who log them, usually more engaged than average.
• Templates on the query date. We read the targets and rest times in templates as they are on the day of the query, not on the day of the workout.
• Strength gains are counted from the first entry. The first time an exercise is logged in the app is not the start of someone's training, and people who stopped logging drop out of the comparison.

Versions in other languages

Versions in other languages are translated from the Polish version the author approved; this English page is one of them. Facts, numbers, sources and data boxes are the same in every language; what changes are the example questions, the exercise names from the atlas in that language and the app's button labels. In the German or Spanish version too, the log data comes mostly from Poland, and we describe it that way.

Updates and corrections

Every article shows the date of its last update and, when something important changed, one sentence on what it was. The tables of volume zones, the RIR to RPE conversion and the exercise group names are not copied into articles: the site builds them from one place in the code with the same rules as the app, so a change to a rule changes them on every page at once. We review sources once a quarter and check feature descriptions against the app version available on the App Store and Google Play on the day of the update.