For further reading about the idea you can look up load velocity profiles.

Additionally, you can compare your warmup speed with the line of best fit to see if you’re having a stronger or weaker day than normal. And if you’re consistently above the best fit line (which should be based on a rolling time window) then you know you’re getting stronger!

Look at the “moved like” estimate of %1RM. That’s based on a much better approach than the linear method in the journals where they have limited and selective data as well as very controlled conditions.

What’s the method used for that? I find the estimates 1rm highly inaccurate and variable (especially at weights below 70%ish) so I assumed the percentage would be off too

It’s based on tens of thousands of tracked sets and some machine learning. The issue you see is probably just an exaggeration of error caused by multiplication. For example, suppose you do a set with the empty bar and the app estimates that was 4% of your 1RM. The real value is 5%. Pretty close, right? It’s just off by 1%. But that leads to big differences in e1RM (500 vs 400). This problem lessens as the weights increase. At 350, 79% vs 80% is 443 vs 438. The main problem with the linear approach is there is no real clear answer to what your terminal velocity is. You’d need to guess. Plus, things are only really linear at heavier weights and your data will be taken from different sets on different days (e.g. one great set at 80% two weeks ago + one great set at 65% last Friday). The result isn’t a great indicator of where you are as things could have changed. You could also get good at one particular velocity and not the other. That would change the slope of the curve and your e1RM, but in reality, that may not be true.

The multiplication issue makes sense.

As far as linearity, I’ve logged a few days of data in excel and have seen pretty much a linear relationship between load and velocity (R^2 = .99 for one days data and .98 for 2 days). I’m going to keep that up for a bit just because I like analyzing data. Being able to see best rep velocity for each set would be useful for that project.

I agree that minimum velocity would be unclear but how does your model get around that? Does it just look at sets by individuals with known e1rms? If you have a moved like metric then whatever velocity moved like 100% is the minimum velocity. Also, minimum velocity is also a useful metric of how well someone can grind so tracking it over time can be useful as it should go down.

Terminal velocity and e1RMs are predicted using more than just a single variable (velocity). You’ve also got ROM. Shorter ROMs are more tolerant to lower terminal velocities. You’ve also got the weight used, which can be somewhat of a proxy for experience/proficiency. You’ve also got your RPE, which adds to the picture. Those are just a few of the inputs but you can see how they’d help determine where that e1RM should be.

Try the latest update. It has better-performing 1RM estimates for more advanced lifters. If you find them working better for you, one thing you could do is disregard the green/red light days and go off of e1RMs on your first three worksets above a 1 in difficulty. If you see e1RM increases across those first 3 sets, do a 4th set. If that’s higher still, do a 5th, and so on. You’ll likely wind up with mostly days of 3 sets and a few days with 4 and 5 here and there. That’s your monthly distribution of stimulating work based directly on performance instead of estimated indirectly through wellness.