When Will You Hit Your Next PR? The Math Behind the Prediction
<em>How WodPilot uses your lifting history to forecast your breakthrough—and why the math only works when you're ready.</em>
You're three weeks into a solid training block. Your squat felt fast last week. Your bench press numbers are climbing. And somewhere in the back of your mind, you're wondering: when is the next PR actually going to happen?
It's not just curiosity. Knowing when a breakthrough is likely changes how you approach the next few weeks of training. It shifts your mindset from "I hope this happens" to "this is the plan." It helps you decide whether to push hard or dial back before a meet. It tells you whether you're on track or if something needs to adjust.
The problem is, most athletes guess. They feel their lifts are improving, extrapolate that feeling forward, and hope for the best. But your body leaves a trail of data—every lift, every session, every attempt. That data has a pattern. And patterns can be predicted.
The Linear Regression Model: Reading the Trajectory
A PR doesn't happen by accident. It happens because you've been getting stronger in a consistent direction. That direction—the slope of your strength curve—is measurable.
WodPilot uses linear regression to analyze your last six personal records and find the trend line that best fits your data. Linear regression is a statistical method that finds the straight line that minimizes the distance between your actual data points and the predicted line. In lifting terms: it answers the question "if this trend continues at this rate, when will you hit the next milestone?"
Linear regression requires a minimum of 3 data points to establish a trend. WodPilot uses your last 6 LiftPrHistory records to calculate the slope (rate of strength gain per day) and intercept (your baseline). The model's reliability is measured by R², where R² >= 0.85 indicates a high-confidence prediction, R² >= 0.50 indicates medium confidence, and R² < 0.50 indicates low confidence. The higher the R², the more consistent your progression has been.
Think of it this way: if your last six squat PRs were 405, 410, 415, 420, 422, 425 over the past 12 weeks, the regression line captures that upward trend. The slope tells you how many pounds per day you're typically gaining. If that slope is +0.5 lbs/day, the model predicts the next PR when you've added another 1% to your current max.
But here's the catch: linear regression only works when your data is actually linear. If your progression is erratic—huge jumps followed by plateaus, or recent attempts that broke the pattern—the model loses confidence. That's not a flaw. That's the model being honest about uncertainty.
Every time you log a new PR, WodPilot recalculates the regression on your six most recent records. It computes the slope (your rate of progression) and projects forward to find the date when your predicted strength gain will equal a 1% increase from your current one-rep max. This becomes your "days to next PR" estimate. The R² value is calculated and compared against the confidence thresholds to determine whether to show you a high, medium, or low confidence prediction on your athlete hub.
A high-confidence prediction tells you your progression is consistent and predictable—you can trust the timeline and plan your training blocks accordingly. A low-confidence prediction signals that your recent data is noisy, which usually means you're either in a phase of high variability (trying new attempts, recovering from fatigue, or testing different rep ranges) or your progression has stalled. Both are valuable signals, but they require different responses.
The 1% Rule: Why One Percent?
The next PR target isn't arbitrary. It's built on a principle from strength sports: meaningful progress is incremental.
In elite powerlifting and weightlifting, a PR is typically defined as any weight that exceeds your previous one-rep max. But for programming purposes, coaches often think in terms of minimum viable increments. A 1% increase is the smallest jump that registers as real progress without being so small it's noise, and without being so large it requires a disproportionate effort spike.
The next PR target is calculated as current_1rm × 1.01. For example, if your current squat PR is 400 lbs, the next predicted PR is 400 × 1.01 = 404 lbs. This 1% increment aligns with the principle of progressive overload: small, consistent increases in load over time produce strength adaptation without excessive fatigue accumulation.
Why not 2%? Because 2% jumps are harder to hit consistently, and they compress your training timeline. Why not 0.5%? Because at that threshold, measurement error and day-to-day variation become noise—you can't distinguish real progress from fluctuation.
The 1% target also respects individual differences. A 1% jump for a 400 lb squat is 4 pounds. A 1% jump for a 200 lb squat is 2 pounds. The model scales automatically to your current strength level.
When the regression model calculates your trajectory, it projects forward until your predicted one-rep max reaches current_1rm × 1.01. The number of days between now and that date becomes your "days to next PR" estimate. This target is fixed; the timeline adjusts based on your recent rate of progression.
When the Model Stays Silent: The Prediction Holds
There are moments when WodPilot won't show you a PR prediction, even if you have enough data. These aren't failures—they're guardrails.
The model returns None (no prediction) in three scenarios:
- You're in a deload or taper. These phases are intentionally designed to reduce load and volume. Your lifting numbers will naturally dip. Projecting forward from a deload is meaningless because you're not following the normal progression curve.
- Your recent trend is negative. If your last few attempts show declining numbers, the regression slope is pointing downward. The model recognizes this and stops predicting—because predicting a PR from a downward trend would be false confidence. Instead, this is a signal to investigate what's happening: fatigue, technique breakdown, life stress, or a genuine need to reset.
- You have fewer than three data points. You need a minimum of
3 LiftPrHistory recordsto fit a line. With fewer, there's no trend to measure.
Linear regression requires minimum 3 data points to establish a trend. The model explicitly checks for deload/taper phases and negative slopes before returning a prediction. If either condition is true, days_to_pr = None. This prevents false confidence in scenarios where the underlying assumptions (consistent progression, normal training state) are violated.
When the prediction disappears, it's not because the system broke—it's because the conditions that make prediction meaningful have changed. A deload needs to happen; a negative trend needs investigation. Respecting these pauses is how you stay healthy and keep progressing long-term. The prediction will return when you're back on the curve.
Confidence Tiers: Reading the Signal Strength
Not all predictions are created equal. Some are based on rock-solid data; others are educated guesses.
R² (R-squared) measures how well the regression line fits your actual data points. It ranges from 0 to 1. An R² of 1.0 means the line passes through every single point—perfect fit. An R² of 0.5 means the line explains half the variance in your data; the other half is noise. An R² of 0.0 means the line explains nothing.
WodPilot classifies prediction confidence into three tiers: R² >= 0.85 = high confidence, R² >= 0.50 = medium confidence, R² < 0.50 = low confidence. These thresholds reflect the statistical reliability of the underlying regression. A high-confidence prediction means your recent progression has been consistent and predictable. A low-confidence prediction means your data is scattered, and the timeline should be treated as a rough estimate, not a guarantee.
What creates high confidence? Consistency. If your last six PRs came at predictable intervals with steady strength gains, the regression line fits tightly, and R² climbs. What creates low confidence? Variability. If you hit a PR, then took a week off, then hit another, then deloaded, then came back strong—your data points jump around, and the line doesn't fit well.
Low confidence doesn't mean you're not making progress. It means your progression isn't following a smooth curve right now. That's normal during transitions between training blocks, after injuries, or when you're experimenting with new programming.
The confidence tier is displayed alongside your PR prediction on the athlete hub. High-confidence predictions are shown prominently because they're reliable guides for planning. Medium-confidence predictions are shown with appropriate context. Low-confidence predictions are shown but flagged as uncertain, so you don't make training decisions based on a noisy signal.
Confidence tiers help you separate signal from noise. A high-confidence prediction is worth planning around—adjust your meet prep, your volume, your intensity. A low-confidence prediction is worth noting but not betting the farm on. It tells you to focus on execution and consistency first, and let the timeline reveal itself as your data becomes clearer.
The Feedback Loop: How Better Data Improves Predictions
Every time you log a new PR, the model gets smarter. Your six-record window slides forward. The regression recalculates. The slope updates. The confidence adjusts.
This is why consistency matters. If you log PRs on a regular schedule—every 3-4 weeks during a strength block—your data becomes increasingly predictable, R² climbs, and the model's timeline becomes more reliable. If you log PRs sporadically—months apart, or multiple times in one week—the data is noisier, and the prediction is less certain.
The model isn't judging your training. It's reflecting what your data shows. Frequent, consistent PRs signal a well-structured program with good progression. Sporadic PRs signal either a program that doesn't prioritize regular testing, or an athlete who's still finding their rhythm.
WodPilot's job is to make that pattern visible so you can decide: Am I happy with this progression rate? Do I need to adjust my program? Is this the right time to peak, or should I extend the block?
The Bottom Line
Your next PR isn't magic. It's math. It's the inevitable result of consistent training applied to your body's capacity for adaptation. Linear regression reads that pattern in your data and projects it forward, giving you a realistic timeline based on how you've actually been progressing—not how you hope to progress.
The prediction works best when your training is consistent, your data is clean, and you're in a normal training phase. It steps back when conditions change, because that's when honest uncertainty matters more than false confidence.
The real value isn't the number of days. It's the signal: you're on track, or you're not. Everything else is coaching.
Ready to train with real science?
WodPilot uses the models described in this article to personalize your training every day.
Get started