Brave AI on Lactate Threshold, Orange Theory, Whoop, and Garmin:
Conversation had 5.Sep-2026:
Brave uses a mix of open-weight large language models, primarily Mixtral 8x7B and Llama, with Qwen models (including Qwen3) now powering its flagship Ask Brave feature — all hosted in-house and grounded via retrieval-augmented generation (RAG) over Brave's own search index, rather than relying on a single third-party model.
Q: Are Splat Points by Orange Theory the same as Whoop Strain Scores:
No, Splat Points and Whoop Strain scores are not the same.
Splat Points measure minutes spent in the Orange (84–91% of max HR) and Red (>91% of max HR) heart rate zones during a workout, primarily to track potential post-exercise calorie burn (EPOC).
In contrast, Whoop Strain is a logarithmic score from 0 to 21 that measures the total cardiovascular load your heart worked relative to your estimated max heart rate, accounting for both intensity and duration across all zones.
While both metrics relate to workout intensity, they differ in calculation and purpose:
Splat Points are a linear count of specific high-intensity minutes; only time in the Orange and Red zones counts, while lower-intensity aerobic work (Green zone) contributes nothing.
Whoop Strain is a cumulative load score that incorporates effort across the entire spectrum of heart rate zones, providing a holistic view of daily cardiovascular stress rather than just peak intensity time.
Q: Does either one account for lactate threshold or heat training:
Splat Points are anchored to a fixed percentage of your estimated max heart rate (84%+ = Orange, 92%+ = Red). There's no way to re-anchor those zones to your lactate threshold, and no heat adjustment exists.
Whoop Strain is also built on heart-rate zones defined as percentages of your max HR. A common complaint in the Whoop community is that you cannot set custom zones based on lactate threshold HR — the only adjustment you can make is manually override your max HR value in settings. No heat acclimation factor is applied to the strain calculation either.
So both systems treat a given heart rate the same way whether you're training at 95°F or 55°F, and neither uses your actual lactate threshold as a reference point. If you want a metric that accounts for lactate threshold, you'd need a device like Garmin that lets you build HR zones from a lactate threshold test. For heat, most training platforms (e.g., Coros) recommend manual pace/effort adjustments rather than automatic score correction.
Q: Which devices support lactate threshold HR zones?
Here's the landscape:
Garmin — The most robust support. You can set HR zones to % LTHR (percentage of lactate threshold HR) as the zone base, with auto-detection via Firstbeat Analytics. Available on all current mid-tier+ devices: Fenix 8 series, Forerunner 970, Forerunner 570, Venu 4, Venu X1, Forerunner 965, Forerunner 265, Epix Pro Gen 2, and prior flagships. You can also manually enter a lab-tested LTHR value.
Coros — Provides Threshold Pace and Threshold Heart Rate as part of its training zones framework, with 6 threshold-based pace zones. Its documentation doesn't detail the modelling approach, but it's a first-class feature.
Polar — Offers a Threshold Heart Rate derived from a guided field test scheduled in Polar Flow (not passive auto-detection during regular runs).
Wahoo — Its SYSTM platform uses a threshold concept via the Four Dimensional Power profile, but this is cycling-specific and derived from structured test efforts.
Does NOT support it:
Apple Watch — No lactate threshold estimate; VO2 max is its primary aerobic fitness metric.
Suunto — No discrete lactate threshold value exposed in the app as of early 2026.
So if LT-based zoning is a priority, Garmin and Coros are your two main options for running/multisport, with Polar as a third if you're willing to do a dedicated guided test.
Q: How accurate are Firstbeat Analytics estimates?
Lactate Threshold HR (LTHR)
The most reliable of Firstbeat's estimates.
A 2025 Frontiers in Physiology study (comparing Garmin Forerunner 265, Huawei GT Runner, and Coros Pace 3 to the Dmax blood-lactate method) found:
MAE ≈ 8.9 bpm (MAPE ~6%) for Garmin's LT HR estimate
No statistically significant difference from the lab reference (p > 0.05)
However, none of the devices met strict equivalence criteria (TOST), and individual variability was wide — limits of agreement were broad enough to cause zone misclassification near threshold boundaries
A 2024 Open Access J Sports Med study on the Fenix 7 found HR at LT was within ~2% of a standardized field test (p > 0.05), though the authors noted the small sample size.
Practical takeaway: LTHR is generally good enough for zone-setting in trained athletes, but expect ±5–10 bpm of error and treat it as a trend tracker rather than a precise number.
Lactate Threshold Pace
Much less reliable. The same 2025 study found all three devices significantly overestimated LT pace:
Device MAE (km/h) MAPE
Garmin 2.17 25.8%
Coros 1.93 22.6%
Huawei 1.22 12.7%
A 2024 study found the Fenix 7 underestimated LT pace by ~12% compared to a blood-lactate field test. The inconsistency (over- vs. under-estimation across studies) reflects the fact that pace at LT is heavily influenced by terrain, wind, heat, and fatigue — none of which Firstbeat controls for.
Garmin Fénix 7® Underestimates Performance at the Lactate Threshold in Comparison to Standardized Blood Lactate Field Test
Authors: Heiber M et al
Volume 2024:15 Pages 47—58
https://doi.org/10.2147/OAJSM.S444568
Open access journal of sports medicine
The Garmin® manual does not specify the number of training sessions that must be completed to begin the lactate threshold test. Second, to standardize the validation, we instructed participants to run twice per week for 5 km. 1.) At 70% of the individual’s maximum HR (70% HRmax), and 2.) at 85% HRmax. The use of smartwatches for other exercise activities was prohibited. We calculated the maximum heart rate (HRmax) using the standard sports science formula (220 minus age). Third, after five weeks, they should execute the lactate threshold test with the watch. Because this test required a maximum load run, one of the authors was presented to provide first aid if needed. After activating the GPS, the participants started the lactate threshold test on the Garmin smartwatch as follows:
From the watchface, press START
Navigate to the activity outdoor running,
Hold MENU
Select training
Select lactate threshold test > start test following the instructions on the screen.
During the test, the watch displayed both current and intended target HR. The task was to run the given intervals at each displayed target HR for a specific time. After completing the test intervals, the participants were notified by the vibration signals from the watch. They had to stop the timer and save the activity. Subsequently, the watch allows users to update their HR zones for future training based on the ascertained lactate threshold HR. However, users can always decide whether to accept or reject the latest lactate threshold calculation. In this study, all the participants were instructed to accept the calculated values. Fourth, the lactate threshold values, lactate threshold pace, and lactate threshold HR determined by the Fenix 7® were compared with a graded test in the field using blood lactate taken from the earlobe ≥ 48 hours later.
The modified Dmax method was used to analyze the lactate threshold. The method contains:
The first increase in lactate (> 0.4 mmol/l compared to the previous stage) served as a rule for determining the aerobic threshold
The point on the lactate curve at which the aerobic threshold is determined is related to the lactate termination value.
Next, we determined the point on the lactate curve with the maximum perpendicular distance to the connection between the two points.
Consequently, the anaerobic threshold corresponds to the load at the point determined in step 3.
We found that the pace at the LT calculated by the Garmin Fenix 7® (M 11.87 km/h ± 1.26 km/h) is 11.96 % lower than the pace at the LT obtained during the field test (M 13.28 km/h ± 1.72 km/h). Similarly, the HR acquired by the Garmin Fenix 7® at LT (174.42 bpm ± 4.62 bpm) is 1.71 % lower than the HR at LT calculated with the Dmax Method (177.42 bpm ± 9.99 bpm). We observed a higher variance in pace and HR at LT for the data gathered during the lactate field test.
If the deviations in LT pace and HR are constant, potential training adaptions, performance gains, or losses can still be detected over time. However, considering a competitive or professional athlete, the same estimation error may have more extensive consequences. The pace at LT was, on average, underestimated by 11.96% by the smartwatch compared to the field test. For example, at an actual LT pace of 4:00 min/km, this deviation can be as high as 28s/km. This distinct underestimation might lead to the permanent application of insufficient or inadequate training stimuli, which would prevent athletes from achieving their best possible performance enhancement within a given training period. In contrast, a potential overestimation of LT pace or HR could lead to unplanned fatigue or over-training symptoms.
https://www.gneta.app/blog/garmin-lactate-threshold-explained
Q: "You should train at lactate threshold to improve it"
Partly true, but incomplete. Training at threshold intensity does improve your lactate threshold, but so does high-volume zone 2 training (which develops the aerobic base that raises the threshold from below) and high-intensity interval training (which improves your lactate clearance capacity). A well-rounded program includes all three.
Q: "Garmin's lactate threshold is the same as FTP in cycling"
Related but not identical. FTP (Functional Threshold Power) is defined as the power you can sustain for approximately one hour, which corresponds closely to lactate threshold for most athletes. But FTP is a performance measure (watts) while lactate threshold is a physiological measure (blood lactate concentration at a given intensity). They usually align well but can diverge, particularly in athletes with unusual metabolic profiles.
The key principles:
Use the guided test rather than relying solely on automatic detection. Control the data quality.
Trust the LTHR more than the pace. Heart rate at threshold is more stable across conditions.
Set your zones based on LTHR, not max heart rate. The zones will be more accurate and more useful.
Retest every 8-12 weeks during structured training. Watch for the trend, not individual readings.
Cross-reference with running power zones if your watch supports running power. Power and heart rate together give the most complete picture.
Changda Lu
Front Physio 2025
https://doi.org/10.3389/fphys.2025.1621996
Validity of smartwatch-derived estimates of lactate threshold heart rate and pace compared to graded exercise testing
Results: The success rates for a single test were 78% for Huawei®, 65.22% for Garmin®, and 47.06% for Coros®. For LT HR, no significant differences were observed between smartwatch and DmaxMod estimates across all devices (p > 0.05). The Huawei® watch showed MAE = 10.66 bpm, MAPE = 6.32%; Garmin®: MAE = 11.44 bpm, MAPE = 7.15%; Coros®: MAE = 8.93 bpm, MAPE = 5.95%. Corresponding Pearson correlation coefficients ranged from r = 0.13 to 0.67, and R2 values ranged from 0.02 to 0.45. In contrast, LT Pace predictions demonstrated significant overestimation for all devices. Huawei® reported the smallest error (MAE = 1.22 km/h, MAPE = 12.70%, p = 0.01, r = 0.88, R2 = 0.78), followed by Garmin® (MAE = 2.17 km/h, MAPE = 25.78%, p < 0.01, r = 0.73, R2 = 0.53), and Coros® (MAE = 1.93 km/h, MAPE = 22.63%, p = 0.08, r = 0.79, R2 = 0.62). Bland–Altman plots confirmed systematic biases and variable agreement patterns, particularly for LT Pace
Q: Lets bring in DFA-alpha-1, Garmin has it, why not use that more often, even exclusively, for LT heart rate and LT pace, compared to the other methods described above?
Explored DFA-a1 limitations for Garmin lactate threshold tracking
Great question, and it's a fair one — DFA-a1 is arguably the most physiologically "honest" threshold signal available to a consumer. But there are several practical and methodological reasons it hasn't (and probably won't) replace HR-based methods as the primary engine:
1. It's a different threshold than most people mean by "LT"
This is the biggest confusion. DFA-a1's 0.75 crossover maps to the aerobic threshold (VT1, ~2 mmol/L), and 0.5 maps to the anaerobic threshold (VT2, ~4 mmol/L). The evidence base is strongest for the aerobic threshold. Most of the "lactate threshold" work that Firstbeat does is anchored to the anaerobic threshold (VT2/OBLA), where the evidence for DFA-a1 is thinner and more recent. So it's not a drop-in replacement for what Garmin's LTHR actually represents.
2. It requires a chest strap
This is the hard blocker. Wrist-based PPG cannot produce the RR-interval quality needed for reliable DFA-a1. You need a Bluetooth strap like a Polar H10 or Garmin HRM-Pro. For the vast majority of Garmin users who train with just their watch, this metric simply isn't available. Firstbeat's HR-based approach works with the built-in optical sensor, which is a massive usability advantage.
3. It needs sustained, controlled effort to establish
To get a clean threshold reading, you need several minutes of steady-state or slow-ramp exercise at a given intensity. AI Endurance specifically requires at least 4 minutes of continuously declining DFA-a1 past 0.75 for the aerobic threshold, and 6 minutes past 0.5 for the anaerobic threshold.
This makes it:
Poorly suited for interval training or variable efforts
Less useful for passive auto-detection from random daily runs (which is how Firstbeat's LTHR updates)
More of a structured test than a background metric
4. It gives you a threshold HR, not a threshold pace
Even with a perfect DFA-a1-derived threshold HR, you'd still need to correlate that HR to a pace, which reintroduces the same terrain/heat/wind/fatigue problems that plague Firstbeat's LT pace estimates. DFA-a1 solves the HR side of the problem but not the pace side.
5. It's heat and hydration sensitive (arguably a feature, arguably a bug)
DFA-a1 reflects your actual autonomic state at the moment of measurement. That means heat, dehydration, caffeine, sleep debt, and illness all shift the crossover point. This is both its strength (it tells you where you actually are today) and its weakness (the number isn't stable across conditions the way a "training" threshold should be for zone-setting purposes). Firstbeat's model tries to produce a trait-like threshold that's consistent across sessions; DFA-a1 is more of a state measurement.
6. Garmin is using it — just not as the primary method
Garmin records in-activity HRV in .fit files, and there's an alphaHRV data field in beta that reports DFA-a1 in real-time on Garmin watches and bike computers. The 2022 Frontiers in Physiology update paper (Rogers et al.) explicitly discusses this. So Garmin isn't ignoring it — they're treating it as a supplementary data field alongside Firstbeat's primary estimates, not as a replacement.
Rogers B, Gronwald T
Front Physio 2022
https://doi.org/10.3389/fphys.2022.879071
Fractal Correlation Properites of Heart Rate Variability as a Biomarker for Intensity Distribution and Training Prescription in Endurance Exercise: An Update
Where DFA-a1 wins: If you already wear a chest strap, do structured ramp tests, and want a personalized, physiologically grounded threshold that reflects your actual autonomic state on the day — it's arguably better than Firstbeat's estimate.
Where it can't win: As a universal, passive, background metric for 50 million Garmin users who train with just a wrist sensor. That's what Firstbeat was built for.
The ideal future setup is probably both: Firstbeat for passive daily zone updates, with DFA-a1 from a chest strap used periodically (monthly or after a big fitness change) to calibrate and validate the Firstbeat estimate.
AI Endurance and Runalyze are already moving in that direction:
https://aiendurance.com/blog/dfa-alpha-1-app-to-track-your-heart-rate-variability-thresholds
https://blog.runalyze.com/features/aerobic-threshold-estimation-based-on-hrv/