Running index should be graphed over time

[thomasturrell]

Personally I find my running index over time one of my most useful metrics when assessing my long term progression.

An individual running index is of little value to me but the moving average is tremendously useful.

This is an example of a running index report in Polar flow.

Incidentally, I believe that running index in elevate should be renamed. It is not a the same algorithm as Polar's running index and it would be a shame if people got confused because the two algorithms have the same name.

[thomaschampagne]

Hi @thomasturrell, could you provide me (by a private way) an export of your files (option 2 on this strava link)?

I'm currently tackling with @tsourbier the following issue: #871

In version 7, it should be renamed as "Running Performance Index", "Running Scale" or "Running Ratio" or better idea?

Thanks for your help,

Thomas

[thomasturrell]

Glad to be of help. Elevate is a very useful training tool.

Some naming suggestions:

• Running Rating or Run Rating
• Running Score or Run Score

I like the literation in Run Rating, so that would be my preference. Elevate should use a name that isn’t used by anyone else.

I will gladly share my runs with you. I will message you privately.

[tsourbier]

I'll side with "Running Rating" it is perfect imho to encompass the fact it is the fruit of some calculations and allows to compare runs & runners.

[thomasturrell]

@thomaschampagne I have sent you DM in twitter.

[thomaschampagne]

@thomaschampagne I have sent you DM in twitter.

I replied you ;)

[thomaschampagne]

Let's go for "Running Rating" then. I have a first result that compares both source. @thomasturrell Let my know if I can post it here.

[thomasturrell]

@thomaschampagne no problem, please post the comparisons here. I am expecting my running rating to have less spread than my running index.

I am looking forward to seeing the results.

[thomaschampagne]

Here you go on your 2021 activities:

• Matplotlib + scipy

• Into a desmos graph: (https://www.desmos.com/calculator/jxrpl8maq4)

[thomaschampagne]

The CSV polar-vs-elevate-RI.zip

[thomasturrell]

That is really interesting. Elevate does a really good job.

As I suspected, Elevate results do not have as much spread (which is a good thing).

@thomaschampagne there is one surprising result, data point 94 is higher but I can not see an obvious reason why.

[tsourbier]

@thomasturrell as we discovered with my data 2 issues could be at play:

• bad HR data that Polar was able to filter out but not Elevate.
• bad Elevation data that Strava corrects when you look at it online but the error can still be present in the export.

may be looking into that training deeper will reveal some other oddities 😂

T.

[thomasturrell]

• bad Elevation data that Strava corrects when you look at it online but the error can still be present in the export.

@thomaschampagne I have seen this in your results. I did a 10K race with significant elevation gain, Strava does a good job of correcting it. However most watches struggle to give accurate elevation data from GPS.

I also suspect you are right about the 'bad' heart rate data. But filtering that out would be very hard. I notice that even Strava doesn't have an algorithm for that. Stava does not spot people who accidentally mark a bike ride as a run (trivial for a human to spot, difficult for an algorithm).

I think that people should be able to remove a running rating for certain runs in Elevate. I don't want my anomalous readings to affect my moving average but I also do not want to delete a run with bad heart rate data (or bad GPS). Having said that, with enough activities it will average out.

[thomaschampagne]

@thomasturrell @tsourbier I implemented a Flag mecanism able to detect flaws in speed, power, heartrate... For example if activity is flagged with a strange HRSS then his score will not count for Fitness Trend (we use another valid one or none). I can do the same for running index/rating...

However most watches struggle to give accurate elevation data from GPS.

If you have an altimeter that's OK. But indeed it's complicated for Elevate without. It's fixable using a Kalman filter but I have to know the measurement errors of the altitude signal first (get the standard deviation on a 100m flat section ?!).

The "alone point" is the below activity. And indeed altitude is broken:

@thomasturrell Do you want a dev build including the current "Running Rating"?

[thomasturrell]

@thomaschampagne A dev build would be great. I have a Mac (if that makes any difference).

[thomaschampagne]

@thomasturrell I sent you a dm on twitter

[thomaschampagne]

Results with new assertions before computing Running Rating:

  /**
   * Computes equivalent Polar running index
   * https://support.polar.com/en/support/tips/Running_Index_feature
   * https://www.polar.com/en/smart_coaching/features/running_index/chart
   */
  public static runningRating(
    athleteSnapshot: AthleteSnapshot,
    stats: ActivityStats,
    timeArray: number[],
    distanceArray: number[],
    velocityArray: number[]
  ): number {
    if (
      !stats.movingTime ||
      !stats?.heartRate?.avg ||
      !stats?.pace?.gapAvg ||
      !athleteSnapshot?.athleteSettings?.maxHr ||
      _.isEmpty(timeArray) ||
      _.isEmpty(distanceArray) ||
      _.isEmpty(velocityArray)
    ) {
      return null;
    }

    // Verify running rating requirements
    const COOPER_KPH_THRESHOLD = 6;

    // Verify legit ascent speed to get proper grade adjusted distance value
    if (Number.isFinite(stats?.elevation?.ascentSpeed) && stats?.elevation?.ascentSpeed > this.MAX_ASCENT_SPEED) {
      return null;
    }

    // Verify avg heart-rate against athlete current max hr. Should not be higher!
    if (Number.isFinite(stats?.heartRate?.avg) && stats?.heartRate?.avg > athleteSnapshot.athleteSettings.maxHr) {
      return null;
    }

    // Verify standard deviation. Missing variation means heart rate problem
    if (Number.isFinite(stats?.heartRate?.stdDev) && stats?.heartRate?.stdDev === 0) {
      return null;
    }

    // Ensure to have legit hr effort using HRR%
    if (
      (Number.isFinite(stats?.heartRate?.avgReserve) && stats.heartRate.avgReserve < 52) ||
      stats.heartRate.avgReserve > 98
    ) {
      return null;
    }


    // Verify that athlete ran at least 12 minutes over 6 kph. If not, no running performance index
    const bestCooperSpeed = ActivityComputer.computeSplit(velocityArray, timeArray, 60 * 12) * Constant.MPS_KPH_FACTOR;
    if (bestCooperSpeed < COOPER_KPH_THRESHOLD) {
      return null;
    }

    // Compute the grade adjusted speed in meters per seconds
    const gradeAdjustedMetersPerSec = Movement.paceToSpeed(stats.pace.gapAvg) / Constant.MPS_KPH_FACTOR;

    // Keep the greatest adjusted distance
    const gradeAdjustedDistance = Math.max(gradeAdjustedMetersPerSec * stats.movingTime, _.last(distanceArray));

    // Now compute the running performance index
    const distanceRate: number = (213.9 / (stats.movingTime / 60)) * (gradeAdjustedDistance / 1000) ** 1.06 + 3.5;
    const intensity: number = Math.min((stats.heartRate.avg / athleteSnapshot.athleteSettings.maxHr) * 1.45 - 0.3, 1);
    return _.round(distanceRate / intensity, ActivityComputer.RND) || null;
  }

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically closed because it didn't had recent activity.