Version 1.5 released! Try out the new algorithm

Smoke Point has been updated to v1.5 for iPhone and Android. In the new version, there are visual changes to how readings are displayed, as well as an early version of the new target prediction algorithm.

Visual changes to reading display

Reading display has been updated to more clearly indicate the quality of the reading. Quality here means how likely the heading of the line on the map is to the real heading from the observer’s position to the target. Since mobile phone compasses are sensitive to interference, these headings are not always accurate. Sometimes, they can be wildly inaccurate. This is called “compass error”.

Early results from the Smoke Point data collection project, the goal of which is to collect real-world compass error values, have been integrated into the new visual display of readings.

Previous version: a gray “error beam” surrounded the line representing the compass direction. The error value was derived from the phone’s self-reported compass error. Ironically, this error value was itself wrong most of the time.
New version: the gray beam is gone. For readings that don’t intersect, confidence bars are displayed. The colors red/orange/yellow indicate a 66%/95%/99% chance that the real heading of the target is within that color.
New version: short, thick lines indicate low confidence (and high compass error).

New version: long, thin lines indicate higher confidence (and low confidence error).

To further visualize the accuracy of a reading, a reading’s detail view shows the accuracy of that reading on an accuracy gauge, like this:

Preview of the new algorithm

We’re excited to bring to you a preview of the new target prediction algorithm. The development of this algorithm was made possible by some excellent mathematicians who have volunteered their time and expertise. This version outperforms the existing algorithm in most cases, and importantly, shows confidence ellipses on the map indicating zones of increasing target likelihood.

An intersection of 3 readings. Each ellipse represents a different confidence that the real target is in that zone. The red dot is the target prediction. The dark orange zone = 66% likelihood, medium orange = 95%, and yellow = 99%.
The red/yellow/green meter (labeled “Accuracy”) displayed in the target information box, tells you how large of an area you must search to find the target. A large area will put the needle on red, while a smaller area will put it on green. Adding more readings makes the search area smaller.

Why is it a preview? Good question. For one, the error values that are used to calculate confidence zones are not final. Second, we would like to perform more quality testing on the new algorithm. There are some cases where the new algorithm is not able to calculate a target location at all. I’ll write more about how the algorithm works in a future post.

To use the new algorithm in your app, go to Settings (on the main screen) > Advanced Settings > Use new target prediction algorithm:

Other improvements

  • UI tweaks
  • Opening an incident zooms to it during loading
  • Automatic crash reporting
  • Scale the UI down for smaller phones
  • Ask fewer post-reading survey questions

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: