Statistics about StreetComplete usage - how many elements were edited by each quest? #1749
Comments
|
What is the difference between "Add Sidewalk" and "Add Sidewalks"? |
|
Looking at just usage - likely AddSidewalks is the old name, used during developement (looking at git history may be able to confirm this). It is possible that this name used by quest in some private fork that is also using StreetComplete changeset tags. It may be investigated by finding specific changesets that used this tags. |
|
Nice! I am balancing when to award certain achievements. For the achievement that is awarded for supplying the road surface, it makes sense to award it later (100s of quests solved) while for things like pedestrian ferry access, it almost makes sense to hand it out after the first one solved. Another information I'd need is the median (not average) of how many quests a user solved. Then I can calculate how many of each quest type a user typically solved. |
That's pretty nice! Didn't expect that much, good job @matkoniecz , this quest is very valuable for keeping the map up-to-date! So the aliases because of renamed quests are: AddAccessibleForPedestrians -> AddProhibitedForPedestrians |
I was thinking about what kind of pool of users should be taken. All who solved any quest? All who solved any quest of this type? All who solved >10 quests? |
|
"All who completed one session" would be a good measure. So perhaps >=3 or so. |
|
SC has 20430 accounts with >=3 edited elements. For nearly all quests median is 0, given sample of users who made >=3 edits. See https://gist.github.com/matkoniecz/f5ebdae67317737f008335636eb7c3d3 AddRoadName - median 1 Everything else - median 0. I suspect that giving some message "It is not only AddRoadSurface app" may be helpful. I suspect that large number of users have never solved any quests except AddRoadSurface. |
|
I am now running script for statistics with users who edited >=3 elements except road surface quest. EDIT: turned out to have a minimal impact - see https://gist.github.com/matkoniecz/431571939bac223b0770fce02fcdef64 But note that about 3000 of 20 000 accounts solved only road surface quests. |
|
Oh, ok |
|
I guess we need another filter to only return users which are really using SC. For example, in the editor usage stats, SC appears consistently with about 7k users, not 20k. |
|
Even after counting median only among people who solved at least 1 quest of given type it is really low - see https://gist.github.com/matkoniecz/11ef437c8a6a4f3edc756831ee3d1bdb https://gist.github.com/matkoniecz/231c5afce9c6b943ada4493d7aafbb28 is counted with median among top 6000 editors https://gist.github.com/matkoniecz/2f25f82a4362650cf48852da3dbb3f10 median among top 3000 editors |
How is the number of people sometimes higher than the number of users? |
|
In this case I manually override number of users by specific value (here 3000). In other words, I put threshold to be counted as user higher than "solve some quests". "Active users" would fit better there, but it was anyway repeated run of hackish script. |
|
Hmm, do you also know the median edit count (all quests summed up) for a user? (median of +top 3000, +top 6000)? |
|
Looking at the pivot table output - posted to https://gist.github.com/matkoniecz/73aabf3cbcc0517f82eabe9b5b1aca92 3000th user has 309 edits 6000th one has 117 10000th one has 44 |
|
BTW another bias you may have to take into account for the stats in the OP: Of course, the later a quest has been introduced, the less users have seen it, i.e. old quests likely got more answers just because they are old. (though this could be mitigated by the fact that the user base is obviously always growing per time) |
|
Is it possible to calculate the 'impact' of SC? E.g. of all the highways with surface=*, what percentage was tagged by a mapper using SC? |
|
It should be, although due to the quantity of data it might need to be geofenced to particular areas |
I think that fairly good answer can be achieved by assuming that every SC edit added one case of surface. Ignores splitting ways (reduces SC impact) and ignores undoes (inflates SC impact). See https://taginfo.openstreetmap.org/keys/surface - 29 564 798 cases, 1722k SC edits, for 5% of tags added by SC - https://duckduckgo.com/?q=1722*1000%2F29564798&t=canonical&ia=calculator
|
|
Isn't 29M the total, so it should be the denominator? That is, 1.722M/29M, or ~5.8% |
https://taginfo.openstreetmap.org/keys/surface#combinations |
Yeah, right - thanks for spotting what went wrong! |
For start - thanks to all mappers who edited together over 5 million elements in total!
I made a list of the popularity of every StreetComplete quest - what is the most popular quest? how many elements were modified by AddOpeningHours quest? Have you added quest and were curious how many elements it edited? Scroll below for list or read some about additional detail.
I made it by a little number crunching of metadata of all changesets from the first changeset to 2020-02-24 (date of Weekly Changeset Metadata file on the mirror that I used). Unlike full planet file that requires a beefy computer to load, this file containing metadata of all changesets ever made can be easily processed.
The list below includes all edits made by apps using StreetComplete changeset tag format done more than 36 times. It means that it includes
If someone is interested in code used to do that - look at https://github.com/matkoniecz/StreetComplete_usage_changeset_analysis
If someone is interested in similar statistics
Note that usefulness and effort needed to solve quest for one element differs significantly depending on a quest. For example AddPlaceName solved 20 thousand times likely took more time and is more useful than many quests that edited more elements in total.
In some cases further research may be interesting about usefulness - how many AddParkingAccess and AddPlaygroundAccess confirmed
access=yesand how many helped to mark this objects as private?The text was updated successfully, but these errors were encountered: