News PlayTracker Insight - a new website for tracking Steam owners (and more)

lashman

Dead & Forgotten
Sep 5, 2018
30,365
85,134
113

View stats and graphs on Steam games. Find out how many people own a game, how many are actively playing it, how it has fared over the years, and more. All free, no strings attached.



How does PlayTracker Insight work and is it accurate?

PlayTracker Insight gathers public data from gamers around the world in a random sample and extrapolates that data to estimate total values.
PlayTracker correlates that random data with internal data from PlayTracker users to get more detailed information, for example the "Achievements over time" graph, which shows how a game has fared over the years.
PlayTracker's current sample consists of over 20 thousand Steam accounts whose full game libraries are synced into a database with over 20 million data points. This makes PlayTracker Insight a reasonably reliable market analysis with a margin of error under 1% for 95% confidence. It is, however, still an estimate, and can be off. This is no different to other similiar services.
The sample grows every day as PlayTracker crawls new accounts. The current sample is updated on a cycle lasting around 2 days, meaning data for a game's launch weekend should be fully accurate by the beginning of the next week.
PlayTracker started gathering data in early March 2019.

Why are you giving everyone access to this data?

After Steam changed profiles to be private by default, SteamSpy locked most useful data and graphs behind a paywall because a lot of things were broken and needed fixing. I disagree with this philosophy and believe it will be easier to fix and improve Insight with the help of the public. Furthermore, it's simply useful and fun to browse data, and I think it can survive off Patreon donations encouraged by convenience features.

What is "popularity"?

Popularity is a metric derived from the number of people that own a game and the number of people that are actively (in the past two weeks) playing it. The active players have a 4x multiplier, which evens the two numbers out in importance. I tested various popularity algorithm and found this one to be the best.

Can I use your data for my own thing?

Yes. Be it in an article or elsewhere, you may use PlayTracker Insight data, article text, and article cover images. Please give credit if it is not too inconvenient, and consider donating on Patreon. Feel free to contact me on this topic if needed.

If you have further questions, feel free to contact me at marijan@playtracker.net.
 

Digoman

Lurking in the Shadows
Dec 21, 2018
854
2,390
93
Steamspy getting C O M P E T I T I O N
Not gonna lie, even though I don't see how this can be any more accurate because of the biased sampling (only public pages and/or achievement data), I would be pretty happy if takes money away from the Patreon Galyonkin is still running despite working for competition now.

Yes... I'm being petty... and I don't care :)
 
Last edited:

OSatan

OFather OSun
Not gonna lie, even though I don't see how this can be any more accurate because the biased sampling (only public pages and/or achievement data), I would be pretty happy if takes money away from the Patreon Galyonkin is still running despite working for competition now.

Yes... I'm being petty... and I don't care :)
Even worst he gets money for lying, the current stats are useless but a lot of his patreon do not know.
 
OP
lashman

lashman

Dead & Forgotten
Sep 5, 2018
30,365
85,134
113
update from the dev, about inaccurate owner numbers:

The launch of Insight was way bigger than anticipated thanks to retweets from very cool people! <3 It is late here in Europe and I cannot respond to everyone tonight, but before bed I'd like to quickly address the topic I saw the most questions about - the estimated owner count.

The estimated owner count is calculated from a sample of crawled public Steam profiles and extrapolated to the entire Steam audience to get an estimate. It is a method like the one described on ArsTechnica, I believe SteamSpy works similarly

Because it is still a sample and not the whole population being polled, the estimate is sometimes incorrect. While building Insight, I built it to accomodate a system that can deal with these anomalies by assigning a "bias" to an estimate proven to be incorrect.

This was future-proofing on my end, intended for when the service one day grew to a level where developers were actually aware it existed. That is why I made Insight data public for everyone - the estimates are flawed, and only solid data can help make them better.

If your game is displaying (wildly) inaccurate estimates, and you're willing to share your data, feel free to contact me via PM or email. I will go more into detail on this topic tomorrow in an article on Insight. I apologize for any fake richness Insight may have projected. ;)

P.S. Forgot to clarify that the "bias" should indeed affect games other than the actual game in question, it is not just a manual override. To put it simply, it teaches the system which games are over/under represented in the sample.
 

texhnolyze

Child at heart
Oct 19, 2018
3,581
8,584
113
Indonesia
I remember a couple recently disclosed sales number games, Slay the Spire and Dead Cells.

The former is quite accurate, they announced 1.5 million copies sold a couple of weeks ago and it's showing 1.6 million copies here.

The latter though, is way off. They said it has sold over 1 million copies (accross all platforms) last week, and this site shows 2 million.
 

Wok

Wok
Oct 30, 2018
4,923
13,188
113
France
Unless he explains more clearly what he does, people will be suspicious.

I remember a couple recently disclosed sales number games, Slay the Spire and Dead Cells.

The former is quite accurate, they announced 1.5 million copies sold a couple of weeks ago and it's showing 1.6 million copies here.

The latter though, is way off. They said it has sold over 1 million copies (accross all platforms) last week, and this site shows 2 million.
Maybe the webmaster adjusted the "bias" after reading the sales number for StS, as you did.

Maybe the webmaster missed the sales number for DC.

If that is the case, well, numbers are only going to be accurate after the webmaster has the correct data, which would defeat the purpose of the site: the algorithm has to extrapolate/generalize well in order for the website to be useful.

For instance, this is weirdly written:

PlayTracker's current sample consists of over 20 thousand Steam accounts whose full game libraries are synced into a database with over 20 million data points. This makes PlayTracker Insight a reasonably reliable market analysis with a margin of error under 1% for 95% confidence.
Reference: About PlayTracker Insight
 
Last edited:
  • Like
Reactions: lashman

Alextended

Segata's Disciple
Jan 28, 2019
5,461
8,524
113
How can you get correct data for hundreds of thousands of sales or millions of sales in some games out of just 20k Steam accounts that are public or whatever? Surely it takes a lot of guesswork and assumptions which can vary greatly depending on the type of game and who it appeals to? Same for SteamSpy really.
 
Last edited:
  • Like
Reactions: lashman
OP
lashman

lashman

Dead & Forgotten
Sep 5, 2018
30,365
85,134
113
to be fair - they literally launched that thing ... yesterday ;)

give it a month or two ... we'll see how it evolves
 

bobnowhere

Careful Icarus
Sep 20, 2018
1,665
4,288
113
There is nothing inherently wrong with sampling as long as the sampling method is good. Political sampling in Australia is usually within 2% with samples of less than 1000 in 15mil voters. My concern is I suspect those that have conscientiously made their account public are more likely to be more invested in steam and would own more games, hence the massive over estimation currently.

Also, the more a game is know, the better the estimate it seems. This is to be expected, the more you are invested in steam, the more you are likely to delve deeper into the dark corners of the store.
 
Last edited:
OP
lashman

lashman

Dead & Forgotten
Sep 5, 2018
30,365
85,134
113
There is nothing inherently wrong with sampling as long as the sampling method is good. Political sampling in Australia is usually within 2% with samples of less than 1000 in 15mil voters. My concern is I suspect those that have conscientiously made their account public are more likely to be more invested in steam and would own more games, hence the massive over estimation currently.

Also, the more a game is know, the better the estimate it seems. This is to be expected, the more you are invested in steam, the more you are likely top delve deeper into the dark corners of the store.
yeah, that seems to be the case so far ... hopefully it'll get better with time