Jump to content
  • Sign Up

WvW API real time data


Pemberly.6305

Recommended Posts

Has anyone else seen this? Is it old news?

 

I spent some time looking at the time interval between updates to the # of kills, and # of deaths, for each match, map, and team in the API. I sampled the API every 2 minutes give or take a few seconds to account for the time it takes the API to respond each time through the loop. I then step through each list of data to determine the number of consecutive identical values, and plot the results in a histogram. For example, if list_kills=[120, 122, 122, 122, 126, 134, 134, 179], list_consecutive_values= [1, 3, 1, 2, 1]. There are over 45k entries of consecutive identical values, where data was collected from Saturday @ Noon to Sunday @ 22:30 New York City time.

 

https://drive.google.com/file/d/1hG8Z7w4YFx50uPca9UtNEi5YuhIbm1mY/view?usp=sharing

 

The plot above shows the vast majority of API updates to the # of kills, # of deaths corresponds to 1 or 2 consecutive values, or approximately 4 minutes or less. This lines up well with the 5 minute RI timer for flipping objectives, assuming the API updates objectives status, war score, # of kills, and # of deaths at the same time. The large number of 1 consecutive values indicates that the API has the potential to update every 2 minutes, but does not always do so.

 

Zooming in on the y-axis shows a collection of values in the 15 to 30 range. This corresponds to time intervals of 30min to 60min between updates to the API reporting # of kills and # of deaths. I cannot explain the local minima at 15, and local maxima at 27. There are two possibilities, one is map inactivity, and the second is a minimum threshold before the game engine updates the API server. It is, in my experience, not un-common for maps to be completely inactive for periods of time. This would lead to consecutive identical values of # of kills, and # of deaths, with the API being updated at regular time intervals. A second possible explanation is a minimum number of new kills, such as 5, must occur before the game engine updates the API. There are probably other explanations too, and I am interested in learning what people think.

 

https://drive.google.com/file/d/1Q2M4RoZYv2Kd-Yyh-iriJ_mGKKlity8T/view?usp=sharing

 

Zooming in further on the y-axis, and expanding the x-axis to max values yields the plot below.

 

https://drive.google.com/file/d/1I8KvZY_MejXvruwd9IAKt1zt28ix4BHB/view?usp=sharing

 

This is the plot that confuses me the most. 120 consecutive identical values corresponds to 4 hours. There is no way a map has no kills or deaths for a period of 4 hours, and so this time interval between API updates of # or kills (or deaths) has no explanation I can think of. Even 40 consecutive identical values, 80 minutes, of no kills seems very unlikely to me.

 

Comments and Questions Welcome.

Pemb.

 

 

Link to comment
Share on other sites

Hm, it's known: The APi isn't always up to date or hangs a while from time to time, not only with kills, also with scores (e.g. on http://gw2stats.com/eu/matchups it sometimes takes quite a while till a skirmish that has ended is shown to be ended, and I think it's the API and it hangs differently long in different matches, so T4 may have a skirmish finished, that "still running" in T3)

 

So figure's like yours would show periods where even score hangs for a while longer than this is possible technically (as PPT should increase every 5min)

Link to comment
Share on other sites

https://drive.google.com/file/d/1bC6AB1XKoqYVlT-EAXsxsdfZk4PwQjB5/view?usp=sharing

New data set. Wednesday at Noon to Friday at Noon.

Good News, the maximum number of consecutive identical values has dropped from 160 down to 65.

 

In the plot above, we limit the data set to sequences of 25 or more consecutive identical numbers of kills. For each identified such sequence, we plot it against the number of consecutive identical deaths at that time.

 

The takeaway is that because the resulting scatter plot is a line, whenever the sequence of kills stops being updated, the sequence of deaths also stops being updated, for the same amount of time.

 

I think this rules out the game engine as the place where the lock up occurs, because the running list of kills and deaths are different processes in the game engine. I think that leaves either the transfer of data from the game engine to the API data base, or the API data base itself as the source of the lock up.

 

Comments?

Link to comment
Share on other sites

https://drive.google.com/file/d/1pRq1o6ZvtoV1HcJy-aG8tNKJ8iW5CuXT/view?usp=sharing

https://drive.google.com/file/d/1xdlexlxbZapYcgZfeHJzPFVZ7ruxawbS/view?usp=sharing

 

New Data Set: Wednesday 18Nov to Friday 20Nov.

Things definitely to a step backwards for this data set.

Maximum length of consecutive identical values is 107, 235 min, ~4hours, 2 skirmishes.

Link to comment
Share on other sites

Looking at the Gaussian distribution around 25 consecutive identical values.

~ 3200 events of avg 23 entries duration times 2.2 minutes per sample = ~162,000 minutes in 48 hours.

dividing by (9 matches * 4 maps * 3 teams * 2(kills and deaths)) 216 individual data streams.

= avg 12 hours of downtime for each stream in 48 hours.

 

So, if my math is right, 25% of the time, the WvW API is hung up and not working.

I hope my math is wrong.

Link to comment
Share on other sites

https://drive.google.com/file/d/19n4XSrBc-UissJg07Eok6quPsPLSj24F/view?usp=sharing

https://drive.google.com/file/d/1H0qeqbhz0PWac9yfrXme2EDWxOx9kG7a/view?usp=sharing

Latest Data Set: NA Reset to Monday @ Noon, 23Nov2020

 

No improvement in results.

 

From now on, all results will be posted at https://sites.google.com/view/gw2-wvw-for-beginners/wvw-live

I guess I am naive in expecting Anet to fix something that is broken 25% of the time.

Link to comment
Share on other sites

  • 3 weeks later...

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...