Mismatch between timeout rates in endpoints and webpage data

Why don't these agree? It's been pointed out to me that the timeout rates for two player in the following match don't agree with the figures given in the web pages:-

https://api.chess.com/pub/match/1654189

"username": "irijeteveliova",
"rating": 1183,
"timeout_percent": 25 (web page shows Timeouts: 7.94% Last 90 days)

.

"username": "ljuba1009",
"rating": 1231,
"timeout_percent": 100 (web page shows Timeouts: 10.59% Last 90 days)


They can't both be right so why does the endpoint in each case disagree so widely with the web page data?

There's a ticket in for staff but the API is pulling the data the back end has.

So it's the web-page timeout figures that are incorrect, not that in the match endpoint?

Yes, I sent a feedback report to chess.com about a week back. I spoke with their support and they said the devs are aware of it now (since my ticket) and will work on it when they have time but they don't know when that will be. 
The endpoint information is incorrect. It is drawing values from somewhere and it's not always clear where that is.

My understanding from the ticket is that there was a format change in how that's being calculated and the API hasn't been updated to take advantage of that.

Ok, so the webpage timeout ratios are reliable and accurate but it's the endpoint data that's incorrect in some cases?

I only ask because as we speak team admins are removing players from their matches in a tournament I'm helping to organise based on the data I've provided them with, data that in some cases is simply wrong.

Just checked a few. Confirm that the API TO stat is wrong in a lot of cases.

Tried to establish what C.C devs are calculating but can't see any pattern.

The Web page stat looks accurate for TO over last 90 days.

Another own goal. No checking before release and no urgency to fix the rubbish created by poor work.

"No checking before release and no urgency to fix the rubbish created by poor work"

Absolutely - I despair of this site!!

In fact you might add to that no attempt by the site to notify/warn users of the API about this even after it's been identified? There's such a lack of professionalism in the way this site is managed and supervised/overseen.

That endpoint is one of the key parameter I use for my team management and recruiting for tournaments (because "auto tournament invite" feature doesn't respect the timeout constraint, maybe for the same reason? I have to do a pre-filtering).

I wonder now if the data is totally random, if it is wrongly calibrated, or if it's calculated on some other value. I can't say the players I select with my filtering have ever had a worse timeout rate than my criteria, so it's probably not random data.

I have a feeling the API values are being calculated as the percentage of timeouts in the last 'n' games instead of in the past 90 days of games? This would then make sense...

"username": "ljuba1009",
"rating": 1231,
"timeout_percent": 100 (web page shows Timeouts: 10.59% Last 90 days)

I tried to check if it was a number of games instead of time period but couldn't see any correlation

I am debating writing a script this PM to do my own calc from archives

Public API updates are very low priority and it appears it was work outside the API that is causing the API to return incorrect data. Developers are assigned to higher priority projects so it may be a while before updates are made

So yet another blow to the group admins and league organisers.

We had previously been told that that C.C considered API to be priority due to being used to organise online competitions for group members. Oh well. Good to know where we figure in the plans.

Think I have a script now that will provide 90 day TO's for matches in registration.

However .... This requires an additional 5 endpoint calls for each player. 1 for the archive list and then check the last four for daily games with lose by TO.

So a match with 500 registered players that is an additional 2500 calls to API. No doubt the resources will not be stretched by me doing this as it's a trivial number in the grand scheme.

It does however take a lot longer than extracting TO from each member stats at the client end.

Appreciate the candour and my remarks weren't aimed at you but might it be possible to start a thread for pending bugs/issues in which Jon, Ben and youself could post warnings of new problems and update us on existing ones?

It would be so useful if those of us who use the API extensively could be kept up to date with emerging 'issues' and what's happening with known ones.

Intead I've just posted a slew of data on player timeouts in the World League and I now realise much of it is complete junk!

Barring direct developer work on the API heavy users of the data are likely going to notice issues before anyone else. Keeping track of topics here is likely going to be the best way to see if there's an issue and generally how I find out about things that we're not direct API changes

"heavy users of the data are likely going to notice issues before anyone else. Keeping track of topics here is likely going to be the best way to see if there's an issue..."

The trouble is that there seems to have been a serious problem with the alerts system on the site for some time now, with many members failing to receive them for new topics in clubs etc. I'm still puzzled why I hadn't seen this one in the forum here....

https://www.chess.com/clubs/forum/view/stats-api-shows-incorrect-daily-chess-timeouts

I don't remember receiving an alert for that but notifications for new comments do seem to be a lot more reliable.

Yeah, that's the topic where I reported it to staff initially

Actually, I reported it two days before that was posted. Maybe from a chat or beta post 🤔