My latest algo submitted, 116942, is currently marked as crashing, but none of the 100 matches last played have any crashes!
Either it was mislabelled as crashing, or enough matches were ordered that after the crashes happened, 100 newer matches that were already scheduled ended up hiding the all the crashing ones!
Could there at least be some way of keeping these statistics in sync? So an algo might be marked as crashing explicitly because x% of the last 100 matches crashed, but not because the crash total was too high?
I took a quick peek at this endpoint:
https://terminal.c1games.com/api/game/algo/116942/matches and it certainly claims to have seen 5 crashes in it’s life (so your second suspicion seems correct):
From experience with my own crashing algos, since the crashCount is tracked with an actual number it has been accurate.
Personally I wouldn’t advocate for a formula based on x% of last 100 matches. Continuing to play matches such that the actual crashes fall out of the history needs some kind of fix - I’d say either they are totally removed from the match making pool, or the crashed matches somehow get archived in such a way they don’t fall off the end of the list.
Thanks for that look up!
Yeah, I’d say whichever is easier for C1Ryan to get working would be fine with me!
Ok, will look at this. Might just keep crashed games at the top of the matchmaking history, since they will usually be the most important.
That would be a great solution!
This is now done and will be released this week.
While working on this, I realized that the 100 match limit for that endpoint was originally implemented for technical reasons that are no longer relevant. I will be removing the 100 match limit.
Well that is a wonderful extra development! Thank you!