@Janis [22,13],[21,13],[21, 12],[20,12],[20,11],…
The unit should prefer [21,13] because it has moved vertically before (and the distance is the same)
So something like this green+purple path?
Yes @876584635678890 seems like you also see it the same way.
Another occurence of buggy pathing: https://terminal.c1games.com/watch/564957
Turn 7, 12 and 13
Should have targeted [1,15] but instead go around the blockage.
This is currently being worked on. The path 8 highlighted is correct.
Just more examples of this for review. This one took a chunk off my ELO too
https://terminal.c1games.com/watch/567576
In rounds 6 and 13, my pings on the left should have just breached at [1,12] once they destroyed the walls there, but instead they go on a hilarious wild detour through the death trap. Bizarrely, in round 17, a very similar attack succeeds without problems.
Indeed a detour.
Explanation to bizarreness is probably cause on 17th round there are 12 pings whereas before there are 10 of yours. This makes attack more powerful and kills filter just a couple of frames earlier.
When fixing an extremely rare pathing bug last week, I introduced a pathing bug that has been occurring in most games this week.
The previous bug was due to a very rare edge case which would only occur when a tower was destroyed while a unit was standing on top of an endpoint, while there was another friendly unit in a location more than 2 tiles away attempting to reach the same edge. The unit would take their friends path instead.
I introduced a new and much more common bug to pathing when I ‘fixed’ that last week. I fixed the issue by forcing endpoints to always be open, valid, and ideal. You don’t need to worry about these values, but the problem was that in the process of validating these tiles we would gain information that we no longer had after these steps were being skipped, which led to the new issues.
Everything should be fixed now. For real this time. The fixes will be released tomorrow midday.
Happy to hear that! Thanks for the work.
This bug (even wierder) might be related:
on the round 5 of this match : https://terminal.c1games.com/watch/592608, the pings go for suicide even if several paths are available
Edit: As pointed out in the comment below, nothing wrong here. I did not watch properly
I don’t see the bug you are talking about: those pings on turn 5 had to reach the top right edge and while there was many paths to the top left edge, the units could not reach the top right edge, so they exploded.
oh damn
I looked twice, I should have looked a third time to be sure.
Thanks for that!
I think the bug is back!
This was in the playground between 2 versions of Aelgoo
Player 2 scramblers took the green + red path (I expected the green + cyan)
Ooph. Ill investigate. Could you drop the IDs of the two algos used if they are not random and you still have them
Player 1: 31389 Aelgoo_47pre2
Player 2: 24676 Aelgoo42_The_Answer
Turn: 27
They are not random but they implement some time control, so if the computer running them is slower or faster than the playground it can change a lot.
Ya go for it I have them
Quick update, a few matchmaking things are on dev servers and going through the QA flow, i’m looking at this now.
@Aeldrexan You can download all of your previous algos (if you have the id) by using the algo
API endpoint.
Original upload: algo/{id}
Extracted version: algo/{id}/compiled
I think this may be another pathing bug in the leaderboard matches;
https://terminal.c1games.com/watch/711825
My pings in round 10 and round 15 don’t follow the usual “diagonal” rules, and instead make a sharp left-hand turn traveling in a straight line for maximum self-damage
Not sure what the current state for this bug is, if I missed somewhere that it is being worked on but not implemented yet. Either way, here is another game I noticed with weird pathing at round 15 (and a couple others). I did not affect the outcome, but thought I’d bring it up in case we thought this was fixed.
I was working on it earlier this week, but was delayed because there is another engine change that has to be done first. The current bug is fairly rare and there are no major competitions coming up so its not an extremely high priority for us right now.
Sorry pathing has been an issue, its kinda 100% my module so I take full responsibility for that