On the 5th of August this year, I set 3 new world record in a fully continuous 9 minute video recording for the scenario 'Police Patrol' of 18.25, 17.75 and finally 17.25 (with a successful run of 18.50 shortly after the 18.25) after i had realised not long after starting to record the speedrun session that i could pit the suspect at around the 10 second mark on the in-game timer. 8 days later (13th August), King had rejected my times for the following reason: 'Decreased gameplay performance is affecting the AI behavior, after doing some testing it seems like slow AI reaction times caused from low framerate(sub 20FPS) is allowing the chase vehicle to catch up far too soon when compared to a run where the FPS consistently remains above 20. Due to this reason, the submission will be rejected.'

After what i admit was a somewhat heated debate on the run, King had basically said that, because of the game was running slower than it should and it had been running on a framerate slower than 20fps, AI performance was slower allowing for the possibility of me catching up. I uploaded a video showing 3 lots of runs i had done on Beam. These were the first ever run i submitted to the leaderboards on the 'Small Island Chase' scenario, my previous world record of 44.75 uploaded on the 31st of May and my new world record which was right after i installed the update from just before the 5th August.

The SMC run was recorded on a laptop that i used at the time and isnt that good in framerate with some stutters. The 44.75 run clearly shows it has a lower framerate than the 17.25 but was verified to the same ruleset as the 17.25. I pointed out that there was an update and that i believe it was that that had caused this since before that it was pretty much impossible to even touch the suspect but its easy to pit the car now after that update.

The discrepancy lies within a record that was verified approximately 20 hours ago by Mattcalz and was verified by King. It was submitted on the 25th of September and was verified on the 29th. I was surprised to see that the time was lower than my record and as i viewed it, i noticed it was the exact same situation of catching up early and using the strat i used of pitting the car into a tree on the left-hand side, however it was cleaner in my opinion. How can this run be verified when it is very similar to my run from almost 2 months earlier? It cant be framerate/game performance like my run since the framerates were similar and the game performance were also similar. I could not see any other reason for why this is so. I am confused on how this discrepancy exists on these leaderboards and my run to be rejected when a very similar run in most aspects was verified with the same ruleset.

It was verified simply because in the newest version of the game, the AI now acts this way even at 30FPS or higher. In the update prior to v0.26, the behavior demonstrated in your submissions was only able to be replicated by limiting the FPS to sub 20s. Which I felt as though limiting the framerate to nearly unplayable rates becoming the new meta for WR AI runs wasn't something that current and future runners would find agreeable.

But as far as Police Patrol itself is concerned, I had to go back to check for myself after the discussion we had about this exact issue a month or so ago, and discovered the AI behavior as it is now. In a perfect world it'd remain consistent across updates obviously, but since BeamNG is still in Alpha constantly undergoing major changes such alterations are expected to continue happening. Entire scenarios including Police Patrol fundamentally change at times making standing WR runs impossible and are archived to preserve the outdated runs.

Given the current circumstances with AI behavior it may happen again soon, but we'll see what changes the updates bring.

Why then should there not be a split between the 2 groups of versions where the ai act differently? It actually would benefit all involved. People who are able to use the older versions can still submit runs, people with the newer versions wont be left on the table, and those who happen to play on a version where the game runs fine but the ai acts in a different way depending on many different aspects wont be given a middle finger by both sides.

It's not prudent to need to verify for several separate versions of the game for each instance of the AI acting differently or if something changes, as it wouldn't just be two or three groups especially when considering the game changing going forward. It opens a Pandora's box of intricacies and issues to verify around that even implemented at its best would only clutter the leaderboards even more then they are now, all for something that only a very limited amount of people even seem to care about.

Odd to just jump in but newer Beam player here trying to improve my times, worked on the police patrol scenario for a while and wanted to post my time but seems kind of silly when the times are 15 seconds with AI not working as intended. Found this thread after looking for what was happening. Are there any plans to do anything with the Police patrol scenario or should I look elsewhere to post my run?

Id say with how its changed, dont sub to the one with my 15 run in, go for the other since the ai in it is alot like how it was before the version i got my runs on.