SecondMonitor (Timing & status App)

Discussion in 'Community Workshop' started by Winzarten, Oct 26, 2017.

  1. ravey1981

    ravey1981 Well-Known Member Beta tester

    Joined:
    Apr 15, 2018
    Ratings:
    +322 / 0 / -0
    I wish the sim would tell us what level the AI is using, always run adaptive and it works well but I have no idea at what level I am actually racing at.... I can't see why it would be hard to have this info in the shared memory, assuming the level is already locked in as soon as you fire up a race "weekend"
     
    • Agree Agree x 1
  2. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    I agree, the rating is more or less useless in conjunction with adaptive AI. But there is nothing I can do about it, and as only 1 of 5 supported sims have this feature, I'm not realy bothered.

    For me personaly, the adaptive ai never worked well. I find it too eager to switch difficulties, especially increasing it. I can have a nice, close fight for the lead, only to find the difficulty increased several levels the next race. This, coupled with the non-transparency of the system is why I was never able to stick with it.
     
    • Agree Agree x 1
  3. Skidmark

    Skidmark Well-Known Member

    Joined:
    Oct 9, 2015
    Ratings:
    +146 / 0 / -0
    Me too.

    I don't hate it or anything but it just doesn't work for me... I find it unpredictable at times which won't work for a championship and I want to set the AI so that winning is just within reach for me if I am at my best.

    After using adaptive AI for several months I found it set itself so that I was in the upper-midfield. Nothing wrong with that but not what I want. So I'm back to setting the AI manually - which is a shame cos I want to do a championship but variable AI performance levels across tracks combined with my own variable performance across tracks means I can't find a single level that will give me a consistently competitive championship.

    I would LOVE to be able to set up a championship and, as I choose each track, choose the AI level for that track. Heaven on a stick.
     
    • Agree Agree x 1
  4. CheerfullyInsane

    CheerfullyInsane Well-Known Member

    Joined:
    Oct 24, 2015
    Ratings:
    +336 / 0 / -0
    Fair enough.
    Only a finite amount of hours in the day, so I can certainly respect that. :)
    Had hoped there was some log somewhere that showed the chosen AI, but no such luck I guess.
     
  5. ravey1981

    ravey1981 Well-Known Member Beta tester

    Joined:
    Apr 15, 2018
    Ratings:
    +322 / 0 / -0
    You are completely correct here, I have never really thought about it fully as I usually race at tracks I like (and am good at) so rarely encounter those problems. I think a good solution would be if the AAI reacted to the drivers performance in the practice session rather than just in the race. That way by the time you have practiced and got a set up sorted the AI should be training themselves to your level, especially on relatively unknown tracks.... I would agree that the difficulty ramp is not quite correct in that they do seem to get very good suddenly. I find that top 10s are easily achievable but getting on podiums can be very hard.
     
  6. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    Yeah, If there is a way to get the AI level, I would definitely go for that. Not only because of adaptive AI, but also because I generaly hate these kinds of "manual synchronization". I like apps that just work :D

    I also find this to be a nuisance, and tbh R3E is the most consistent when it comes to AI performance across different tracks. Other sims worse and it makes having good races in SP very difficult.

    In the future I would like to have a map with relative difficulties for all tracks. These relative difficulties would most likely have to be user defined, as I don't think I can easily extrapolate them from user performance. So you would at least have a way how to track the tracks difficulties.
    Then I would like to incoroporate these into the ratings, by suggesting different difficulty and by increasing/decreasing the rating of the field.
     
  7. ravey1981

    ravey1981 Well-Known Member Beta tester

    Joined:
    Apr 15, 2018
    Ratings:
    +322 / 0 / -0
    AAI is worked out on a per class per track basis though, so once you have a couple of races in a particular combination the AI should be trained. Only when no data for a track exists does it use an estimate based on overall performance (I think). I usually do a couple of "test races" using the restart race button to make sure if in doubt.
     
  8. Skidmark

    Skidmark Well-Known Member

    Joined:
    Oct 9, 2015
    Ratings:
    +146 / 0 / -0
    I agree, although with the exponentially increasing number of car/track combinations (keep 'em coming S3!! :)) it's much harder to maintain this consistency.

    I remember posting in the early days (2015-ish) about how amazingly consistent the AI was across different tracks but these days there are some (for me) examples of where in the same car I'm a 95 at one track and 110 at another which isn't ideal if you want to do championships.

    On the other hand its very hard, as an individual, to work out how much is of this "AI variability" is actually my own personal variability between tracks. It will of course be some of it.

    It would be an interesting project to get a large number of players to race at, say, 6 tracks and then rank them from fastest AI to slowest for each player. The top 2 then get their AI tweaked down a bit by S3 and bottom two get AI tweaked up a bit. Rinse and repeat over a number of trials. Not gunna happen I know (and there would be cleverer ways of doing it I'm sure) but its the type of thing that I think would be worthwhile if we think AI consistency is important (I obviously do!).
     
  9. BIGROBIN

    BIGROBIN New Member

    Joined:
    Jun 25, 2019
    Ratings:
    +0 / 0 / -0
    i have a problem with installing this app because i can't see the installer exe file. when i open the document
     
  10. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
  11. Alex Hodgkinson

    Alex Hodgkinson Sector3 Developer

    Joined:
    Jun 5, 2017
    Ratings:
    +1,248 / 0 / -0
    I'm having some problems with slow loading times with recent updated. If I do 4 laps of the Nordschleife, the laps take around 5 minutes to load when I tick them in session telemetry box. I even upgraded my computer from 8 to 16gb of RAM but it's the same, and I don't think I'm the only one having this sort of thing.

    Hope it helps that I've mentioned it!
     
  12. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    It is your fault, you keep adding things to shared memory, then it takes longer to load :D

    Just kidding, it is not only you, and it is problem that is very high on the list. The cause is that telemetry storing/loading right now is very basic and not efficient (just simple .net binary serialization of everything). If you couple it with track like Nordschleife, than it can take an ethernity to complete... and sometimes it doesn't even complete, as there is inherit limitation of number of datapoints you can have with binary serialization, and Nordschleife on some slower cars is pushing this.

    In the next version I would like to rework the serialization. I plan to use Protobuf-Net and according to some articles, it should bring significant performance increase (https://theburningmonk.com/2011/08/performance-test-binaryformatter-vs-protobuf-net/)

    This will ofc only affect new telemetry files, old files will still work but there is very little I can do with improving their loading times.
     
    • Wonderful Wonderful x 3
    Last edited: Jun 26, 2019
  13. SvenT

    SvenT New Member

    Joined:
    Mar 18, 2019
    Ratings:
    +1 / 0 / -0
    Hey,
    maybe it´s a known issue.
    Yesterday I did some laps on Nordschleife and used the "restart" function of Raceroom.
    It seems as if this messes up the fastest lap and also the deltatime display.

    What is the intended procedure to restart a race/training and keep SecondMonitor in sync?
    Also restart it ? Is there a fast option just to initiate a new session ?

    Regards
    Sven
     
  14. Alex Hodgkinson

    Alex Hodgkinson Sector3 Developer

    Joined:
    Jun 5, 2017
    Ratings:
    +1,248 / 0 / -0
    I have an idea for a histogram to evaluate camber behaviour, but it may need some tweaking to work.

    First thing, because it's mostly relevant in a corner, there should be an adjustable filter which can be used to ignore certain values. For example, when lateral G < a certain amount. Maybe there's some way to automatically calculate a value from this with the captured data?

    Next, you should have 4 histograms, one per corner. On each one, you would set the "optimum" camber angle, which would be your 0 on the X-axis. Then the charts would display time spent above or below the given value, remembering the G-filter so it doesn't skew the data with straight-line values.

    Does this make sense?
     
  15. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    @SvenT
    The application should detect a session restart, and start a new session automatically. You can also start a new sesion manually using the very small "round back arrow" button on the bottom right. There is a bug with ghost cars, as the application cannot distinguish between player and ghost car, so the times can get messed up.

    Partially :D Just so I got it right, x axis would be camber, with zero being the optimal selected?
    Just a curiosity question, why the setting for zero is necessary, and why a regular camber histogram would not be sufficient? My quess would be that it improves the readability of the chart, as the general shape of the chart would be much more telling with a set ideal value.

    But It shouldn't be that hard to do. Filtering for lateral G should work, as that is already used for some charts. One thing that I need to do, from which also the traction circle would benefit, is to filter outliers in lateral/horizontal G. So the data don't get scewed with instances where you i.e. hit a curb to hard... producing 5+ lateral G for few datapoints. :)

    And btw I send you a reques on discord, dunno if it reached you, or got lost in the interwebz :)
     
  16. SvenT

    SvenT New Member

    Joined:
    Mar 18, 2019
    Ratings:
    +1 / 0 / -0
    Hmm okay, any hints where my specific problem might have the cause?
    As mentioned I did some laps in the IDR challenge on the Nordschleife.
    I restarted the session and the delttatime was approx. 300 seconds "green" and the sector times didn´t get any update ().

    I will try it today again and come back with details if this would be helpful ?

    S.
     
  17. Thomas Jansen

    Thomas Jansen Sector3 Developer Beta tester

    Joined:
    Apr 5, 2018
    Ratings:
    +355 / 0 / -0
    A useful plot with the tire forces would be in histogram form per tire as with the damper velocities, with some useful calculations being the mean and variance of the data. In theory, the better the suspension, the lower this variance should be! :D
     
    • Agree Agree x 1
  18. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    You can have another try. If it doesn't work, could you provide the logs (the is a open log directory button in the settings)? I will also have a look. There is a possibility that the app has problem with handling the "challenge" mode

    I should be able to do that :)
     
  19. Winzarten

    Winzarten Well-Known Member

    Joined:
    Jun 15, 2016
    Ratings:
    +161 / 0 / -0
    I got some interesting first testing result from the telemetry loading rework.
    Single lap at Algarve is 17MB big using the binary serialization, and around 7MB big using the protobuf. This is nice. :)
    The performance improvement on the other hand is insane. That laps takes 14.60 seconds to load using the old serialization... Using the new, it takes 387...milliseconds... :D

    I will most likely create a migration, so if you load some of the older telemetry files, it will convert it to the new format, so you get the perfomance increase next time you load the lap.
     
    • Like Like x 4
    • Wonderful Wonderful x 1
  20. Alex Hodgkinson

    Alex Hodgkinson Sector3 Developer

    Joined:
    Jun 5, 2017
    Ratings:
    +1,248 / 0 / -0
    Yes, exactly that and for exactly that reason and a regular camber histogram would be harder to spot a trend with quickly.

    I got your discord request and I think I accepted it? Send me a message asking what you like when you're ready.
     
    • Like Like x 1