iRacing semi-automated incident wibbletron

[Dave 'Gizmo' Gymer]

Rather than have this rumble on in the Silverstone race thread, I thought I'd start a new one.

I'll have a play at getting a system that can parse the data into a more readable format from just the SessionID.  From looking at it, it seems pretty do-able.

Before we get too carried away, let me point out (again!) the gaps between the available data and what needs to go into the court system as reports that the drivers can respond to:

• 'SR flags' involving several cars as part of a single incident need to be grouped together so that each driver knows which incident it is they're responding to.
• A car may have more than one incident of the same type on a single lap, and there needs to be an incident report for each one.

These two may overlap, of course.

If only we had a GPLRA-type tool that could spot related incidents in the replay and provide timestamped groups of data from which incident reports could be generated.

[Paul968]

As you pointed out a few weeks ago when I asked about this, we can't get all the data, just the laps where things occur. If we had a report which collated all the data into one place though it would help enormously, and this is all I would suggest is needed. A mod can always check the rest of a contact lapto see if there is anything else untoward. The system wouldn't be foolproof but as long as the vast majority of incidents were picked up then then it would do its job.

[ross.mcw]

The basics are certainly do-able - here's a quick test I did for the last race in the Skip at Silverstone (not automated, but used the logic I'd implement to get the data pulled together):-

http://www.zensoftware.co.uk/downloads/temp/iracing/20100316.Silverstone.Skippy/incidents.htm

Certainly makes it quick and easy for mods to see where they need to be looking in the replay.

I should be able to put together an automated solution for this for mods to use which would present the data in a nicer way (mainly to include driver names instead of just their number), but that data can easily be copy/pasted into a spreadsheet.

Cheers, Ross.

(updated to point to revised URL)

[popabawa]

Great work Ross! looks very promising

Looks like there were 6 instances of car contact then (12 mentions, I'm assuming 2 cars involved each time?) so not too onerous for the moderator to check each one.

Having the driver name would be v. helpful, I think that might be enough really.

[ross.mcw]

Yeah, should also be make it easy to pick-up on anyone consistently over-driving or track clipping.

[ross.mcw]

Having the driver name would be v. helpful, I think that might be enough really.

Yeah, that shouldn't be a problem at all, quite a few things I can do with the data once it's done programatically.

[popabawa]

Could you do a summary per driver then? e.g.

name	1x off track	2x lost control	4x car contact	Total Incidents
Iain Mabbott	4	2	0	8

[Paul968]

Yes, that would be good. It would also be useful to see if the number of incidents added up to the total reported on the results page, as this would show if some were hidden.

[ross.mcw]

Yes that might be do-able as it'll just be a different database query but can't car contacts be 2x or 4x?  If so, the data doesn't seem to provide any differentiation.

I did find that there's a flag (the lowest bit of the binary version of the flags value) for an incident that doesn't have a description which is a bit odd, but it's only occasionally incurred so probably isn't important at all - could be anything really and is perhaps just a status flag for something relatively unrelated.

[ross.mcw]

...ah except that, as Dave's highlighted above, the data available only flags any incident type once per lap so trying to provide a breakdown of incident points won't be do-able I'm afraid.

I guess it should just be viewed as a tool to help figure out where moderation for non-reported incidents may be needed.

Cheers, Ross.

[Paul968]

What I meant was that the results give an incident total for each driver. It would be good if the report could flag up the difference between this count and the calcualted total from the different types of incident.

[ross.mcw]

Thing is it could be wrong because you could have 10 off track incidents and get 10 points, but this would only be flagged in the lap chart once.  Just seems to be a limitation of the available data I'm afraid.

[Fred Basset]

But even if this just tells the mod which laps/drivers to look at it would be an enormous help wouldn't it.

Looks like good work to me.

Gary

[Paul968]

I know - that is why I want to see the difference.

e.g

Dave Gymer       53 total incidents           4 x car contact     7 x loss of control     14 x off track    =    44 incident points         discrepency = 9 inc points

[ross.mcw]

Dave Gymer       53 total incidents           4 x car contact     7 x loss of control     14 x off track    =    44 incident points         discrepency = 9 inc points

The day that happens, I'll eat my computer....

I'm still puzzled as to why you want to see the discrepancy though given that we know the data we have access doesn't allow us to calculate the incident points total with any certainty at all?  Am I missing something?