Laps Split

[TheoSugarelis]

Hello everyone,

First of all, great job—and thank you for making this work public.

I have a question about how the lap splitting works and how accurate it is.

Am I correct in assuming that the split is based on the timingdata messages when the car crosses the start/finish line (sector3 as identifier?), and that the lap timing accuracy is therefore limited by the message frequency?
Is there any additional processing that adjusts the lap boundaries (for example, trimming or extending them) based on the official lap time?

[theOehrly]

The lap time itself and all sector times are 100% accurate. The timestamp/date at which a lap was set contains an error. But this error will be constant for all laps, since FastF1 syncs these timestamps based on the other available data.

[TheoSugarelis]

Hello, thanks for the reply.
Yes, exactly that was my point (maybe I did not explain myself well enough):
-Since we know that the laptime itself and sector times are 100% accurate
-But the timestamp/date at which a lap stats/finishes contains an error...
---> a) I assume this error/info about the timestamp comes from the timingdata endpoint? (eg "the timestamp we get the laptime = the timestamp lap N finishes, and lap N+1 starts)
---> b) i was wondering if any correction logic (based on the accurate laptime) takes place.
eg. if the laptime is 70.0s , but the calculated form timestamp is 70.3s, is the lap "trimmed" , trying to reflect the actual laptime/lap start/finish time? or even better: if laptime of lap N is 70.0, but the calculated is 70.3... and the laptime from lap N-1 is 70.5 but the calculated is 70.2.... --> would it make sense to "move" the lap change timestmap by 0.3s earlier?

[theOehrly]

Yes, the error comes from the endpoint. My assumption is that the timestamps we receive resembles the point in time when the data is sent out via the data stream. It is not the time, when the sample was taken.

The correction logic that you are describing already exists. FastF1 recalculates the timestamps based on the lap times. If you were to calculate lap times from their timestamps, you will notice that they line up.
Timestamps are adjusted so that no timestamp is moved to a later point in time, since that would imply that the data was received "too early". But that would be physically impossible.

In race and sprint sessions, we additionally have data for "gap to leader", equivalent to what the TV graphics show. When this data is available, it is used to also adjust the timestamps across drivers in a similar way.

What remains is one constant delay that is equivalent across all drivers and laps and we have no idea how large that offset is.