F1 Fanatic

This sounds like my dissertation at University, best get Matlab out and crunch some numbers!

The obvious problems are car differences, and then track differences too.

An idea I had for equalising the performance differences of the cars was to look at things like braking distances, top speeds, cornering speeds. This data would be available, maybe not to us fans, but if a person with good connections was to do it then they could probably get this data. There is obviously still driver skill in this. So maybe you could look at wind tunnel data, which would be extremely hard to get hold of.

I love statistics but sometimes they just aren’t applicable, and maybe in this case they aren’t. Sport has so many deciding factors, it’s not just the equipment, it’s not just the participant. It can be based on mood, and other external factors that are very hard to judge.

I still like the fact that math nerds are trying to solve the unanswerable question, and maybe they will prove us wrong.

It is an unanswersable question. But Patrick O’Brien methodology makes sense and it seems a very good aproximation, at least considering the subjective opinion. Still, too bad they don’t give us a real example, but I guess it does involves a lot of data, as sector/lap times for all sessions for all cars for all GPs.
BasCB, I did saw a couple of episodes, but got bored easily (I guess it depends the quality of the intervieweds). The blog though seems top quality with very good articles, almost no publicity, makes us wonder what else is out there.

I was wondering whether you could put ultimate lap times into the mix, taking the best sectors across the race weekend and adding them together across the team. This would give you a relative view of what the cars were achieving. It still would not separate driver performance from the system but should help to rank the potential of each of the cars. Drivers could then be assessed against this benchmark by looking at their best times and how they compare to the ultimate time. This could possibly include fuel adjusted laps if the data is available to look at consistency.

The advantage of this, and the difficulty of other methods, is the dynamic development of the equipment. One update can quite literally transfer the performance of the equipment. Looking at Historic data for past tracks wouldn’t capture this.

Also what variables would you use to define the circuit influence. Monaco would need to be reviewed in a very different way to Silverstone!

I’m new to this forum, and I have “resurrected” this thread because I’m deeply interested in this subject and quite curious about why it is so poorly explored – I hope the discussion can be resumed. There’s an interest master thesis about F1 driver ranking here:

http://david.stadelmann-online.com/pdf/0015_formula1.pdf

The problem with this ranking (and with many others) is that it deals poorly with the car factor, which is of course extremely important in F1. The f1-facts method is quite interesting, although it will be biased by the amount of weak/strong team mates a driver has had in his career. For example – Prost was team mate to many ex and future WDCs during his career, while Schumacher, on the other hand, had just a few races pairing Piquet. So I came up with an idea: what if we rank drivers as in an incomplete round-robin tournament where “matches” are only considered when in the same car? The scoring scheme could be the same one used by f1-facts. So, if driver A and B never drove the same car, but driver C was team mate to both of them, we could rank A and B using their f1-fact scoring against C: If A vs C equals +35 and B vs C equals -15, a A>C>B rank is produced. This is repeated using all team mate results until a final rank is produced (given that noise is expected, this may need numerical optimization, but there’s certainly literature about dealing with that). My first impression after thinking about it is that it would still leave the problem that it may not be possible to generate a single network of “matches” (e.g., being impossible to rank drivers A and B due to the lack of a “team mate string” connecting them), but after five minutes with pencil and paper I was able to connect *all* drivers in the current grid in a single network adding just one external driver (Barrichello was needed to connect di Resta and Hulkenberg to the rest of the network, which had plenty of connections) – so perhaps it is possible that most important drivers throughout F1 history may be “connected” by such team mate strings. Yes, it will definitely not the be “ultimate driver ranking”, but it certainly looks fun for those who like statistics, graph theory, etc. Anyone interested?