Ranking drivers 1950 to 2013 ... a simplistic approach ...

ZakspeedYakspeed

NeverUnderestimateThePredictabilityOfStupidity
Valued Member
What a ornery old badger this issue is ... generates plenty of debate ... plenty of angst .. plenty of "you canna compare across different era's, etc" ... but we have the technology to do something relatively simple that might guide us down a certain path that makes a little sense ...


So ... how to construct something simple and meaningful that does represent a fairly robust ranking of F1 drivers since 1950 ...

So bear with me (drink a double expresso to give you that little awake zing) as I lay out the ....

Parameters

* Take the full results from all events from 1950 Silverstone to 2013 Brazil

* Remove all Indy 500 drivers who only raced in the Indy 500

* Using the 1981 to 1984 points system (top 6 paid, 964321, all results count), recalibrate the actual points scored for each event ... the nominator ... my favourite points system ...

* Calculate the "races competed in" - the denominator ... basically count only events where the driver either (a) tried to qualify and (b) didn't qualify but started the race. Excluded are the F2 category results from the shared series back in the olden days, driver withdrawals prior to qually, drivers entered but who didn't attend the event, 3rd drivers, etc. This is a little more refined than what are commonly referred to as "prescences" ... for which you get a tick if you are the 3rd driver in the garage during the weekend maybe running in P1 or more than likely making the coffee and trying to figure out who you to work on to get a drive next year...

* Calculate the average points per "race competed in" ... nominator / denominator


Additional Measures for each driver

But to further differentiate the drivers, why not take each drivers win count, podium count, pole count and fastest lap count and calculate a Strike Rate for each ... this removes the "how many did he win" and replaces it with "how often did he win" ... a simple measure of effectiveness rather than quantity ... sorry Michael ... you should never have come back my friend ....

i.e.
Win count / Races competed in = xx %
Podium count / Races competed in = xx %
Pole count / Races competed in = xx %
Fastest Lap count / Races competed in = xx %


So now we have five different data points (the average points per event competed in, and these four Strike Rates) for each driver during their career (or continuing career) ... and we can rank each of these categories highest to lowest, best to worst.

I have taken each of these five categories and have assigned an even 20% weighting to each of them ... i.e. I have taken the actual ranking against each Strike Rate for each category for each driver and multiplied by 20% ...

eg. Jim Clark - win S/R rank = 3 ... multiply by 20% = 0.6


Blended Rank Sum

Once I have calculated all of them you can add them together ... I call this the Blended Rank Sum ... then we can rank each driver on the basis of their calculated total for Blended Rank Sum ... which has captured the five different categories above. Again ... this is about effectiveness rather than quantity ...

The pdf attached shows the Blended Rank Sum for the Top 30 drivers ...

What does this tell me ...

=> that no-one will be better than Fangio ... not in my lifetime I believe ...
=> that Jim Clark was the real deal ... RIP
=> that young Seb is 4th behind Fangio, Ascari and Clark ... but ahead of Schumacher, Prost, Senna and some rather impressive talent ...
=> that Jackie was no slouch either ...
=> that Stirling Moss was desperately unlucky not to claim a deserved WDC
=> that J-P Montoya was pretty special too ...


I have attached the EXCEL file for those like to tinker (I have saved it EXCEL 97-03 for those who might not have a current version) ... I have also added a second file using the current points system (25,18, etc) that opens up a whole new set of rankings for those who finish 7th thru 10th ... I see you your Andrea de Cesaris and raise you my Mauricio Gugelmin ...

The sheets have filter buttons activated (see row 4) ... so if you want to filter to all Italian drivers, simply click the filter button on the Country column, uncheck the box (Select All) which clears the filter, then find Italy, and select it, then hit OK ... and you will see only Italian drivers ...

For those who have their own opinions on the importance of the weightings ... who might value a win more than a FL for example, then you can change the percentages above each column (AH to AL) in row 3 on either sheet and and it will recalculate the Blended Rank Sum and automatically calculate the new Rank for you each driver.


Comments, questions, critiques (oh ...is that the time) and queries are welcome ...
 

Attachments

  • Avg Points per Race 2013 11 (CTA) - 8184 top 20.pdf
    207.8 KB · Views: 208
  • Avg Points per Race 2013 11 (CTA) 8184.xls
    549.5 KB · Views: 187
  • Avg Points per Race 2013 11 (CTA) 2010.xls
    546.5 KB · Views: 179
Last edited:
mjo ... calculating the "races competed in" rather than using the "presences" took a little grey matter and time out of my work day to sort out ... most of the other stuff I have in a relatively easily manipulatable format ... but to get it to that state took a shedload of time ... spread out over boatload of time ...

I think it gives a good representation of each drivers relative position to each other ... lots of talk in the Mark Webber thread the past week spurred me on to analyze some of the data I have and see what shakes loose ... but why not go back to where it all began (officially) at Silverstone in 1950 ...
 
If I had known you were doing this I would have given you the source data for the F1 Database we used to have here.

It has the race results for every GP, including all the qualifying and race statuses such as DNQ, DNS, FAT*, RET, DSQ, etc.
As well as the retirement reasons, broken down into mechanical, driver error, etc.

Nice work by the way, I can appreciate how much time and effort has gone into it.

* = fatality
 
Brogan ... thanks ... I probably have the same / almost identical base data ... however yours is probably in a little bit better shape than mine !!
 
Fernando, Kimi and Lewis splits

Lewis has never had a drop off ... i.e. driving for a team that is not capable, or has won in the past ...

Kimi drove for Sauber in 2001 ... scoring 9pts ... no wins, poles, podiums or fastest laps ... his first season with McLaren was marginally better ... his last with Ferrari in 2009 was average

Nando had 2001 with Minardi ... two reasonable seasons in 2003 and 2004 ... average 2008 and a forgettable 2009
 

Attachments

  • Avg Points per Race 2013 11 (FAKRLH).pdf
    193.7 KB · Views: 173
Top stuff as always ZakspeedYakspeed !

It certainly looks like your results are close to how a subjective opinion may come out. It shows how adept people are at making calculations like this without ever realising it :D. One of the best descriptions of how we use maths in real life (way beyond what we can do on paper) is catching a ball because this involves complex differential equations that most people could never explain!

It is much harder to describe the formulas that people innately use, but that is something you have done here :thumbsup:.

One thing I have always thought to do is to compare drivers by comparing them with their team mates and future / past team mates of their team mates. The mesh you create by doing this can cover most drivers that have ever raced F1. You can see how Vettel -> Webber -> Coulthard -> Hakkinen -> Senna although that is quite a weak chain with Hak & Senna racing each other in McLarens only 3 times). I'm sure there is something in the chess grading system (or the like) that could be really powerful here, for next time ;)
 
thanks jez101 ... you have summarized it pretty damn well ...

are you suggesting a six degrees of separation from Rubens Barrichello !?! ... just kidding ...
 
UPDATE

I have added another measure ... Points Finishes => Strike Rate => Rank ... adds a small reward for point scoring consistency / effectiveness

I have changed the Blended Rank Sum for each Strike Rate category to:

Avg Points per race competed in = 20%
Wins = 20%
Podiums = 20%
Poles = 20%
Fastest Laps = 10%
Points Finishes = 10%

Top 50 attached 1981 to 1984 points system 964321 all scores count

The rankings change up a little ...
 

Attachments

  • Avg Points per Race 2013 11 (CTA) - 8184 top 50 (R1).pdf
    229.1 KB · Views: 178
  • Avg Points per Race 2013 11 (CTA).xls
    985.5 KB · Views: 172
Back
Top Bottom