Hamilton unlucky or not: how about a statistic model.

[Phil]

That's a very excellend post there - in fact you beat me to the points about the definition of luck. I was going to write the same thing, more or less. Nothing really much to add there.

If you want to nullfy the element of luck where tge drivers are concerned, might as well have them build their oen car using their own organised/built materials. Then they can be held fully respobsible and in control of what they race and what failures they receive.

Since this isnt what it is, motorsport is still very much a team effort. Win as/with the team, fail with them.

There are some arguments that Lewis could be heavier on his car/equipment because he's an agressive driver (opposed to the likes of say Button who are known to be smooth drivers). That may be true, and it may be also true that because of that, that he gets his increased share of failures. I however dont see how you could put that blame or responsibility on him - after all, as a team and especially in todays F1, the team has hundreds of sensors and computers to monitor and simulate everything on the car. If a driver is using the car beyond its intended scope and limits, it should show up eventually and its the teams respobsibility to inform their driver.

We've had various cases of this being highlighted on race weekends - a quick glance at the team radio is sufficient. One example that springs to my mind was i think P3 in Hubgary when Nico was adviced to use a different gear in a particular corner to put less strain on his engine.

These things are crucial and absolutely necessary. The driver relies on the team doing its job properly and to the best of their ability. Up to that point, it's absolutely the drivers job to maximize what ever he can out of his car to be as close to that limit as possible. After all F1 is about finding limits - as a team and as a driver.

Want more driver influence without 'the luck' (from his POV)? Easy: start with reducing the extensive monitoring of the cars and seize all communication while they race. Let them drive dark. This still wouldnt avoid unlucky failures, but as drivers they would need to rely less on team information and drive more cautiously. Or as i said above; let them build their own cars too.

If we do statistics to analyze if Hamilton suffers more failures, it would also be nice to include some sort of nunbers how much potential points those failure cost and perhaps if they realistically were avoidable or just unfortunate circumstances.

Edit: sorry for all the typos. I actually typed this up while sitting at the pool using the 5inch display of my HTC. Yeah, i should get a life, i know...

[Waywardism]

It's far simpler and IMO also more accurate to class all mechanical failures as bad luck for the driver. It's the team's job to ensure the mechanical integrity of the car is enough to reach the end of the race. It is also the team's responsibility, given modern telemetry, to identify any potential risks to mechanical integrity and give appropriate instruction to the driver. So I can't see how a destructive driving style can play a factor in this, if it is/was prevelant then the team really should have addressed the issue long before the 1st race.

Good luck also has to be factored in and should cancel out any accumulated bad luck and vice versa. So if a mechanical failure for the driver is bad luck, then any positions gained due to mechanical failure for someone else has to counted as good luck.

In such a complex event as an F1 race there are many things that can be influenced by luck, if you really want to answer whether driver X has been more or less lucky than his peers then you need to address all of them. Safety cars are massive luck factor at the moment, as we saw yesterday in Hungary. Weather can be a huge luck factor aswell for very similar reasons (if the heavens suddenly open up just as you pass the pit lane entrance you're screwed, for example). Strategy decisions made by the team have to be counted as luck, unless there is evidence to suggest that the driver made the call. Red flags can be lucky too (Monaco 2011 springs to mind with Vettel on dead tyres almost certainly going to be passed by Alonso and Button then we get a red flag which of course allows a free tyre change). I could go on.

If you want to just take into account DNFs and points lost/gained because of them then that's fine but it's only one aspect of luck, it won't really answer the initial question.

[SidSidney]

I am really happy to see that at least a couple of people understand what I am driving at and are willing to consider it seriously. I may not have a fully-functional idea of the framework but I am sure we can get somewhere on this, it's not that difficult.

I have been thinking about how to identify / isolate lucky events. I think it is fairly easy to calculate completely systemic luck ie., the mean outcome of a purely random selection, but I am not so certain how to say if the additional pos. or neg. result is down to skill or some idiosyncratic luck.

I did an quick sample analysis of 2012 last night to get to the two traits you would want to look for in drivers (see post top of page 5), and you can quickly see that the main outlier is Vettel - his results are literally 3 SDs above the mean of all outcomes in that season. Hamilton's were only 1 SD above the mean. So clearly Vettel had either a much more effective process or he was much more idiosyncratically lucky, but I am not sure how to determine which it was just yet.

Stradivarius made this statement: "As long as it is not the driver's own fault, I would say that he is unlucky if a component fails." I have been wondering about exactly that, but I am not sure if that is true, or if you can isolate that effect. The entire package of car, team and driver creates the output, championship points. The only way to isolat ethe driver is to look at his career across different teams and cars.

We also don't know if the driver knew about something from testing and decided to risk it anyway, while his colleague knew as well and took the safer route. Some drivers are well known to be lazy testers, some are known to take more marginal decisions, some are boring pedants and some are cocaine-snorting fiends. There's no way to assess that.

This is for me the trickiest part, deciding the framework to make the test. If it is wrong, the whole analysis is open to criticism.

Just to put my personal bias on the table, I don't believe the whole "unlucky" story, I think it is a preparation issue. I mean, when something goes wrong on his car he is unlucky, but when he comes from the back of the grid to finish on the podium - having hit a wall without any major damage on his first lap after the schoolboy error of ignoring cold tyres and brakes - that is skill and not at least partially luck? I imagine if he had hit the wall properly and knocked off a corner that would have been the icing on the unlucky cake...

I just want to nail it one way or another.

[hollus]

It is going to be very difficult to prove if what happens this season is due to luck, and probably this season has to be taken in isolation to make sense of it.

These are the probability distributions, considering only a single event, like a DNF, after 11 races (Hungary). If the real probability of DNF is 20%, anything from 0 to 4DNFs is quite likely. With a 30% chance, anything from 1 to 5 or 6. There is just too much variability with N=11; and to make it worse we don't know how likely a DNF is for team X, nor, obviously, can we infer it from the observed number of DNFs.



For 19 races it looks a bit better, but still double or triple the number of DNFs of the teammate is just not unlikely, and in real life, is we go to the end of the season, we start running into tactical artifacts, like the extra risk taken simply to add pressure by staying within 1 race worth of points, etc.



If anything is to come out of this, the stats have to be multiseason and multiteam, IMO.

[Just_a_fan]

Stradivarius made this statement: "As long as it is not the driver's own fault, I would say that he is unlucky if a component fails." I have been wondering about exactly that, but I am not sure if that is true, or if you can isolate that effect. The entire package of car, team and driver creates the output, championship points. The only way to isolat ethe driver is to look at his career across different teams and cars.

If you take Hamilton's issue in Australia this year - the component that failed had not failed on any of the Mercedes-engined cars during any running up until that failure. That's the same component across multiple teams with multiple drivers over many thousands of km of engine use in testing and practice. I'd say that was "unlucky" and quite outside of the control of the driver (or, indeed, the team).

[turbof1]

It is going to be very difficult to prove if what happens this season is due to luck, and probably this season has to be taken in isolation to make sense of it.

These are the probability distributions, considering only a single event, like a DNF, after 11 races (Hungary). If the real probability of DNF is 20%, anything from 0 to 4DNFs is quite likely. With a 30% chance, anything from 1 to 5 or 6. There is just too much variability with N=11; and to make it worse we don't know how likely a DNF is for team X, nor, obviously, can we infer it from the observed number of DNFs.

http://i.imgur.com/LmxhvfT.jpg

For 19 races it looks a bit better, but still double or triple the number of DNFs of the teammate is just not unlikely, and in real life, is we go to the end of the season, we start running into tactical artifacts, like the extra risk taken simply to add pressure by staying within 1 race worth of points, etc.

http://i.imgur.com/lm6YL7k.jpg

If anything is to come out of this, the stats have to be multiseason and multiteam, IMO.

I already said it in our PM's, but again very nice job and fully agreed: the issues we currently are having is a too narrow focus on one driver. Data should be gathered from a lot more samples (multi-team & multiseason, as you said). If that can be done, not only can we get more significant results for Hamilton, but we can test every driver included in the sample pool.

[iotar__]

If you take Hamilton's issue in Australia this year - the component that failed had not failed on any of the Mercedes-engined cars during any running up until that failure. That's the same component across multiple teams with multiple drivers over many thousands of km of engine use in testing and practice. I'd say that was "unlucky" and quite outside of the control of the driver (or, indeed, the team).

Not in the races. I'd call it teething problems in the first race of a new engine. Many other drivers in different races had unique engine problems why would this example be special or significant in terms of "luck's" unnecessary dissection?

How would you classify Rosberg's gearbox failure in Silverstone for comparison - another unique across the board problem for the nearest comparison (team-mate): "...but then having a gearbox failure is a bit of a hit. You wouldn't expect in 2014 having a DNF because of a gearbox breaking down. You have to look at the quality aspect of the way we look at things. You can have such a DNF once but not twice." Sounds "unlucky" to me and fits perfectly the theme, the only problem it's a concern of the other driver. Wouldn't that be even bigger lack of "luck" considering how many races gearboxes ran and were tested?

[Just_a_fan]

If you take Hamilton's issue in Australia this year - the component that failed had not failed on any of the Mercedes-engined cars during any running up until that failure. That's the same component across multiple teams with multiple drivers over many thousands of km of engine use in testing and practice. I'd say that was "unlucky" and quite outside of the control of the driver (or, indeed, the team).

Not in the races. I'd call it teething problems in the first race of a new engine. Many other drivers in different races had unique engine problems why would this example be special or significant in terms of "luck's" unnecessary dissection?

It wasn't intended to be "special or significant" it was an example used to make a point. If the failure had been in Rosberg's engine or Button's or...etc. I'd have still used it as an example of the only failure in that component that happened at the one point in time that led to maximum negative impact on the driver's ability to gain championship points. Many hours of running by many drivers in different cars and one driver has it fail at the start of the race. That's "unlucky" for that driver in my view, whichever diver it is.

When I wrote Hamilton's name in my earlier post I guessed you'd jump in on it. I almost said "the Mercedes engine failure in Australia" but I wanted to see if my hunch was right. It was. Well done.

Wouldn't that be even bigger lack of "luck" considering how many races gearboxes ran and were tested?

Which teams run the same gearbox as Mercedes? Not all of the teams that run Mercedes engines also use their gearbox so it's a smaller sample size than the engine component issue. It's still unlucky if he is the only one to have that failure during the entire season out of all of the drivers using that gearbox, yes.

[henry]

I thought it might be interesting to look at pit stops as a sub process in the race.

I wanted to use pit stop time but I couldn't find it so I used pit lane time because it is available on F1.com

I needed a base for comparison. For each race I found the best (shortest) time and then calculated how much longer every stop was than that time.

I edited the data to
: Remove drive throughs, typically 3 seconds faster than any other stop
: Remove any stops more than 10 seconds longer than the best on the assumpton that something other than just changing wheels was happening.
: Deduct 5 seconds from stops which included a 5 second penalty. Help needed here because I could only find Alonso's penalty in GB and I feel there have been a couple more.

I then calculated all the time differences and averaged them per team and per driver.

The results below are the teams ranked in order from least loss to most loss. I also show the difference between team mates.

Code: Select all

Team ID	Ave Diff from best	Ave diff between Drivers
Team 1	 0.680	                0.208
Team 2	 0.847	                0.712
Team 3	 1.043	                0.213
Team 4	 1.057	                0.244
Team 5	 1.143	                0.456
Team 6	 1.156	                0.337
Team 7	 1.643	                1.243
Team 8	 1.946	                0.587
Team 9	 2.118	                0.001
Team 10	 2.250	                0.800
Team 11	 2.476	                0.143
All Teams	 1.466	                0.450

To be honest I am not sure that this tells us much about luck other than drivers would prefer to be in teams 1 and 2 rather than 10 and 11.

I think it might be useful to look at the median stop time to see how skewed the distribution is. Unfortunately I excel doesn't do median in pivot tables. I might do it manually but not tonight.

The drivers list looks like this.

Code: Select all

driver ID	Ave Diff from best
Driver 1	0.491
Driver 2	0.578
Driver 3	0.786
Driver 4	0.922
Driver 5	0.926
Driver 6	0.938
Driver 7	0.991
Driver 8	1.075
Driver 9	1.139
Driver 10	1.182
Driver 11	1.203
Driver 12	1.328
Driver 13	1.378
Driver 14	1.674
Driver 15	1.827
Driver 16	2.118
Driver 17	2.119
Driver 18	2.261
Driver 19	2.318
Driver 20	2.395
Driver 21	2.539
Driver 22	2.627

In this list the numbers themselves are not so interesting but the relative positions of drivers are.

Only one team has two drivers in the first eight. This suggests, at least at the pointy end of this process, the driver makes a difference in pit stop time.

Also in 8 of the 11 teams the driver with the best pit stops is also ahead in the WDC. Nearer 7.5 actually since 16 and 17 are near identical.

The process involves three distinct phases.

1. Driver and car. Brakes to pit lane limit, parks car in box.

2. Pit lane crew and equipment. Change wheels ( and maybe make wing adjustment)

3. Driver and car. Responds to start signal and drives down pit lane on limiter.

The driver obviously makes a difference, particularly in the precision in phases 1 and 2. Maybe this particular characteristic is valuable in modern F1. I'm not sure how lucky that is.

I have deliberately anonamised these figures to avoid "special circumstances" discussions.

If I can find time I will try to go back over previous seasons and look particularly at drivers who have changed team and see if their "luck" follows them round.

[iotar__]

It wasn't intended to be "special or significant" it was an example used to make a point. If the failure had been in Rosberg's engine or Button's or...etc. I'd have still used it as an example of the only failure in that component that happened at the one point in time that led to maximum negative impact on the driver's ability to gain championship points. Many hours of running by many drivers in different cars and one driver has it fail at the start of the race. That's "unlucky" for that driver in my view, whichever diver it is.
When I wrote Hamilton's name in my earlier post I guessed you'd jump in on it. I almost said "the Mercedes engine failure in Australia" but I wanted to see if my hunch was right. It was. Well done.

Certainly, that's why I obliged with an example from his nearest control (team-mate) that nullifies or nullifies and more (depending on interpretation) said Australia engine failure, which you found particularly interesting in defining "luck", in this instance Hamilton's of course. Intention was to create a fuller picture, in case this unfortunate (to avoid using "luck" again) event was omitted and thus results of this rather pointless (IMO) exercise skewed.

[Just_a_fan]

You can't see past the drivers' names - if anyone mentions Hamilton your hatred gets the better of you.

The issue I was discussing was whether it is possible for the driver to be totally blameless in his own "bad luck". I gave an example where it might be considered that the driver was blameless - certainly many commentators at the time thought so. It wasn't about whether Hamilton was more or less lucky than Rosberg in two totally unrelated incidents. Really, because the individual drivers in question are irrelevant in that discussion. They really are.

Now, if you want to try to turn every discussion in to one that allows you to continue to hate Hamilton, that's fine. I shall not be playing your silly game any more.

[Constructr]

It's not luck. It's called Karma. And it can be a real bitch some times!

[SectorOne]

Karma doesn´t exist except in people´s heads so let´s not go into that one.

[SiLo]

I think in some cases the drivers can raise the chance of something bad happening, but the issue occurring in the first place is generally not their fault most of the time.

Like Hamilton in Canada for example, if he had backed off to cool the brakes would he have been more likely to finish the race? Quite possibly.

Whereas in Australia is wasn't really anything to do with him, same for Rosberg in Silverstone. I think Germany and Hungary were not the result of anything Hamilton had done.

Austria I think really was his fault.

My two cents. All my own opinion and conjecture.

[F1Prophet]

This topic is hilarious in itself.

Lewis Hamilton himself stated no long ago – when asked to comment on how terrible and unlucky Michael Schumacher’s accident had been – specifically saying that there is no such thing (as luck) because "everything happens for a reason"...