Mercedes without Hamilton in numbers

[Gillian]

Fell free to open another thread, JAF. But stop derailing this one. "Done is better than perfect". Please stop asking for perfect.

With respect, he isn't asking for perfect he is asking for better!

You must have some criteria to deal with the outliers, so they don't skew your data. For example last year Lewis was on pole for the styrian grand prix by 1.2 seconds over Max. The Merc was not 1.2 seconds faster, Max had an off on his best lap. Outliers like this occur every season and need to be accounted for in some way, ether via a mathematical formula, exclusion, or referencing some other form of clean data.

secondly subtracting the same amount of time from every track is flawed, because the tracks are drastically different in length. Being a fixed amount of time ahead in Austria (4.318 km lap), is much more significant than being the same amount of time ahead in spa (7.004 km lap). Thus the amount of time subtracted needs to somehow be correlated to total lap time or track length.


Off the top of my head, if you want the results to be more statistically relevant, then you need to do something like the following.

• for any given track look at multiple clean qualifying sessions, from previous year (prior to 2014). say 5 to 9 to get a good sample size. More is better.
• for each session calculate the average lap time of the manufacture that took pole, and the next manufacture.
• Calculate the % difference in lap time based on the previous averages. for example if the average for the manufacture who took pole was 60 seconds, and the average for the second fast car was 60.5 seconds we end up with the following. (60.5-60)/60.5*100 = 0.83% faster
• Take the above number and average it over the 5 to 9 samples.

The above gives you a number that takes into account the following things.

• the length of the track
• the fact that the cars get faster or slower based on rule changes
• The difficulty of the track to drive, or general noise reduction depending on how you want to look at it. Some tracks just have larger pole deltas than others.

You can then use the number above to calculate how much better or worse The the car, Lewis, or his teammate was compared to the historical average.

Something like that is possible but then I would want to do it using an actual database, not spreadsheets. I noted it, thanks for the feedback. Any ideas on tackling the races?

[Just_a_fan]

If the only discussion allowed is to agree with the OP, then the thread should be locked. A thread where discussion is not allowed is pointless and rather contrary to the whole point of a forum.

Not wanting to start a new thread about what you want/feel this thread should be about is not anyone else's problem but your own.

What is there to disagree with about OP? The data is doing the talking here, you can discuss/analyse that data all you want.

We're trying to discuss the data but being thrown back at every turn. Data alone means nothing. It requires context. To "discuss the data" just means we can all just sit here and say "yes, the data is there and it's good data". That's it. Anything you say about it from there on is inference. And inference is governed by context such as personal opinions, biases (we all have them because we're imperfect humans), heck even cultural differences.

I take it that some of the commentators on here have never done any serious work. Anyone who has done a Master's Degree or PhD will know that data is just a small part of any piece of work. There is a whole raft of other stuff - why you've complied the data (the reason for your research), how you compiled the data, looking at other work (that's part of the context), analysis of the data achieved and checking it for soundness and then, finally, discussion of the data and attempting to find some meaning in it - and finding meaning requires reference to other sources and expertise amongst other things i.e. putting it in context.

So just saying "discuss the data" is meaningless.

[dans79]

Something like that is possible but then I would want to do it using an actual database, not spreadsheets. I noted it, thanks for the feedback. Any ideas on tackling the races?

The FIA or Liberty probably has a database, that's I'm sure you probably have to pay through the nose to get access to. If the data can't be manually reviewed, then you need to start looking at much larger data sets, and excluding the bad data mathematically via standard deviation, or a similar methodology.

Races will be even harder, as you have even more contributing factors, like safety cars, laped cars, etc etc. You would most likely need to approach it the same way, large data sets and statistical validation.

[Dee]

If the only discussion allowed is to agree with the OP, then the thread should be locked. A thread where discussion is not allowed is pointless and rather contrary to the whole point of a forum.

Not wanting to start a new thread about what you want/feel this thread should be about is not anyone else's problem but your own.

What is there to disagree with about OP? The data is doing the talking here, you can discuss/analyse that data all you want.

We're trying to discuss the data but being thrown back at every turn. Data alone means nothing. It requires context. To "discuss the data" just means we can all just sit here and say "yes, the data is there and it's good data". That's it. Anything you say about it from there on is inference. And inference is governed by context such as personal opinions, biases (we all have them because we're imperfect humans), heck even cultural differences.

I take it that some of the commentators on here have never done any serious work. Anyone who has done a Master's Degree or PhD will know that data is just a small part of any piece of work. There is a whole raft of other stuff - why you've complied the data (the reason for your research), how you compiled the data, looking at other work (that's part of the context), analysis of the data achieved and checking it for soundness and then, finally, discussion of the data and attempting to find some meaning in it - and finding meaning requires reference to other sources and expertise amongst other things i.e. putting it in context.

So just saying "discuss the data" is meaningless.

Example

2020

Analysis
Data given and sound method used

Discussion
How many poles for driver that is 0.1, 0.2, 0.3, 0.4, 0.5 off in Qualy?
How many wins for driver that is 0.1, 0.2, 0.3, 0.4, 0.4 off in race pace?

Conclusion
Trends show that driver with 0.2 off the pace in Q would still/not get a pole/ x amount of poles etc
Trends show that driver with 0.4 off in race pace would still win X amount of races etc

This is not rocket science, this is analysis of data. We are on F1 technical.

You can come to some sort of conscenious at the end of it, you don't need reference to any outside data/context for this thread.

[Just_a_fan]

Not wanting to start a new thread about what you want/feel this thread should be about is not anyone else's problem but your own.

What is there to disagree with about OP? The data is doing the talking here, you can discuss/analyse that data all you want.

We're trying to discuss the data but being thrown back at every turn. Data alone means nothing. It requires context. To "discuss the data" just means we can all just sit here and say "yes, the data is there and it's good data". That's it. Anything you say about it from there on is inference. And inference is governed by context such as personal opinions, biases (we all have them because we're imperfect humans), heck even cultural differences.

I take it that some of the commentators on here have never done any serious work. Anyone who has done a Master's Degree or PhD will know that data is just a small part of any piece of work. There is a whole raft of other stuff - why you've complied the data (the reason for your research), how you compiled the data, looking at other work (that's part of the context), analysis of the data achieved and checking it for soundness and then, finally, discussion of the data and attempting to find some meaning in it - and finding meaning requires reference to other sources and expertise amongst other things i.e. putting it in context.

So just saying "discuss the data" is meaningless.

Example

2020

Analysis
Data given and sound method used

Discussion
How many poles for driver that is 0.1, 0.2, 0.3, 0.4, 0.5 off in Qualy?
How many wins for driver that is 0.1, 0.2, 0.3, 0.4, 0.4 off in race pace?

Conclusion
Trends show that driver with 0.2 off the pace in Q would still/not get a pole/ x amount of poles etc
Trends show that driver with 0.4 off in race pace would still win X amount of races etc

This is not rocket science, this is analysis of data. We are on F1 technical.

You can come to some sort of conscenious at the end of it, you don't need reference to any outside data/context for this thread.

OK, so on the data carefully presented by the OP, what's your conclusion?

[Dee]

We're trying to discuss the data but being thrown back at every turn. Data alone means nothing. It requires context. To "discuss the data" just means we can all just sit here and say "yes, the data is there and it's good data". That's it. Anything you say about it from there on is inference. And inference is governed by context such as personal opinions, biases (we all have them because we're imperfect humans), heck even cultural differences.

I take it that some of the commentators on here have never done any serious work. Anyone who has done a Master's Degree or PhD will know that data is just a small part of any piece of work. There is a whole raft of other stuff - why you've complied the data (the reason for your research), how you compiled the data, looking at other work (that's part of the context), analysis of the data achieved and checking it for soundness and then, finally, discussion of the data and attempting to find some meaning in it - and finding meaning requires reference to other sources and expertise amongst other things i.e. putting it in context.

So just saying "discuss the data" is meaningless.

Example

2020

Analysis
Data given and sound method used

Discussion
How many poles for driver that is 0.1, 0.2, 0.3, 0.4, 0.5 off in Qualy?
How many wins for driver that is 0.1, 0.2, 0.3, 0.4, 0.4 off in race pace?

Conclusion
Trends show that driver with 0.2 off the pace in Q would still/not get a pole/ x amount of poles etc
Trends show that driver with 0.4 off in race pace would still win X amount of races etc

This is not rocket science, this is analysis of data. We are on F1 technical.

You can come to some sort of conscenious at the end of it, you don't need reference to any outside data/context for this thread.

OK, so on the data carefully presented by the OP, what's your conclusion?

I don't know, I'm lazy. I'm waiting on people to do this instead of talking about other stuff. I want to know the trends. And if people want to analyse the data in different ways and they come to similar conclusions, or they don't, it's all going to be interesting.

[dans79]

Analysis
Data given and sound method used

The soundness of any method used to generate statistical data is always a peroneal opinion and thus open for debate.

I've also had to stand in front of a lecture hall full of Phds and post docs, and defend myself. In my case I had to defend how that data was generated, and then how it was analyzed and used to support the thesis.

[Dee]

Analysis
Data given and sound method used

The soundness of any method used to generate statistical data is always a peroneal opinion and thus open for debate.

I've also had to stand in front of a lecture hall full of Phds and post docs, and defend myself. In my case I had to defend how that data was generated, and then how it was analyzed and used to support the thesis.

If you want to do your own analysis, you are free to do that. As of now, you have data and you know how that data was generated. Not using it or not doing your own analysis your own way is entirely up to you.

[dans79]

Analysis
Data given and sound method used

The soundness of any method used to generate statistical data is always a peroneal opinion and thus open for debate.

I've also had to stand in front of a lecture hall full of Phds and post docs, and defend myself. In my case I had to defend how that data was generated, and then how it was analyzed and used to support the thesis.

If you want to do your own analysis, you are free to do that. As of now, you have data and you know how that data was generated. Not using it or not doing your own analysis your own way is entirely up to you.

I think you might have missed the point of my post. when Someone posts an analysis on a forum, tries to publish something in a peer reviewed publication, gives a talk at an event, the methodology will be challenged just as much as the result/findings. That's how the scientific/technical community works.

[Just_a_fan]

Example

2020

Analysis
Data given and sound method used

Discussion
How many poles for driver that is 0.1, 0.2, 0.3, 0.4, 0.5 off in Qualy?
How many wins for driver that is 0.1, 0.2, 0.3, 0.4, 0.4 off in race pace?

Conclusion
Trends show that driver with 0.2 off the pace in Q would still/not get a pole/ x amount of poles etc
Trends show that driver with 0.4 off in race pace would still win X amount of races etc

This is not rocket science, this is analysis of data. We are on F1 technical.

You can come to some sort of conscenious at the end of it, you don't need reference to any outside data/context for this thread.

OK, so on the data carefully presented by the OP, what's your conclusion?

I don't know, I'm lazy. I'm waiting on people to do this instead of talking about other stuff. I want to know the trends. And if people want to analyse the data in different ways and they come to similar conclusions, or they don't, it's all going to be interesting.

Edit: Lol downvoted, the reasoning being "spamming". I was asked a question and answered it.

I've upvoted you. It wasn't spamming. In my humble opinion anyway.

[Gillian]

Something like that is possible but then I would want to do it using an actual database, not spreadsheets. I noted it, thanks for the feedback. Any ideas on tackling the races?

The FIA or Liberty probably has a database, that's I'm sure you probably have to pay through the nose to get access to. If the data can't be manually reviewed, then you need to start looking at much larger data sets, and excluding the bad data mathematically via standard deviation, or a similar methodology.

Races will be even harder, as you have even more contributing factors, like safety cars, laped cars, etc etc. You would most likely need to approach it the same way, large data sets and statistical validation.

Yep. There is an API I know of which I could use which is accessible by python. I have it on my to-do list to port it to Kotlin (so I can use it in a KMP project). I'm not sure how far back it goes but I'm pretty sure it has atleast 25 years of data. Will look into it.

I have played around with races but there are way more factors to consider in races than qualifying so I haven't felt like posting any of it yet.

I think the qualifying "analysis" posted indeed is very limited and can be improved upon but it certainly isn't as crappy as some make it out to be.

Anyway I want to thank you for being constructive. You are the first to provide actual feedback without making me feel like crap.

[Gillian]

The soundness of any method used to generate statistical data is always a peroneal opinion and thus open for debate.

I've also had to stand in front of a lecture hall full of Phds and post docs, and defend myself. In my case I had to defend how that data was generated, and then how it was analyzed and used to support the thesis.

If you want to do your own analysis, you are free to do that. As of now, you have data and you know how that data was generated. Not using it or not doing your own analysis your own way is entirely up to you.

I think you might have missed the point of my post. when Someone posts an analysis on a forum, tries to publish something in a peer reviewed publication, gives a talk at an event, the methodology will be challenged just as much as the result/findings. That's how the scientific/technical community works.

Good points. I would like to add however that questioning someone's intention is not part of that proces.

I also have the feeling part of the discussion is caused by bad communication. I don't see anything provocative in the thread title for example, but one poster here has complained about it a few times now. English is not my first, nor my second language. I don't know if that makes a difference.

[Just_a_fan]

Out of interest, where does Brazil's qualifying sit in this discussion? Not only 4/10s faster than the nearest competing team, 5/10ths faster than his team mate in the same car. Using the OP's method, does that mean that anyone who was 1/10th, 2/10ths, 3/10ths slower would be on pole? If so, does that mean that Bottas is entirely useless even though only a week ago he took pole?

Or this sort of question not helpful?

[Ryar]

Out of interest, where does Brazil's qualifying sit in this discussion? Not only 4/10s faster than the nearest competing team, 5/10ths faster than his team mate in the same car. Using the OP's method, does that mean that anyone who was 1/10th, 2/10ths, 3/10ths slower would be on pole? If so, does that mean that Bottas is entirely useless even though only a week ago he took pole?

Or this sort of question not helpful?

In Singapore 2016, Rosberg outclassed Lewis by 7 tenths in qualifying! The attempts to drive this analytical discussion, in a meaningless way, is of no value to this thread.

[Gillian]

Out of interest, where does Brazil's qualifying sit in this discussion? Not only 4/10s faster than the nearest competing team, 5/10ths faster than his team mate in the same car. Using the OP's method, does that mean that anyone who was 1/10th, 2/10ths, 3/10ths slower would be on pole? If so, does that mean that Bottas is entirely useless even though only a week ago he took pole?

Or this sort of question not helpful?

In this case the Mercedes team would lose a pole position at 0.5 sec and Red Bull would gain one. That's an indication of a more competitive season. Compare this to the posted data of 2014 - 2016 and you will see that even at 0.5 sec there's hardly any difference in those years.