Can I ask what your objections were to the GT4-style of tuning contests?
That's fine. We were prepared for a scenario like this. We wont be pulling out dont worry. 👍
Phew! I'm glad you guys don't have a problem with this
I was worried, as I'm personally a big fan of your tunes and am really anxious (as I think everyone else is) to see how you guys do in a battle against the rest of GTPlanet's finest!
Well, not so much an objection as it is a personal preference. Here's some background info:
I work in the actuarial/investment profession, so putting numbers to real world phenomena is "what I do". Naturally, someone quantitatively-focused like me (ie. math geek hehe) gravitates to endeavors such as car tuning, to see how the effects of variables (camber, ride height, toe, etc) affect lap times.
Back in the GT4 days, I was only on GTVault and if I wanted a good tune I just went to Fumes! Here, on GTPlanet with so much tuning talent and what looks like a great deal more people actively looking for tunes, I figured there needs to be a way to remove the wheat from the chaff in a systematic, and mathematically-sound manner. This leads me to:
Tuner Garage Battle Thesis
In statistics, we use the traditional hypothesis testing methodology. I'll save you the lecture, but basically you change an independent variable (the tune of the car), observe the dependent variable (lap time, as it "depends" on the tune) and your hypothesis either fails or is accepted after normalizing the result and seeing it if falls within a band of acceptable outcomes. However, we don't care if a tune produces a lap time below X'XX.XXX so we can simplify this, skip normalization, and just use the dependent/independent observations to rank a set of variables.
However, all of this hinges on one, simple fact. The number of observations (or population/sample) must be greater than 60, on average, for the result to be statistically significant. The fewer observations we have (lap time submissions), the less meaningful the result will be. The great thing is, we have an excellent community here it seems and a whole lot of people volunteered to test the tunes! Therefore, this method now has even more relevance and the result of these Tuner Garage Battles will be absolute: "The winning tune of the monthly TGB, on average, is the quickest."
How do you decide between tunes with almost identical average lap times?
What I'm proposing is a three round system. The first round, all of our drivers will submit 10 laps for each tune (I'm going to confirm this with our drivers shortly to see if this is too much of a burden or not). This will result in about 200 laps completed, per tuner submission, a very very statistically sound number! I will compile and rank the result accordingly. All but the top 4 tunes (based on mean and median lap times) will be removed. In this second round, 1 will face 4, and 2 will face 3. The drivers will then submit an additional 10 laps for these 4 tunes, resulting in an additional 200 laps for each tune for a total of 400 laps. As "n" (the number of laps submitted per tune) increases from 200 to 400, any difference in lap times is more concrete and relevant. Finally, in the last round where the #1 and #2 tunes face off, an additional 200 laps will be submitted, bumping the grand total of laps driven for these two tunes up to 600. With that sort of n-variable, there is absolutely no doubt that whichever tune is quickest (even if by a small margin), is indeed the quickest without any shadow of doubt. Basically, the statistical "strength" of each victor will increase as they make it through the rounds. If you think about it, it's exactly like other playoffs in sports. The team that wins the championship has proven, through more minutes of actual play against similarly performing teams, that they are the best.
What about drivers with different racing prowess?
The good thing is, we're looking for relative ranking here, not absolute. The winning tune will be the quickest compared to their rivals, not the quickest against some absolute benchmark. The reason why driver talent is irrelevant, is because each driver MUST submit 10 lap times for ALL tunes. That way, the quickest tune will still win relative to the others, all that changes is the absolute value of the lap times (1'41 is still quicker for a slow driver than 1'43, just as 1'31 is still quicker for a fast driver than 1'33). If this doesn't make sense, open up excel and do some quick tests. You should see that the theory holds. I also think its important that we do have a mix of driving talent. A strong tune should be quick in the hands of nearly everyone. It should be quick and forgiving.
But that isn't the point. For us, it's to get as much feedback as possible for each of our tuning styles; since we both tune slightly differently, it's good to see where each of us can possibly improve.
If you look at the first post I made in this thread, you'll see that the #1 benefit which I listed is for the general populace of GTPlanet. I'm afraid that if you volunteer your garage, it's a rather selfless-act. Tuner garage participation will serve to see how your best efforts rank against others, NOT for your benefit through increased feedback.
If this sort of volunteering is an issue and anyone wishes to drop out, then please let me know. But I stand by my decision that the main benefactors should be the people who use the tunes, not the people who make them. Each garage has their own tuning thread for those purposes.
Psychot|k, if you intend on remaining with a single track to test per month, I see no need to limit the tunes.
If you end up doing multiple tracks, then I support the idea of a limit, for time purposes.
Well, if we add to the number of tuners, the first round of testing becomes more and more arduous on our drivers. It means an extra 10 laps they must submit ON TOP of the 120 or so they already need to. Again, I'm going to see just how much of this is reasonable, maybe I'll bump it down to 5 laps, but 10 is a much better number.
If no one has a problem completing 120 laps using 12 tunes in a span of about 14 days or so (test one tune per day), then sure! I'd be happy to increase the numbers of tuners! But until I know that, I don't want to make it even harder on our drivers (yourself included) who have so kindly volunteered their driving skills and time.
Sign me up mate...Im new but willing to give it a crack! 👍
Vengence Tuning
Consider yourself signed up
I saw in the poll thread that Project41 really dislikes how this has been going, so I can only assume he wants no part which I can completely understand. I signed you up in his spot.
You seem to be the Cinderella story here from what everyone is saying, let's see what you can do!
