Time trials - an information science approach

[eran0004]

Hope you like charts and numbers!

I took the liberty to collect and analyze the data of the latest time trials + the FT86 GP event in Japan, all retrieved from Gran-turismo.com today the 18th of November (before the finish of the events)

The charts below are the fastest times for the top 250 drivers for each time trial.

One can see that the curves are the calmest at the right end and becomes more and more upset the further to the left they go. Right before the steep dive that that ends all the curves on the left.

Now, an interesting thing that becomes apparent, is that the curve becomes steeper and steeper at the top range of drivers. As marked below, the difference in time between #1 and #53 at the FT-86 GP is as big as the difference between #53 and #250.
In the NASCAR time trial, the contrast is even larger, #1 to #20 = #20 to #250.
The Trial Mountain event is more similar to the FT-86 event, #1 to #47 = #47 to #250

One theory on why the curves are becoming steeper as the position gets better is that gamers feel more inclined to improve even more the better ranked they are. Someone in the top 10 range might be fighting harder to improve than someone in the 100-200 range. In the top 5, where there's an extremely deep drop in all curves, the fight is also the hardest. This means that the further to the right we go, the flatter the curve becomes, and at the "gold time" the curve would be totally flat for several thousands of positions, and then as we go beyond gold time and further back, the curve would start rising again and once we get beyond bronze time the curve would go rather steep again, for those who haven't reached bronze time and have given up on trying to improve.

This is only a sample from the very top range of drivers. There are several thousands of drivers taking part in the time trial events, and it would be interesting to have access to the entire data list to perform an analysis of the full events and see if the theory above is correct.
It would also be interesting to have access to the lists from old events that are now closed.
It would also be nice to collect data samples ever day during a week and make an animation of how the curve behaves during that time; if it flattens out with time or if it becomes steeper with time.

 

[Forster]

Right... so... how does this affect GT5 in any way?

 

[rhpamman82]

Right......so.........hugely, actually Forster...........

If PD were to track all the TT times by players, when they would be able to predict, with a basic software package roughly what times to expect from GT5 players. And, therefore, the Gold time could be set at a time which is much harder to obtain.

In fact, apart from Drift Trials, the gold standard times have become almost but a joke. Players sign in, do a 10min session, bag the gold with 1/2 million cr, race suit, helmet and call it a day!!!

GT5 is marketed as a racing simulation, is a racing simulation and not an Arcade Game. Gold Standards should be more challenging, but before someone climbs on by back for this, I agree not too challenging so as to put people off the game.

rhp82amman
Thank you erran0004, for an interesting statistical analysis

 

[snowgt]

Actually, I would derive from that statistic, that a) the abilities of drivers are normally distributed, which was to be expected.

The more interesting thing is the fact that b) the NASCAR-graph drops much more sharply towards the top of the leaderboard than the FT-86, which points to the conclusion that good times are a bit more accessible in the FT-86 than in a NASCAR. Which is again, what you would expect as the higher-powered car is more difficult to control.

 

[Energizerrr]

I had a thought about this one today so thanks for putting this up, my explanation is simple - not everyone who completes on seasonal TTs is extremly fast, yes people are fast but not every alien plays every TT.
In seasonal TTs it's more about who spends most time on an event rather than who's actually the fastest (have in mind that I'm not reffering to the Top 5, those are fast).

Competition in WRS > competition on Seasonal TTs.

 

[rhpamman82]

Hay Energizerr,

I don't want to come across as being critical. Perhaps just a chance to agree to disagree. The players who regularly hold top 50, 20 and sometimes 10 places really don't have to try for long to immediately stick a top 20 or 50 time in.

Just with any game or sport. Some people are just better, and know when to brake, find the apex and accelerate at the right point. I think it's everyone who's trying to catch them who spend more time on TTs. However, I do agree that those who are in the top 10 who have a chance of getting top spot would obviously put a lot of time in perhaps!!

It would be nice for some GT5 TT specialist to comment on this and let us know how long it takes to stick in a fast time

 

[eran0004]

I don't think the huge drop in the left end of the curves is because there are 5-10 players that simply is so much better at GT5 than everyone else. It doesn't make sense, there are more than ten thousand people who do the time trials, and that only 1/5000 is capable of achieving the best times doesn't make sense. A more plausible explanation is that most people are happy once the get gold and then they stop trying. However, the better ranking you get, the more you want to improve. For instance, if you get a top ten ranking, then you think that the 1st place is well within reach and so you try again and again and again to improve your time, and eventually you will. That will to improve is not as strong at place 50, and it weakens more and more the further back you are.

A thing that backs up this theory is that the FT-86 event, which is only available in Japan, experiences the same thing. Compare the Trial Mountain curve with the FT-86 curve and they are roughly the same. How is this possible, when the amount of players in Japan, although they are many, is certainly a lot less than amount of players in the entire world? The only explanation is the theory about people being more motivated to keep trying the better position they have.

If we could see the gold, silver and bronze times here we would see that curve would be completely flat just below gold time, because there would be 10 000 people (roughly) who golded it and then left. Then immediately above gold, the curve would rise again, because those who were just tenth's from getting gold did try again and again until they golded it, so there would be very few people at gold time +0.1 to +0.5 (or whatever the actual numbers are and then it would flatten a bit just below silver, rise a bit above silver, flatten out again just below bronze and finally rise again above bronze.

But of course I would need the complete data to be able to prove it...

A lot of skill is naturally involved, but scince this is the top 250 drivers of the world (except from the first chart which is only Japan) I think we can assume that their skill level is pretty equal. That's why I think that the main different is the amount of time they spend on trying to improve their times.

Edit: A reason why the NASCAR curve is straighter than the others might be because those cars are raced stock, which means that different tunings might be what's causing the Trial Mountain chart to be more rough.
Another reason could be that Trial Mountain is a more difficult track than Cape Ring, and harder to produce even lap times on. That could also explain the FT-86 event, because I think that is done in stock cars as well, are they not? Can't remember what track though...

Another thing I just noticed: If you look at FT-86 and Trial Mountain, there are small "peaks" at #60, #30 and #10. This might suggest that a lot of people are aiming on being top 60, top 30 and top 10 and once they reach that they don't try as hard (or at least takes a break to try again later). You can see in NASCAR that top 10 is where the drop really speeds up.

Right... so... how does this affect GT5 in any way?

A bigger sample would have revealed a lot more, but I think this sample reveals quite a lot. Of course, more samples are required to really be able to draw accurate conclusions. For instance, if all Trial Mountain curves are as rough as the one above, then that suggests that Trial Mountain is a difficult track to master. Or if all NASCAR curves are as straight as the one above, that might indicate that the NASCAR's are highly capable of producing even lap times and thus somewhat "easy" to drive, at least for the top range of drivers. These curves in themselves does not say much, but they will say more when compared to other curves.

Data that is collected for the entire range of players will reveal a lot more, and has a potential to be used by PD to fine-tune aspects of the game that might have been hard to notice in other ways (if they have the time, that is ). The more data that is collected, the more and better conclusions can be drawn.

 

[DixonFan]

Interesting graphs (I quite like maths), I think it is a case of there being a few freakishly fast people. I thought firstly it could be the difference between those who race to the spirit of the regulations vs. the detuners, but pretty much everyone in the top 250 seem to be in the same (detuned) car.

 

[zocker_16]

Hi eran0004 i like charts and statistics very much. I even used a system quite similar to that in GT5P to find combos which where easy to get into top 10 position.
I think it would be very interesting to compare these charts with the one from the last GT-Academy: http://eu.gran-turismo.com/gb/academy/2011/ranking/
I think there are clear differences to the just for fun events.
Can you use your tool on that ranking?

I´m sure the chart for all drivers would nearly look like this

 

[eran0004]

Hi eran0004 i like charts and statistics very much. I even used a system quite similar to that in GT5P to find combos which where easy to get into top 10 position.
I think it would be very interesting to compare these charts with the one from the last GT-Academy: http://eu.gran-turismo.com/gb/academy/2011/ranking/
I think there are clear differences to the just for fun events.
Can you use your tool on that ranking?

I´m sure the chart for all drivers would nearly look like this

It's still only the top 250 drivers though, but for each country, so that makes a total of 1500 drivers. Not gonna be accurate though, as it's not certain that driver 250 of Spain (who made it to the list) is faster than driver 251 of France (who didn't make it to the list). Gonna take some time to filter through it and delete the double posts, but luckily I have just the program for it

I'll post the result here once I got it.

 

[duncan77]

Maybe it's just me, but I feel like those axes should be reversed to make it more intuitive. I just turned my head sideways.

 

[eran0004]

Here are the GT Academy charts

First off, each region on it's own (UK & Ireland; France & Switzerland; Italy; Germany & Austria; Spain & Portugal; Netherlands)

And here all regions put together (notice that there's a substantial trend change for drivers positioned >1000, which means that there probably were faster drivers in other regions that should have made it to the list if it was a region-free top 1500 list)

Edit: Just realised that the 1000 point where the curve climbs again happens at the same time as the French and Spanish curves ends (2:18,9 seconds) which means that there should be mostly 500 French/Spanish drivers at position 1000-1500, which didn't make it to the list as they were not top 250 in their own region.

And the second bump at 2:19.8 happens when there are only Netherlands left...

Maybe it's just me, but I feel like those axes should be reversed to make it more intuitive. I just turned my head sideways.

That would make it a really tall curve, or else it would be hard to see the positions. And it's easier to work with images on a computer when they're horizontal rather than vertical But maybe I should make a printer-friendly version where all text is 45 degrees angle, so you can turn it whatever way you want after printing...

Cause you all print and save these charts I hope?

Edit 2: Also, the curves doesn't say that the French and Spanish are the best GT5 players, and that the Dutch are the worst. There are more players in France then in the Netherlands, that's why there are more fast drivers there than in the Netherlands, but there are surely also more slow players in those countries than in the Netherlands, it's just that the slow players doesn't make it to the top charts that often

 

[Nicktune]

Great work Eran0004!!
Just an idea...It would be interesting to see how much the top 250 has changed/evolved (2011-11-18 vs the last day of the seasonal event(s)).

 

[PR1VATEJ0KER]

This is Cool! Excellant work, quite interesting

Here is a small amount of data, that may help plot some of the curve for each TT23,
By looking at my friends rankings through gran-turismo.com.
I can see;

TT23 Nascar Event
Rank Time
375 0:48.443
20607 0:51.923
22051 0:51.971
29797 0:53.184
41302 0:56.998

TT23 Trial Mountain
Rank Time
1896 1:39.704
7606 1:43.792
17918 1:47.223
23363 1:48.739
26933 1:49.859
28543 1:50.392

Taken at time of Post

Another way to get data about the slower times, would be to drive around at a slow pace, going slightly faster each lap, reading and noting the position shown in the bottom left hand corner
Alot of work for that though!

 

[lunarwolf]

interesting. but maybe too early to analyze since events haven't been on for long enough. at least wait for the end of the weekend since i am sure many only get to play on week ends

also the Nascar may be more relevant as the cars are the same. whereas in trial mountain, cars will varry a lot more (although most at the top seem to always use the R390 and some the NSX LM prototype road cars - I didn't but then again i just wanted to get the cash from gold and then stopped as i find time trials boring)

 

[eran0004]

Great work Eran0004!!
Just an idea...It would be interesting to see how much the top 250 has changed/evolved (2011-11-18 vs the last day of the seasonal event(s)).

Absolutely. Do you know when the last day of these events are?

Another way to get data about the slower times, would be to drive around at a slow pace, going slightly faster each lap, reading and noting the position shown in the bottom left hand corner
Alot of work for that though!

Well, the pace could inrease by a second each lap to get a rough idea about the shape of the curve. Good idea, anyone wanna try that and post the results? I'm nowhere near my PS3 at the moment

 

[Monatsende]

Well, to add something that goes with your theory. I take time trials to earn money. If the event is goldened it`s over for me. No need to push for a better time because I don`t have the motivation (and skills!!!) to enter top 100 or 250. I "feel" no difference to be 754th or 1,078th or 6,721th or 23,548th... If I have 100% cash I feel good enough.

 

[eran0004]

Some brand new data straight from the oven, and it's full of stars!

This is the Trial Mountian time trial, the chart shows each position and the date/time when it was set.

If we look at the latest period (12.00 - 15.00 today) we can see that there's a lot of activity in the top. 12 of the top 50 times (nearly 25%) were set during the latest 3 hour period.
If we count the latest 24 hour period (from 15.00 yesterday to 15.00 today) I find that 37 of the top 50 times were set during that period.

Counting top ten, only three times survived from yesterday, the rest were set today.

Positions 1-60 seems to be the ones who are improving the most (most times are fairly new), then positions 60-160 seems to be sitting rather safe, a lot of times there from Thursday and not so many times set today. From 160-250 the spectrum goes wide, with a lot of old times and a lot of new times.

Edit: And yes, this chart should probably be flipped so that date/time is shown on the horizontal axis. Or else it's gonna be a really tall chart by the end of the event.

 

[Nicktune]

Period of Availability: 2011/11/16 22:00 - 2011/11/30 22:00

 

[eran0004]

Here is an updated version of the Date/time chart, that can be updated throughout the event.

 

[noisiaturismo]

I don't understand it, mainly because there is no clear explanation of whats in the graph.

MY TIME TRIAL GRAPH:

 

[eran0004]

Here is a GIF animation of how the Trial Mountain time trial has developed today, from 10.00 to 18.00 UK time. The data was collected once every hour.

The vertical scale is the position of the lap time, and the horizontal scale is the date & time of when the lap time was set.

If you look at top 20, there is a total of 13 new times set there during the period. It's pretty active up there

Top 5 is even more extreme, I think there's 6 - 8 new times set in that range during the period. And by 18.00, all top 5 times are no older than 4 hours!

Will be interesting to do the same graph one week from now and see how the pattern has changed. It will be a lot harder to set a new top 5 time by then.

And I hope that people who finds graphs boring (noisiaturismo?) has better things to do than hanging around in this thread, because it's obviously not for you...

 

[PR1VATEJ0KER]

This is awesome, thanks again!

 

[noisiaturismo]

I don't find graphs boring, you don't give a clear explanation.

 

[eran0004]

I don't find graphs boring, you don't give a clear explanation.

What do you want to have explained better?