17 replies to this topic

[A.Fant]

This is what I predicted for the Hungarian Grand Prix:

 

I was intrigued by the percent at full throttle graph, so I did a regression on m/s and percent at full throttle in dry qualifying and used that model to predict future qualifying results based on track length and percent at full throttle. This is under the assumption that car quality has been constant throughout the season though.

 

I used f1fanatic.com for track length and throttle percent figures, as it was the only site that had them for every track (bar Sochi).

 

This is what I got for Hungary with f1fanatic's 59% full throttle:

 

Mercedes: 1:17.632

Red Bull: +0.698

Williams: +0.736

Ferrari: +0.980

McLaren: +1.148

Force India: +1.419

Toro Rosso: +1.499

Lotus: +1.904

Sauber: +2.432

Marussia: +3.379

Caterham: +3.905

 

Using UBS' 49% this is projected:

 

Mercedes: 1:23.551

Red Bull: +0.652

Ferrari: +0.941

Williams: +1.180

McLaren: +1.364

Toro Rosso: +1.583

Force India: +1.618

Lotus: +2.131

Sauber: +2.728

Marussia: +3.637

Caterham: +4.319

 

The 2nd projection seems more reasonable considering that Hamilton's pole from 2013 was 1:19.388.

 

The model predicts that Mercedes will have at least 0.5s gap to their closest competitor at every GP bar Italy and Brazil where Williams will be closer (+0.242 and +0.362 respectively). Williams will have at least 0.3-0.4s down to the 3rd best team at every GP bar Hungary and Singapore where they will drop to 4th best team.

 

Most teams apart from RBR have been catching up to Mercedes though (esp. Williams) so it could become closer than that.

 

These are the actual results compared to what the model predicted:

 

So while the model was surprised by the general pace of the field, it was actually quite accurate in predicting the gaps to Mercedes. Only two teams were more than 0.4s off the prediction (Williams 0.541s faster, and Lotus 0.491s slower).

 

It was also quite accurate in predicting the order of the teams, though it switched around two pairs of teams (Williams/Ferrari and Lotus/Sauber).

 

Key: Actual/Predicted

 

1. Mercedes/Mercedes

2. Red Bull/Red Bull

3. Williams/Ferrari

4. Ferrari/Williams

5. McLaren/McLaren

6. Toro Rosso/Toro Rosso

7. Force India/Force India

8. Sauber/Lotus

9. Lotus/Sauber

10. Marussia/Marussia

11. Caterham/Caterham

 

As the season goes on, the model should become more accurate as the data sample increases.

 

This is the prediction for Belgium (7004m, 72% full throttle):

 

 

The gaps are larger because of the long lap. Williams are predicted as being 0.42% slower than Mercedes in Belgium whereas Red Bull was 0.59% slower in Hungary (and was predicted to be 0.78% slower).

 

The predicted pole time is of 1:53.613 is 6.040s slower than Button's 2012 pole time, suggesting that the 72% full throttle might be a conservative estimate. I think the prediction will be off in terms of absolute times, but it will be interesting to see how accurate the predicted gaps to Mercedes will be.

 

These are the predicted gaps to Mercedes for the rest of the season (in terms of absolute lap time):

 

I will continue to post the comparisons between the model and the actual results in this thread after each race. I will also update the model with the new data after each race to improve the predictability.

[MortenF1]

Very interesting. 

I believe the Red Bull will be closer than that in Singapore though, and furthermore I think the updated McLaren will be quicker than the Red Bull around Spa. (but the model doesn't do estimates based on such guesswork)

Edited by race addicted, 14 August 2014 - 17:07.

[George Costanza]

1 min and 53 seconds? Yikes that's slow... The pole in 2002 was 1 min and 43 seconds..... (1:43.726) by Michael Schumacher. What a lap that was.  That pole lap in 2002 was unbelieveable; 2 years earlier when Mika got pole, it was 7 seconds slower!

 

In fact the last place in 2002, 1 min and 48 sec, would have gotten pole in 2000-2002 and of course in 2013 and will be in 2014. 

 

Shows you how "slow" F1 has become.

Edited by George Costanza, 14 August 2014 - 17:13.

[scheivlak]

1 min and 53 seconds? Yikes that's slow... The pole in 2002 was 1 min and 43 seconds..... (1:43.726) by Michael Schumacher. What a lap that was.  That pole lap in 2002 was unbelieveable; 2 years earlier when Mika got pole, it was 7 seconds slower!

 

In fact the last place in 2002, 1 min and 48 sec, would have gotten pole in 2000-2002 and of course in 2013 and will be in 2014. 

 

Shows you how "slow" F1 has become.

Don't forget that the fastest Q time last year was 1.48.296 on a drying/green track in Q2. If it had not started to rain again I guess 1.47 was certainly possible in Q3.

 

Certainly interesting to see what times will be this year - of course it's Spa so it might well be raining again!

Edited by scheivlak, 14 August 2014 - 17:26.

[D.M.N.]

1 min and 53 seconds? Yikes that's slow... The pole in 2002 was 1 min and 43 seconds..... (1:43.726) by Michael Schumacher. What a lap that was.  That pole lap in 2002 was unbelieveable; 2 years earlier when Mika got pole, it was 7 seconds slower!

 

In fact the last place in 2002, 1 min and 48 sec, would have gotten pole in 2000-2002 and of course in 2013 and will be in 2014. 

 

Shows you how "slow" F1 has become.

 

Except you're comparing two different circuit layouts, so apples and oranges.

[A.Fant]

1 min and 53 seconds? Yikes that's slow... The pole in 2002 was 1 min and 43 seconds..... (1:43.726) by Michael Schumacher. What a lap that was.  That pole lap in 2002 was unbelieveable; 2 years earlier when Mika got pole, it was 7 seconds slower!

 

In fact the last place in 2002, 1 min and 48 sec, would have gotten pole in 2000-2002 and of course in 2013 and will be in 2014. 

 

Shows you how "slow" F1 has become.

 

Don't forget that the fastest Q time last year was 1.48.296 on a drying/green track in Q2. If it had not started to rain again I guess 1.47 was certainly possible in Q3.

 

Certainly interesting to see what times will be this year - of course it's Spa so it might well be raining again!

 

Except you're comparing two different circuit layouts, so apples and oranges.

The model hasn't been very good at predicting ultimate pace - likely because it can't differentiate between slow/fast corners, just full throttle/not full throttle. But it has been good at predicting the relative pace of the field depending on the type of circuit.

 

So don't focus on the 1:53.613, the pole time will be faster than that if it is dry, but rather the gaps between the teams as the model has been good at predicting those.

Edited by A.Fant, 15 August 2014 - 09:02.

[PayasYouRace]

 

 

 

 

The prediction for Williams at Singapore is looking pretty grim.

[travbrad]

The prediction for Williams at Singapore is looking pretty grim.

 

They can always hope their competitors fall asleep with how much of a snooze fest Singapore usually is.

[A.Fant]

The prediction for Williams at Singapore is looking pretty grim.

The model predicts future results depending on the team's performance relative to the FT% of the different tracks. Most tracks are in the 60-70% range, but Monaco is at 30%. Since Williams did awful in Monaco quali, it has skewed their projections at slower tracks. The Hungary results shaved ~0.350s off of Williams' projected lap time in Singapore, so the earlier projections were even more pessimistic.

 

In Hungary they were projected to be 4th fastest 1.180s behind Mercedes. In reality they were 3rd 0.639s behind. Hungary is at 49% FT, while Singapore is at 48% so the characteristics are similar though Singapore has a longer lap.

 

I think they will outperform the model in Singapore but still be behind Red Bull, though likely ahead of Ferrari.

Edited by A.Fant, 15 August 2014 - 11:25.

[SamH123]

Did we learn much from Belgium?

A small ? over whether Williams are as much quicker than RB as previously thought maybe

[HoldenRT]

A wet qualifying spoiled it, and the nature of the track is hard because of aero.. for example Redbull had a low drag config and went well with it (unexpected and these statistics couldn't have predicted) and some teams had more wing or less etc.  Even Ferrari unexpectedly improved despite having not so great top speed.

 

You have to factor in that teams are always developing and improving things, so it makes it hard.  Monza is a one off track for example, so they will all run a one off aero config for this track and then go back to the usual for future races.  So whatever happens this weekend, it's hard to conclude that it's a constant long term thing, it could be a one off.

 

The statistics have been a very good 'general predictor' though.  Would have been interesting if Spa quali had have been dry.

[MortenF1]

I think it was definitely on to something again, looking at third free.

[Jimisgod]

They can always hope their competitors fall asleep with how much of a snooze fest Singapore usually is.

I like Singapore. It wasn't a du parade like India or Valencia. I can't even remember one single moment from an Indian GP.

[A.Fant]

I've found that the Average Speed from UBS makes for a more accurate model than Full Throttle % and have adapted the model to make the projections based on that instead. This seems reasonable as it incorporates the nature of the track better (slow/fast corners).

 

As HoldenRT said, this type of projection cannot anticipate track-specific aero packages so the low-DF configurations of Red Bull and Ferrari especially will likely be underestimated.

 

This the projection I got for the Italian Grand Prix:

 

 

To underline the new model's accuracy it predicted a Williams pole at Spa with a 1:48.190 ahead of Mercedes 1:48.258. The FP3 results were: Williams 1:49.465, Red Bull 1:49.733, Mercedes 1:49.739. Considering that the fastest lap of the race was a 1:50.511 it seems that the predicted pole time was quite reasonable and a lot better than the previous model's 1:53.613.

 

It'll be interesting to see how the model stacks up for Italy despite the unique aero packages.

 

Then there is the fact that Williams severely underperformed in Monaco qualifying and Mercedes underperformed in Austria qualifying (the slowest and fastest tracks with dry qualifying up to this point) which can skew the projection slightly.

Edited by A.Fant, 05 September 2014 - 16:50.

[wrcva]

Nice analysis  .  Brembo Circuit ID Cards (link) have great data on each circuit and by braking zone up to today's Monza FP.   If you have time to enter them, you can add several other independent vars that may further improve your model.  

[DrivenF1]

Good analysis.

 

There are other ways to predict speed based on typical improvements throughout the weekend. Teams like Ferrari and McLaren flatter on a Friday while Williams always make considerable improvements. This is due to a mixture of fuel effect and car settings. Our team at Formula One Analysis performed calculations and put together a piece here.

 

The predicted qualifying order is -

 

 

The article goes into other areas including long run pace etc.

[DrivenF1]

Anyone happy to model Singapore qualifying times ahead of the weekend?

 

I found our model was accurate except the gap between teams making Q3/Q2 was exaggerated due to track evolution. We took best times from FP2 and Qualifying so Force India who have made Q3 very often this year were flattered in the model compared to Toro Rosso, Sauber, Lotus etc.

 

EDIT: there were also some interesting revelations e.g. Monza favoured Merc engines even more in qualifying due to the extra ERS power it could use on multiple laps. The Ferrari and Renault engine were hampered even more significantly than usual.

Edited by Cult, 14 September 2014 - 18:22.

[DrivenF1]

Is there anywhere where the full throttle % is shown for each track? I thought this was within this thread but I can't see it.