Sebastian Vettel, Ferrari, Albert Park, 2018

Mercedes explains software bug and confirms it cost Hamilton win

2018 F1 season

Posted on

| Written by

Mercedes has explained how a software error cost Lewis Hamilton the lead of the Australian Grand Prix and confirmed he would have been able to win the race.

Hamilton’s chance to win the race disappeared when Sebastian Vettel got ahead of him by making a pit stop during a Virtual Safety Car period. Mercedes has confirmed it found an error in a piece of software which it used to calculate how close Hamilton needed to be to Vettel to ensure the Ferrari driver did not come out ahead.

Safety Car, Albert Park, 2018
Analysis: Mercedes’ missing seconds that cost Hamilton victory in Melbourne
“The issue isn’t actually with the race strategy software that we use,” explained trackside engineering director Andrew Shovlin in a video released by the team (below). “It was an offline tool that we create these delta time laps with. And we found a bug in that tool that meant that it gave us the wrong number.

“The number that we were calculating was around 15 seconds. In reality the number was slightly short of 13 seconds. So that was what created our delta, that was where we thought we were safe, we thought we had a bit of margin, and obviously you saw the result: we dropped out and we were in second place, it’s very difficult to overtake and we couldn’t get through.”

Shovlin also confirmed the team had enough performance in hand that they could have told Hamilton to push to hit the real VSC delta target. However he had been told to back off to save fuel and tyres.

“The simple answer is that we could have pushed the car harder,” said Shovlin. “At this point of the race we were still following a fuel profile that meant we were having to do some saving with fuel. Melbourne’s a very difficult race to do on that 105kg limit.

“And also we were being a bit cautious with the tyres. Because the team thought that we were safe anyway, we thought that Sebastian was going to come out behind us, we weren’t telling Lewis to push any harder.

“That’s really one of the big frustrations we have coming away from Melbourne is the car was clearly quick enough to win the race, if we’d managed the race differently we could have won it. But at this point we were lulled into this false sense of security and we didn’t have Lewis driving as fast as we could.”

2018 F1 season

Browse all 2018 F1 season articles

Author information

Keith Collantine
Lifelong motor sport fan Keith set up RaceFans in 2005 - when it was originally called F1 Fanatic. Having previously worked as a motoring...

Got a potential story, tip or enquiry? Find out more about RaceFans and contact us here.

57 comments on “Mercedes explains software bug and confirms it cost Hamilton win”

  1. Vettel fan 17 (@)
    28th March 2018, 21:58

    The VSC window was 13 seconds, but wasn’t the gap between Vettel and Hamilton 11 seconds? So shouldn’t Hamilton come out ahead.

    1. @vettelfan17
      Yeah, I was going to ask the same. According to the lap charts, it was just over 11 seconds when they crossed the finish line on lap 25, and Vettel pitted at the end of the next lap. I’ve heard people mentioning that the gap ballooned to up to 18 seconds behind the VSC, but I know of no reliable sources for that claim.
      11 seconds, 13 seconds, 15 seconds, 18 seconds – the more I hear about this, the more confusing it gets.

      1. Vettel fan 17 (@)
        28th March 2018, 22:37

        From what I see from the lap charts, I think he lost ten seconds due to the pit-stop and the gap was 11 seconds as you stated. So in theory he should have came out a second ahead, if I’ve thought this through correctly.
        I’ve also heard the gap went up to 18 seconds, but that wouldn’t add up with the 13 second pit-stop delta, since Vettel should have then come out 5 seconds ahead of Hamilton. But I don’t know all the information so maybe I’m wrong.

      2. @vettelfan I think if the vsc gap they needed was 13 seconds Hamilton would have needed a bigger margin eg 15 seconds. In theory if he had an 11 second gap he would come out 2 seconds behind.

      3. If I’m getting it right, these 13 or 15 seconds are the gap under racing conditions to be covered for a possible VSC. The gap is certainly going to increase substantially, to 18 seconds or probably more, once the VSC starts, but the numbers Shovlin discusses only apply in racing conditions.

        Still no answer to why they allowed the racing conditions gap to fall to 11 seconds if they thought the “safe” margin was 13, though.

        1. This. Actual time gaps (as in how long it takes for car #2 to pass the mark where car #1 was) increase when the race is slowed down with VSC, while the distance gaps stay roughly the same. Then in theory once the race resumes that should work out to the same time gaps as they had before, but it does matter in what part of the track the VSC is turned off again, because if that’s on a straight you lose more time speeding up (compared to another driver restarting in a low speed chicane) right?

          1. What are you talking about? A 10s gap at 300KPH is a 10s gap at 100KPH.
            The distance between the cars, in meters, is what will change at different speeds.

            As for the second part of your comment, you are indeed correct, going from VSC to green in a chicken is much better than in the middle of a long straight. It’s a very good point against VSC, but doesn’t apply to the events at Melbourne last Sunday.

          2. actually no; a 10s gap at 300kph is a 30s gap at 100kph. The distance is the same, speed is reduced, time to cover that distance increases. At least with this universe physics laws.

    2. Looking at the place per lap chart Lewis lost First place between the 18th and 19th laps.
      https://www.racefans.net/2018/03/25/australian-gp-interactive-data-lap-charts-times-and-tyres/

      1. But isn’t that just the lap on which he pitted @drycrust? Hamilton did effectively lose a second in the pitstop, bt the mistake was made in laps 22 to the first VSC lap, where you can see Hamilton slightly lowering his pace (line goes up) , when he should have been keeping it constant.

    3. In my opinion, it was due to the fact that during the VSC, the lap time of the cars inside the track went down around nearly half of the normal speed. Whereas the time needed for the cars to pass the pitlane stayed as is.

      To be fair, the speed of the cars inside the pitlane need to go down as much relative to the normal pitlane speed, or shut the pitlane door completely.

      1. Yes the speeding up in pit entry seems very out of place during a VSC period (for good measure: the accident could be right around that area). Why is there no arbitrarily imposed delta in place from the VSC line to the pit entry line?

    4. It’s not hard to understand. Vettel had an 11 second lead over Hamilton when the VSC was deployed because he hadn’t yet pitted. The VSC enableded him to pit without losing the lead. Simple explanation: Vettel’s lead of of 11 seconds at racing speed, say 100 mph, was about 550 yds. If the VSC slowed the field to 50 mph that 550 yds. takes 22 seconds to travel, which is about the pit lane delta time. Voila.

  2. Next time they’ll think twice about using software that has a black prancing horse on the box.

  3. It’s an explanation but even so it’s based on a questionable strategic logic. Mercedes were fuel and tyre saving when there was a risk of Vettel coming out of the pits ahead. So really they should have been focusing on staying ahead as a priority and fuel saving later. Losing the position meant Hamilton undoubtedly faced more tyre wear and engine stress, and maybe fuel use (I don’t know) than he would have done had he emerged from the VSC in first. So the issue seems to be incorrect prioritization and over-confidence on data projections.

    1. You beat me to it @david-br !

      Seems completely crazy that they’d risk the win to try save the car, tyres and fuel on laps where the car was already having an easy time under VSC speed.

      1. @david-br, @Ben indeed, Mercedes seems to have lost the race not so much due to a software issue or the VSC, but rather by focusing on the wrong thing at the wrong time, which is the typical way teams lose on strategy in F1.

        Good, they will hopefully not do that again, and I am sure next time Hamilton will strive to be extra sure before backing down upon request!

      2. Completely agree @david-br you’d think they’d be even more likely employ your logic after the mistake in sacrificing track position last year

    2. I think you got that one wrong, guys, Mercedes didn’t ask Ham to save anything when the VSC was out. An F1 running at VSC pace is already saving everything, it’s not even possible to save more. Mercedes asked Ham to save tyres and fuel BEFORE the VSC came because they thought they had a possible VSC scenario covered.

      1. Yep, that is true@alonshow, the team got Hamilton to back off before the VSC came, and then could not have reacted during it, even if they had realised their strategy error.

      2. Indeed, the error was under racing conditions before the VSC, where they should have been accounting for the hypothetical scenario of a VSC. In most cases this strategic error would go unnoticed to the viewers and unpunished in the race, but in this race it did manifest itself when the VSC actually occurred.

  4. Times aside, it seems stupid not to push as much as the rules allow. The teams go all out and spend God knows how much on pit stops to gain 0.1 of a second, yet while in an actual race situation tell the driver to go slower than he’s allowed to ? How much stress on can running maximum VSC speed for a few laps really put on the car, fuel usage or tyres ?

    Back to the numbers, so they thought they were safe with 15 seconds but only had 13. A two second gap seems too small to start playing games with. The cars are already super slow under VSC so what’s the point in slowing even more with only a 2 second gap ? The more I read the more I realise Mercedes really shot themselves in the foot here. Had they actually raced flat out rather than trying to be too clever they’d actually have saved the car more and won the race.

    1. Was it Fangio who said that you try to win in F1 as slowly as you can. I know Lewis repeated it after the heated China 2015 race.

  5. Why did they need software? Before the race did they not have pre determined points for safety cars and vsc? Surely the pit wall would know they need to be over x amount of seconds to be safe? Also why would you not push like crazy during the pit stop phase to maximise until all round of pit stops were complete no matter what?

  6. Maybe there’s a place for something like a cruise control which the driver activates and it controls the car’s speed to keep it just above the minimum time requirements for each sector of the race track and leaves the driver to just turn the steering wheel.

  7. Another sad example of how F1 has gone from drivers racing each other to an off track computer generated game. Looking on the bright side we can solve a lot of the cost problems and get rid of the “Halo” simply by allowing the drivers to drive the cars remotely from the simulator at team headquarters.

    1. Agreed. Not sure how you avoid computation when conducting team strategy though, it seems inevitable in this day and age.
      As mentioned by others, you just need a slower pit lane speed limit during a VSC period, then no algorithm for this is required.

      1. Ban all communication from the team to the driver. No radio, just a generic gap pit board operated by the FIA.

      2. @john-h, Avoiding computers for race strategy is easy, just get rid of pit stops and leave the driver work out how to get ahead, which is how F1 used to be before Bernie.

    2. Would having Horners idea of one engine mode not maybe help? The way the only way to save the engine would be upto the driver to short sift, lift and cost etc. Atleast its back to the driver again….

  8. Seems like such an easy calculation that even a high-schooler could do during an exam. How can 4-times world champions Mercedes get this wrong?

    After Monaco 2015 you’d think they would have these SC/pitstop calculations triple-checked. Apparently not.

    1. The calculation is not simple at all. I’ve no idea how it’s done and after reading quite a few articles about it I still haven’t seen anybody who knows. You’d expect an F1 team to do it properly, however complex it might be, sure, but then again it’s just one of one million things they need to do right to have a good race, and they did right the other 999.999.

      1. Perhaps the easiest way is just to take a lap in practice and do a simulation?

      2. @alonshow, how is it complex? All they are trying to calculate is how many seconds a car loses during a VSC pitstop compared to the car behind. They don’t need, nor would they want, millisecond precision so a back-of-the-envelope calculation would suffice:

        t = pitd*maxpitv + minpitt – trackd*maxvscv + margin

        where

        pitd = pitlane length (distance)
        maxpitv = max pitlane speed
        minpitt = min pitstop time
        trackd = track length (distance) from pitlane entry to exit
        maxvscv = max speed during VSC
        margin = security margin

        Sure you can include accelerations and decelerations in and out of the pitlane and pitstop, lower speeds going to the last one or two corner if applicable, but again, not only it wouldn’t be that much more complex but this level of precision is not desired since you’ll want a security margin in case of any eventualities anyway.

        1. @paulk: That’s wrong in so many ways. For starters, there is no max speed during a VSC. The list goes on, but I’ll spare you the details. That equation is not even wrong, for in order to be wrong it would have to be possible for it to yield a result, and it isn’t. So it’s beyond wrong, it’s just meaningless.

          About the idea of doing a simulation in practice, it doesn’t sound like a very good idea. You’d have to do that simulation for every single race, which would waste lots of precious and extremely limited practice time and resources in getting information that a piece of software can get for you anyway (assuming it doesn’t have bugs, of course).

  9. Interesting answers from the engineer anyway.

  10. F1 became a matter of saving so much of the car… engine life is precious now with the stupid new 3 engine rule, plus the fuel and whatever. It’s frustrating. In any other era, Hamilton wouldn’t have lost the race no matter what niggling software bug.

  11. Yay Mercedes! glad you have pinpointed the reason I’ve watched F1 all these years: probabilistic equations.

  12. Wow, last time I checked, RoboRace was still testing their cars, now they’re competing in F1!

  13. Winning is winning for ferrari. Lose is lose for mercedes, u lose already so dont give excuse anymore. U cant get back the Australian trophy. Just go ahead n stay ahead.

  14. Although I like Niki Lauda, I do not like his philosophy: to win every race at the slowest possible speed. In this case it backfired. Good, now I propose a ban on radio and telemetry, let the driver do his job, in stead of listening to an algorithm witch produces energy saving delta times.
    I want drama, speed and battles, if that means blown up tires and engines, nerve wracking empty tanks etc.: yes please. Everything is better than a Mercedes winning every race in energy saving mode.

    1. * witch = which. Even though the algorithm is a… witch.
      Damn autocorrect. Same feeling Lewis must have had ;-)

  15. I am a bit confused.
    The subroutine that would calculate such a time delta, involving initial gap, locations and velocities of two cars at t0, versus the required gap under a VSC is so simple and basic, and could be programmed, tested and debugged in an afternoon.
    Compare this with the CFD-mesh subroutines used to solve the navier stokes equations for tubulent airflow over the front wings; or the subroutines used to control the MGU-H, MGU-K, hydraulic brakes, batteries and ICE. The levels of complexity and difficulty are orders of magnitude greater.
    How can you get the latter right and the former wrong?
    This literally makes no sense to someone that has done a lot of scientific computer programming.

    1. Potentially because the more complex equations get triple checked whereas the assumption is the simple one has been done correctly?

  16. Lewis didn’t even want to pit. Why can’t his brain be the computer that counts? Ban radio comms now.

    1. fyi, Hamilton was the one who masterminded his pit stop in the 2015 Monaco GP as well …
      Can we stop blowing this out of proportion? Hamilton and Mercedes lost a race, sucks for them. Is there anything fundamentally wrong with the VSC? Maybe, but if so, that has been the case ever since it was introduced, and it never really bothered anyone.
      So, I’m really sick of those cries to change everything just to accomodate that one driver who lost that one race. There’s exactly one necessary consequence of that race: Mercedes need to get their act together. End of story.

      1. Hamilton was the one who masterminded his pit stop in the 2015 Monaco GP as well

        Ah Monaco 2015 where Merc gave Hamilton incorrect infomation and then incorrectly calculated whether he could pit safely or not. Merc threw away a perfectly good 1 2 there.

  17. Fuel saving. Tyre saving. Engine saving. I hate these 3 things more than the halo, pay drivers or engine modes.

  18. See it this way, a modern day F1 driver can mention in their post racing career resume about the dynamic management skills they have learnt while on the job of driving at 300 kmph.

  19. Formula 1 is getting more and more boring, hearing Hamilton on radio asking why Vettel had came out in front, because you were not racing just saving fuel and engine. Gone are the days of Senna changing tyres on last lap and trashing the car to get in front….

  20. Warren Stilwell
    29th March 2018, 14:52

    Big picture is the race wasn’t really played out between two battling cars, it was decided by a leapfrog pit stop part way through the race. That’s a problem.

  21. Just curious, I’m a fairly casual fan but I thought giving the drivers instructions over radio was banned a while back. Was that rolled back? It’s just giving the drivers such minute instruction and telling them to confirm to very specific timing seems like it’s full on race management done for the driver.

    Would be nice to just have a formula where the drivers go as fast as they can to win and manage their own car and tired and have to optimise it themselves with no outside interference. Am I mad or naive?

    1. It was un-banned in the second half of the 2016 season after the general consensus arose that the ban on ‘excessive’ radio communication turned out to be a lot more farcical than the problem it tried to solve.

      Am I mad or naive?

      Naive, I’m afraid. Pandora’s box has been opened, there is no turning back.

  22. To me it sounds like they are admitting it was not a software problem as much as they didn’t plot in a good enough VSC in lap.

  23. I’d rather the teams use their pit-speed limiter (or some other fixed speed setting) during the VSC period — do away with this Delta nonsense. The FIA WEC go around at a constant rate, why does F1 require a different system?

  24. Should the FIA consider implementing a track specific stationary period during pitstops under VSC to eliminate the advantage of pitting under VSC?

Comments are closed.