I'm not American and I don't really know much about the DQed athletes but I do care about fairness in the sport. The RTs are clearly lower at this event than other events and I don't believe the athletes have suddenly found a way of starting better. This seems to be across all events (with a large sample size). Even if you have no love for the athletes involved, surely you can see that this isn't right.
I wonder if the blocks construction and track construction come into play at all? The blocks require 25kg of force to be applied before they trigger movement (I think). So if the blocks were more rigidly in place (longer spikes, better grip to the surface) then any flex in the blocks might be reduced and the whole of the 25kg force applied to their surface might be recorded slightly earlier than if the blocks had some flex in them? Maybe?
That is also an interesting point. Track surface technology is almost always hyped to have improved since the last championship, being all new and whatnot. If (more) true this time there might be an issue with starter block compatibility change (or what to call it).
What if they've improved the technology, and they're finally getting accurate reaction times? The Oregon data is different...not necessary wrong.
Then they need to change the rule, 4 athletes got DQ'd in one night. I also find it unlikely that 4 athletes were able to anticipate the gun within .01 before it was fired. Usually when there are false starts, there is much more dispersion in times, including negative reaction times meaning they jumped fully before the gun.
I think he's probably right because I counted 40 reaction times under 0.120, which is a pretty hard # to beat usually, and quite a few of those were under 0.110. None of them were absolutely ludicrous and obvious false starts such as 0.05 or under. I believe that 0.085 was the lowest, with three or four under .100, and numerous close to it including the .099 and .102/3 for Allen prior to the final and a couple .104, including one by Coleman. Kerley had a .107 in the first round. Something's not normal in the stats and their own study from around 2010 showed that reactions to sound begin around 0.08, so between the one and the other, Allen and the other dq's were jobbed.
Out of 279 starts of 100M/110M race (including the Heptathlon), I counted only 13 under .110 (including the 3 DQs) or about 4.6% of starts. I counted 37 between .119 and .110 or around 13% of starts.
Allen (.099) and the other DQs (.096, .095) seem to be significant outliers. DQs were 1% of starts - 3 out of 279 starts under .100. Is that an unexpected amount in a track meet of this size? Doesn't seem like it.
I didn't see any .085 - the lowest I saw was .095 (DQ). There were not "numerous" times close to .100 - there were 3 DQs (under .100), two .101s, three .104s and five between .105 and .109. So very few starts under .110 (13 total), and 3 were DQed. Anything under .110 seems pretty rare and under .100 seems incredibly rare.
There was .099 and .101 or .102 for Allen, as well as .104 for Coleman and another guy, and .107 for Kerley. Those aren't outliers, particularly not among the top runners. Remember that most of the starters in these races are not among world elite but from small undeveloped countries with slow sprinters who usually have poor reaction times. When they are dq'd, it is usually blatant--before the gun or around half the low limit on a legal start. Here there were far more sub .120s than in any other World or Olympic meet in at least the last ten years. That's not because they got trained up on electrical stimulator gizmos to improve their reaction time. The system is failing. Remember last year when everyone was getting false starts registering before the gun. That's probably the same over-sensitive system involved, just tuned a little better to avoid all the call backs.
I think the sensors in the blocks are just "better" now and all of the studies in the past that showed 0.100 seconds as the fastest a human can really react to hearing a gun was based on slower block sensors.
If you have to have a limit, it should be adjusted down to 0.080 seconds.
And 0.08 happens to be the legal limit for blood alcohol content, so the number somehow makes sense and can be remembered.
This speaks volumes. I know we are in a day and age where facts are conveniently ignored, however the fact that there have been 25 reaction times measured at under .115 so far and a total of 5 in the previous 5 championships screams there is a massive problem that should be dealt with immediately, not later.
If it is possible for reaction times to be as fast as 0.08 then why have previous championships produced so few times faster that 0.15. shouldn't most of these sprinters be pretty darn close to reacting as fast as humanly possible?
It seems like the timing has been flawed for years and they have finally corrected it to be accurate.
If it is possible to react as fast as .08 then they should lower the false start threshold to .08. However, if it is not possible to react faster than 0.1 then keep it the way it is, there is nothing wrong with the system.
Thank you, Monkeys Skyping! That's a great compilation. Does it include RTs for all the false starts? (I guess I could browse it in more depth to see...)
I am a statistician with 30+ years experience and 100+ research publications in science journals. What I've seen others post already (e.g., user: Timer....) is fairly conclusive. The difference is so large, unusual, and systematic that a 'fancy' [comprehensive] analysis is not really even necessary, IMHO. But I may do one just for fun.
Thank you, Monkeys Skyping! That's a great compilation. Does it include RTs for all the false starts? (I guess I could browse it in more depth to see...)
The data does have reaction times for false starts, at least for some cases. There is a gap from Devon Allen's 0.099 to the next highest recorded false start time of 0.053, which, I think suggests that the threshold could be lowered to, say, 0.08 without much risk of "false negatives". Reaction time for the 7th to 12th WC meet is recorded as "0" for DQs.
Most of the recorded false starts are blatant, that is, negative numbers indicating they moved before the gun.
Do you have the data and methodology for this graph? Because you could very easily just put that graph together. Not saying that Eugene isn't an outlier, but you need to provide the evidence.
I believe it. Since the false starts that didn't actually look like false starts I've been looking at the reaction times. The slowest reaction times are like 0.18 which is still quite fast, there are a ton of 0.12's being recorded which should be insanely rare but they are happening in every race by more than one person.
My bet on this issue is the prevalence of the new super spikes that Nike, adidas now have most athletes wearing that is causing this reaction time. The materials in the spikes are intended to provide a much more immediate push upon ground contact and that is the issue in play here that no one has mentioned. Every athlete here is trained and conditioned to have an excellent start like all other Worlds and Olympics but the new issue is the spikes.
Thank you, Monkeys Skyping! That's a great compilation. Does it include RTs for all the false starts? (I guess I could browse it in more depth to see...)
The data does have reaction times for false starts, at least for some cases. There is a gap from Devon Allen's 0.099 to the next highest recorded false start time of 0.053, which, I think suggests that the threshold could be lowered to, say, 0.08 without much risk of "false negatives". Reaction time for the 7th to 12th WC meet is recorded as "0" for DQs.
Most of the recorded false starts are blatant, that is, negative numbers indicating they moved before the gun.
I agree on the need to move the threshold just to be safe (and I was going to comment on FNs vs FPs, etc, but edited it out). Whether it is due to equipment calibration, or to starter command timing being ideal (as SprintGeezer and others alluded to in another thread), the obvious FS threshold needs to have a meaningful gap from the reasonable end of the distribution's tail. There were way too many RTs within +/- 10ms of the threshold to be confident it is a fair threshold. 90ms would be ok, 85 would be better. I'm afraid at 80ms you might provide incentive for more anticipation guessing.
I will go back and see if other archived sites (or waybacking) can fill in the gaps on those RTs for DQs that are recorded as zeros. The distribution of verifiably "real" False Starts' RTs is itself pretty demonstrative, it would be possible to empirically determine a threshold (e.g., using SDT framework) that would optimize the FP vs FN tradeoff once you postulate the relative costs of each mistake.
The starter being consistent is possibly an issue. Difficult to prove though as timing the precise gap between their words and the gun isn't recorded anywhere (that I know of).
Every one of them is recorded on video! Probably at framerates of 60 or greater, so precision of +/- .016s at worst.
Really the first question that should be answered is to look at the curves and find out, is this an overall-faster thing or a cluster of faster RTs dragging down the mean. If it's a cluster, then see if that correlates with specific equipment. If not, it's not likely the equipment.
It's a U.S. track meet and everyone, including athletes, are more sensitive (jumpy) to the sound of gunshots than they would be elsewhere in the world.
It's a U.S. track meet and everyone, including athletes, are more sensitive (jumpy) to the sound of gunshots than they would be elsewhere in the world.
If they grew up in Chicago or Baltimore, the shots would just lull them to sleep.