Was Kilty penalized for superior reflexes?

"Long known for his exceptionally fast reaction time at the start of races (as automatically recorded by pressure pads embedded in starting blocks), Kilty routinely reacts to the starter's gun in less than 0.130 seconds, several hundredths of a second faster than his competitors. Under rules established by the IAAF to discourage sprinters from "anticipating" the gun, any racer who reacts faster than 0.10, is deemed to have "false started" and is disqualified. " -

https://en.wikipedia.org/wiki/Richard_Kilty#.28U.29_Lightning_Fast_Reactions.2C_but_Not_Blinding_Speed

Under rules established by the IAAF to discourage sprinters from "anticipating" the gun - can someone bring up the actual IAAF regulations here for a quote?

The way I see it, if it can be demonstrated that a sprinter did not anticipate the gun (that is, that it the reaction time is consistent with past performances by that athlete) then in the name of fairness the DQ cutoff of 0.1 should be adjusted correspondingly. It wouldn't be difficult to hook a database up to the timing systems and make this calculation in real time.

It's not clear. Probably the key is in the absence of flinches on the blocks.

Tim Montgomery had a 0.104 reaction in his 9.78 WR, He clearly jumped the gun but wasn't called. Because he didn't move.

In Daegu 2011, Bolt also had 0.104 reaction but he flinched, following the blinches of Yohan Blake (uncalled) and that was a FS.

stochastic_track wrote:

It was pretty obvious that Kilty didn't anticipate the start, he just has some of the quickest reflexes out any sprinter in the field,

Not obvious at all... Many think the theoretical limit is around 0.13 so even when you end up with a reaction time of 0.10x you probably got away with jumping the gun. What is Richard Kilty's usual reaction time?

HardLoper wrote:

stochastic_track wrote:

It was pretty obvious that Kilty didn't anticipate the start, he just has some of the quickest reflexes out any sprinter in the field,

Not obvious at all... Many think the theoretical limit is around 0.13 so even when you end up with a reaction time of 0.10x you probably got away with jumping the gun. What is Richard Kilty's usual reaction time?

Ok I concede that whether an athlete anticipates or not is very difficult to make clear, apart from visible flinching. That's why we rely on the "reaction barrier", which is set to 0.1. My question is, why should it be a static barrier? If someone indeed possesses superior reactions, then wouldn't it be better to establish a statistical system for determining whether a modification of a "soft" 0.1 barrier is needed?

Isn't this a bit like suggesting if an athlete sets a new PB by a margin greater than X, s/he is DQ'ed?

What's the outlier cutoff? You're still setting some arbitrary number as a cutoff.

I think drop it to 0.00, and be done with it. If the athlete anticipates the gun correctly, good luck to her/him.

stochastic_track wrote:

HardLoper wrote:

Not obvious at all... Many think the theoretical limit is around 0.13 so even when you end up with a reaction time of 0.10x you probably got away with jumping the gun. What is Richard Kilty's usual reaction time?

Ok I concede that whether an athlete anticipates or not is very difficult to make clear, apart from visible flinching. That's why we rely on the "reaction barrier", which is set to 0.1. My question is, why should it be a static barrier? If someone indeed possesses superior reactions, then wouldn't it be better to establish a statistical system for determining whether a modification of a "soft" 0.1 barrier is needed?

Kilty would no doubt be penalized in that case. He isn't consistently under 0.110, all his reaction times are 0.12x-0.13x... which may be the best in the world. There's only so fast a neural impulse can travel and in that way the 0.100 barrier is like the 4:1 T:E ratio, assumed to be well beyond impossible to achieve naturally.

Wtfunny wrote:

Isn't this a bit like suggesting if an athlete sets a new PB by a margin greater than X, s/he is DQ'ed?

What's the outlier cutoff? You're still setting some arbitrary number as a cutoff.

I think drop it to 0.00, and be done with it. If the athlete anticipates the gun correctly, good luck to her/him.

When was the .100 rule instituted? It was well before the current false start (one and done) rules which might be enough to discourage anticipating the gun

Wtfunny wrote:

Isn't this a bit like suggesting if an athlete sets a new PB by a margin greater than X, s/he is DQ'ed?

What's the outlier cutoff? You're still setting some arbitrary number as a cutoff.

I think drop it to 0.00, and be done with it. If the athlete anticipates the gun correctly, good luck to her/him.

No. No, they would only be DQ'd if they BOTH broke 0.1 for reaction time AND their reaction time was an outlier based on their historic performances (take 95% or 99% confidence).

Do you see the massive distinction?

Under my proposed system, nobody would be DQ'd that wouldn't be DQ'd under the current system. However, the "flexible cutoff" gives someone the opportunity to compete if they simply have faster reaction time, based on their past record.

I'm in the process of acquiring historical data so I can give some hard numbers.

For the moment being, let us assume that Athlete X has had 20 IAAF-recorded reaction times over the past three years: 0.112, 0.104, 0.119, 0.226, 0.103...etc. Now, let's say he reacts with a 0.099 one day (like Kilty today). If we just plug that time into a suitable formula (https://en.wikipedia.org/wiki/Dixon%27s_Q_test for example) we can register 99% confidence of whether that 0.099 was an outlier or not. And we can DQ or keep him accordingly.

It looks like the IAAF posts reaction times for championship meets once a year, but not for Diamond League meets, at least not the 5-6 meets I checked.

Why don't they post them all? They even have the column in the table for it. At the least, it would make for an interesting analysis of reaction times and they record them anyway so why not make them publicly accessible?

This mark would actually probably pass that test. Dixon's Q test assumes a normal distribution, so a 0.099 would be fewer standard deviations from the mean thanks to his slow outliers. However reaction times are not normally distributed, they are concentrated towards the athlete's natural limit with a few bad starts on the higher end. At this point neuroscientists pretty much agree that 0.100 is safely below the human limit and nobody has empirically challenged that.

So you're suggesting different rules and standards for different athletes? That would be odd.

Flo'day boy ....

I don't know for sure, but wasn't it the early 90's sometime? I remember them discussing it when Christie was DQed in Atlanta 96, and, IIRC, saying he had something to do with the rule's development. He was an advocate for the .10 rule, and it cost him a race in that final.

I agree it seems to be a bit redundant having one and done AND the .10 rule. Personally, I don't like either of those rules.

HardLoper wrote:

This mark would actually probably pass that test. Dixon's Q test assumes a normal distribution, so a 0.099 would be fewer standard deviations from the mean thanks to his slow outliers. However reaction times are not normally distributed, they are concentrated towards the athlete's natural limit with a few bad starts on the higher end. At this point neuroscientists pretty much agree that 0.100 is safely below the human limit and nobody has empirically challenged that.

No need to fixate on Dixon's Q test, I absolutely understand that reaction times are not normally distributed. We can sample from a database of reaction times (once we obtain one!) to determine the distribution shape and develop a model for detecting outliers accordingly.

"At this point neuroscientists pretty much agree that 0.100 is safely below the human limit and nobody has empirically challenged that." The fact is that 0.100 would have had Kilty compete today, but his 0.099 cut him out, and I don't know why there's not more people asking how can we know whether that's fair.

Wtfunny wrote:

So you're suggesting different rules and standards for different athletes? That would be odd.

Seriously? NO. Why don't you reread what I've written and give it another thought.

What I'm suggesting is that under our current system, an athlete with phenomenal reaction ability (such as Kilty) may be unfairly excluded from competition by the arbitrary 0.1 second standard.

I am proposing one modification to the standard to be held to all athletes. The new modification is, that if an athlete dips under the arbitrary 0.1 (say 0.998) and they have a record which gives us a high statistical confidence that this was a natural reaction and not an outlier, indicating anticipation, then we can allow that athlete to go through. For example, if Insane Flash ran with reactions of 0.102, 0.105, 0.101 so far at Diamond League this summer, then 0.099 would seem reasonable, right? Why cut him out when he could run at 0.101 all summer long? On the other hand, our statistical model would demonstrate that Flash's 0.099 was an outlier, and thus likely to be illegal anticipation, if his previous record was 0.182, 0.199, 0.211 so far this summer.

Does this make more sense?

Right, I see what you're saying, it's like the longitudinal drug tests (like bio passport) but for reaction times. Maybe it would work if you managed to get enough sample reaction times for every runner. But if Kilty ran 0.100 today, he would have just gotten a very rare lucky break... I don't see it as an injustice unless he were consistently reacting close to those times.

PS I can see all reaction times for this year's Diamond League results but not the archived ones.

HardLoper wrote:

PS I can see all reaction times for this year's Diamond League results but not the archived ones.

Nice, where are you seeing those times - on the IAAF site? Link please?

Also:

stochastic_track wrote:

For example, if Insane Flash ran with reactions of 0.102, 0.105, 0.101 so far at Diamond League this summer, then 0.099 would seem reasonable, right?

Definitely, though no one is close to doing that, so it's still a solution to a problem that likely doesn't exist.

stochastic_track wrote:

HardLoper wrote:

PS I can see all reaction times for this year's Diamond League results but not the archived ones.

Nice, where are you seeing those times - on the IAAF site? Link please?

The 100m dash has the reaction times

Thanks for the link. I'll either assemble an extensive repository of reaction times myself or find one, and I'll post it when I do (though someone might be able to save me some time).

I acknowledge that this situation is rare, but the idea that Kilty may have not anticipated today is troubling. If we could at least get his stats for the past few races that would be great.

My made-up numbers above are exactly that - made up - but where do we draw the line? (We draw it at 0.1) If we apply analogous stats to those used in blood profiling, are there any current sprinters who would be affected? Is that the best we can do? If "scientifically rigorous" statistics were applied to develop a model as discussed, would that be better than the current hard 0.1 rule?

HardLoper wrote:

Right, I see what you're saying, it's like the longitudinal drug tests (like bio passport) but for reaction times. Maybe it would work if you managed to get enough sample reaction times for every runner. But if Kilty ran 0.100 today, he would have just gotten a very rare lucky break... I don't see it as an injustice unless he were consistently reacting close to those times.

PS I can see all reaction times for this year's Diamond League results but not the archived ones.

To be more clear, I pretty much agree with everything here. Thanks for the productive discussion.