Games Event 4 Sprint Couplet: Were some lanes faster than others? /w Stats

[bengalinath19]

@theparadox116 It does sure seem unlikely that those 3 lanes just happened to get those results of all 4 heat runners doing better. ESP when you view the names in them.

I think Castro being infallible after the games is a mental armor he has to wear to maintain his status as programmer but I do hope in his mind somewhere he thinks - ok this is a little weird and maybe next time/year we do X instead because of the possibility of equipment/lane bias.

 

[puttincomputers]

@bengalinath19 Doubtful, sled events have had this same issue in the past, and Castro still programs them. Same with Sprint events. If they wanted to test pure speed they could just do it at a regular track like it's been done for years at track meets. That will prevent turf issues. Or just Sprint without direction changes? But he still likes the zig zags on turf.

Castro seems to make adjustments based only on what he thinks needs to be adjusted, and he clearly likes these types of events so doubtful he'll adjust anything.

 

[bengalinath19]

@puttincomputers Idc about zig zag or running on turf. But equipment on a surface that shows a predisposition to be better than others (post event analytics) indicates not a fair test. Fair amongst the competitors.

No one is claiming the sprints zig zag was unfair because the placement of the barriers were not equal distances apart in each lane. That’s the equivalent comparison here. Or that some lanes could wear cleats and others have to run barefoot. Something to that extent.

 

[eternityintheirhearts]

@theparadox116 But your analysis includes the muscle ups.

That totally invalidates all data because it includes another movement. Plus everyone would be taxed from muscle ups differently and the first sprint differently. It makes comparison impossible.

You could do the analysis if you recorded how long it took everyone to do just the first sprint but even then it doesn't account for the athletes having good or bad technique.

The event being biased to some lanes is much less likely than it being uniform. You could assume the turf is of uniform length. It was all exposed to the same amount of sunlight. I would think the sleds are all brand new so are very likely to have all been manufactured the same and still in the same condition. Even if the turf was uneven, it would be very noticeable to have more than one or two second effect. Which it wasn't. Where is there room for mechanical variance?

 

[jayhiinkz]

@eternityintheirhearts I agree there are confounding factors, but I'm not sure it makes the analysis worthless. The muscle-ups are a separate variable, but I think any non-random effect (i.e. any bias introduced) would tend to make it less likely for what we see to have happened by chance.

[Specifically, over the 2 heats for each gender you explicitly have 1 athlete with points below the median and 1 above. Which you'd expect to bias towards that lane having 1 time below the median and 1 above.]

Watching, it certainly looked like some people were having a much harder time than others - at this speed/weight, it seemed like there was a definite 'sweet spot' where the sled moved easily, and some athletes couldn't find it. Whether that's variations in the surface, variations in technique, or just plain luck I don't know.

Personally, I find all the sprint events somewhat unsatisfying, because it feels like one bit of bad luck, or one small mistake, often makes a bigger difference than the actual difference in speed/skill/strength. (Or if you want to say "it's not luck", then often you're testing something insanely specific, like ability to judge how hard you can brake on turf without sliding, or judge where the barbell will bounce so you can jump over it).

 

[jacquiluvsjesus]

@eternityintheirhearts Sleds are extremely sensitive to minute differences. For example: near some lanes (but not the center ones), there were massive yellow fans spraying water onto the field.

 

[theparadox116]

@eternityintheirhearts As someone posted above muscle ups add noise which would make a difference even harder to detect yet it still shows up. It's not 100% clear but it adds a bit to the argument that the athletes were right.

I thought about subtracting the muscle up time but it would severely limit the sample size since not everyone is on screen, and it's especially unlikely that all 4 athletes in each lane are on screen.

The variance comes from the fans that mist some part of the field. Some lanes could even be used more due to seeding. Some lanes could also be exposed to more/less shade.

 

[eternityintheirhearts]

@theparadox116 True, I forgot about the misting fans, that would likely influence things. Can you see what lanes they were in / blowing over?

Limiting the sample size is just unfortunate. But if the overall time involves something that isn't a sprint and everyone will perform it differently then it's significant.

 

[lisarose001]

@eternityintheirhearts

Plus everyone would be taxed from muscle ups differently and the first sprint differently.

nope.

 

[praynmama]

@lisarose001 Yep.

 

[eternityintheirhearts]

@lisarose001 Nope?

Of course everyone would be taxed differently. People who are very good at muscle ups will find 18 much less taxing than people who are less good. The second sprint is influenced by the individuals anaerobic capacity much more than the first.

 

[lisarose001]

@eternityintheirhearts that's the beauty of seeing Tia and Mat look like shit. They are both great at muscle ups and got monkeyhammered on the back sprint. That right there is enough to justify a hypothesis, and the analysis is enough to conclude that the lanes were unequal.

You are criticizing the conclusion because it does not specify the precise extent to which the lanes were unbalanced and the impact on the athletes. That's true, it does not quantify those things and no test can. But that doesn't even matter. An analysis need only show that the test was biased to invalidate the result.

The OP has accomplished that. Take this shit in stages:

1) was the test fucked? Yep.
2) how did the fucked-ness of the test affect the results? Cannot be determined

But the fact that we cannot answer (2) doesn't make (1) go away.

Thus, "nope".

 

[eternityintheirhearts]

@lisarose001 Actually my point was the muscle ups are too great an unknown to be included in the analysis.

However, it most certainly does make (1) go away. If you can't determine how much of an effect something had then you can't determine there was an effect at all.

Unless you know every sled was pushed in the same way, with the same amount of force applied, and the muscle ups were done at the same speed, and everyone fatigued at the same rate, then comparing total time is too full of variables. Only considering the first sprint would be better, but still pretty inaccurate because people like Mat drove their sled into the ground right away.

 

[lisarose001]

@eternityintheirhearts Now I see the issue: You don't understand the OP, and appear to have a poor grasp of statistics generally. Here's a hint:

If you can't determine how much of an effect something had then you can't determine there was an effect at all.

This is absolutely and totally wrong. You should apply for a job at HQ.

 

[eternityintheirhearts]

@lisarose001 I bed to differ.

How does this conversation play out in your mind.

You: The lanes are biased.

Me: How do you know?

You: ...

If you can't determine the effect what on earth are you going to say? They are because I think they are? You can just tell?

You yourself said the effect of the lanes " Cannot be determined ". So how are you going to claim something is true if you can't prove there was an effect. The statistics would prove it if they weren't compromised by being on the total time.

 

[odonnell619]

@eternityintheirhearts I'm sorry, this tells me you absolutely do not understand how statistical analysis works. You do not need to control for every variable, known or unknown, in order to test a hypothesis. The only things you need to control for are non-random effects. Random effects on the other hand will present as noise in the analysis and mute rather than amplify the findings. If we still find something despite NOT CONTROLLING for random variance factors then there is most definitely something happening.

The only non-random effects here are lane assignment and all of it's correlates (tested ranking to this point in the competition which may also correlate with ability level, games experience, etc). This effect is defacto cancelled out by the analysis itself because the hypothesis that specific lanes are faster and the effect does not go from middle lane --> outside in the same order as ranking, thus there is no need to statistically enter it as a control variable. This is further amplified given there are repeated measures of the same lanes despite the overall ranking in that lane improving from one heat to the next.

The things you are talking about (small differences in gymnastic ability, pushing power, sled experience, anaerobic proficiency) are more than likely randomly distributed across lanes. Even if they weren't it would be very hard to explain how they were distributed in a fashion that was different than overall ranking at the time of seeding given that they are all fitness variables that would contribute to.... ranking.

Short: I can't dream up a scenario in which gymnastic ability and/or sled pushing power/proficiency was non-randomly assigned to the specific lanes that outperformed the others while also being different than overall ranking in the competition up to this point. Can you? If you can't then the analysis holds with the given probability of error.

 

[eternityintheirhearts]

@odonnell619 Muscle ups are not sled pushes so can't be included in analysis about sled pushes! How is that so complicated!

I understand statistics well enough. You do have to control every variable. Or as many as possible. Otherwise what are you testing? It's the scientific method. There are too many random variables in this analysis to have meaningful conclusions drawn. Differences in sled technique and muscle up proficiency are not small differences so absolutely must be controlled or eliminated to test the claim the lanes are different. If you can't control them you can still draw conclusions but the uncertainty is so high that you can't accept the results as accurate.

 

[lisarose001]

@eternityintheirhearts Statistics don't prove anything. It's not how statistics works as a disciploine. We already know you don't understand statistics. Stop telling us.

The IG account @wodscience has more info; I don't wanna copy paste, but they identified specific lanes which produce results unlikely to be explained by chance.

 

[lonelyone0111]

@theparadox116 This is why xpo sleds should have been used instead.

https://www.roguefitness.com/xpo-trainer-2

 

[sonas]

@theparadox116 WOD Science did a similar analysis here:

View this content on Instagram

View this content on Instagram