|
I'm the kind of person who wants to make informed decisions based on data. Since getting into motorcycling I wanted to know how to be the most safe on a motorcycle. So I emailed the Federal Highway Administration to get the data of the Motorcycle Crash Causation Study, and they sent it to me, then I used my mathematics of computation degree to write some python to process the data and I'll post them here. If you have any questions, like "are crashes more common if you have X" then post them here and I'll crunch the data. A note on the statistics For any "COMPARISON" section, I did a chi squared test on the contingency table of the values presented here. "save" is these values plugged into Bayes' theorem with probability of crash at 100%, yielding how likely you are to get in a crash if you are wearing retroreflective materials. For the "nighttime" versions I removed any data point in high light conditions from the samples, leaving only mid light and darkness. Control samples were gathered at the same time of day as the crash sample, so if the crash sample was during high light conditions then the control was removed as well. It's important to note that the study's not randomized (and a purely randomized study wouldn't pass the ethics review board) so it's tough to establish a causation ie that a particular piece of gear reduces crashes, because it could be that the sort of person who buys that gear also tends to drive safer. But we can establish whether there's a correlation, and if there's no correlation there's definitely no causation. In my book if there's a correlation then there's likely to be a causation as well, worth enough to spend the money. Here's how to read this example: code:As a disclaimer, I do have a background in mathematics and I learned all this statistics stuff at uni, but it's not like I do this professionally so take everything with an I-don't-do-this-for-a-living sized portion of salt. CRASH CONFIGURATIONS This is a chart of the most common crash configurations and how lethal they were, ordered from most common to least common. MC=motorcycle, OV=other vehicle. Here are my interpretations for the configurations that aren't obvious.  "OV turning left in front of MC, MC proceeding in either direction perpendicular to OV path"  "MC and OV traveling in opposite directions, OV turns in front of MC, crossing MC path, MC impacting OV" code:* Vehicles taking a left turn in front of the motorcycle are the most common and therefore important to watch out for, but aren't necessarily the most dangerous setup. * The most lethal setup is actually the motorcycle running off the roadway, almost half!! of accidents causing fatalities. * A full 19% of crashes are motorcycles falling on the road with no other vehicle involved. Avoid falling on the road like a dingus and you avoid 1/5 of accidents, including the most lethal type. CRASH CONFIGURATIONS (ZERO BAC) I pretty absolutely never drink when I drive anything with an engine. So my first question is: How are the results affected if the rider has a blood alcohol level of 0? This chart is the same data, except every case where a rider had a positive blood alcohol level was removed first. code:* I expected fatalities from running off the roadway to be less common and a result of impairment by alcohol, but I was wrong. Sober motorists run off the roadway all the time. Very important to slow down before curves. * Actually, none of the numbers have changed drastically. Turns out, drunk riders make roughly the same mistakes as sober, just apparently more often. (Well I don't know that for sure but I'm assuming drink riders have accidents more often.) RETROREFLECTIVE MATERIALS COMPARISON I want to start asking questions like "Will doing X make me safer?" and I figure answering the question "Will putting retroreflective materials on the motorcycle and myself make me less likely to be in a crash?" is a pretty simple place to start and a cheap way to become more safe. In this test, the motorcycle was considered to have retroreflective parts if it had any retroreflective surface, material or device anywhere on it. The rider was considered to be wearing retroreflective clothing if their upper body, lower body, or gloves had any retroreflective material or if they had an arm band or similar item. Material on the helmet or boots wasn't considered here. code:* Putting retroreflective materials on you and your bike definitely makes you significantly less likely to get into a crash, enough that probably not all of the effect is because of a safety bias frm the rider. Retroreflective material is cheap. Definitely buy and place retroreflective materials on yourself and your bike. * The reduction in nighttime crashes when wearing retroreflective materials on the rider at night is kinda high and the p value is still very strong. I'll definitely be buying a retroreflective riding jacket. AFTERMARKET FRONT TURN SIGNALS COMPARISON Next question I want to ask is "will buying the Admore light bar (review) and turn signals make me less likely to get into a crash?" The problem is that the study didn't track aftermarket modifications to rear lighting, only front turn signals. I don't actually expect front turn signals to reduce crashes much, especially because most replacement front turn signals aren't that much brighter than the standard ones. code:* p is too high for me to be confident in the analysis. I'll call this inconclusive. * I'll probably still buy that light bar and turn signals, but only if I'm planning on doing a lot of riding at night at some point. At the moment I'll only be returning from the occasional outing with friends. SADDLE BAGS COMPARISON I intend to get saddle bags so that I can go to the grocery store or whatever and not have to carry a backpack. Will buying saddle bags make me less likely to get into a crash? I actually don't think so, because why would it? If my code is correct I expect this to turn out to have negligible effect. code:* I'm somewhat surprised. There are only two causes I can think of: 1. saddlebags make the motorcycle appear bigger and cars see it more often 2. saddlebags are typically installed by people who aren't interested in reckless behavior. I'm getting saddlebags either way I suppose, I'm just skeptical that #1 is a significant part of the effect. Vino fucked around with this message at 22:28 on Jun 3, 2023 |
#
?
Jun 1, 2023 02:45
|
|
|
|
| # ? May 3, 2024 07:58 |
|
|
More questions I want answers to: * Does ABS reduce crashes? * Do engine guards reduce leg injuries? * What mechanical failures most likely lead to the crashes in this study? Vino fucked around with this message at 18:35 on Jun 6, 2023 |
|
#
?
Jun 1, 2023 02:45
|
|
|
Saving a post for myself to add results and edit later.
|
|
#
?
Jun 1, 2023 02:45
|
|
|
OK I have added the below to the OP. TLDR is it looks like retroreflective materials are very useful in reducing crashes, especially at night. RETROREFLECTIVE MATERIALS COMPARISON I want to start asking questions like "Will doing X make me safer?" and I figure answering the question "Will putting retroreflective materials on the motorcycle and myself make me less likely to be in a crash?" is a pretty simple place to start and a cheap way to become more safe. In this test, the motorcycle was considered to have retroreflective parts if it had any retroreflective surface, material or device anywhere on it. The rider was considered to be wearing retroreflective clothing if their upper body, lower body, or gloves had any retroreflective material or if they had an arm band or similar item. Material on the helmet or boots wasn't considered here. code:* Putting retroreflective materials on you and your bike definitely makes you significantly less likely to get into a crash. Retroreflective material is cheap. Definitely buy and place retroreflective materials on yourself and your bike. * The reduction in nighttime crashes when wearing retroreflective materials on the rider at night is kinda high and the p value is still very strong. I'll definitely be buying a retroreflective riding jacket. Note: See the OP for for more details on the math Vino fucked around with this message at 08:58 on Jun 2, 2023 |
|
#
?
Jun 2, 2023 08:52
|
|
|
Vino posted:then I used my mathematics of computation degree dang what kind of batteries does it take
|
|
#
?
Jun 2, 2023 13:33
|
|
|
I mean this complementary: your posts make so much more sense now.
|
|
#
?
Jun 3, 2023 05:08
|
|
|
lmao i suppose it's a complimentshacked up with Brenda posted:dang what kind of batteries does it take These batteries are old and lose charge real quick
|
|
#
?
Jun 3, 2023 06:33
|
|
|
This is cool! Glad you put this together. I don’t have any good ideas for pulling interesting insights out of this but makes me want to get that highlighter yellow turtle vest.
|
|
#
?
Jun 3, 2023 06:42
|
|
|
I wear highlighter yellow jackets with retroreflective materials when I ride. Not sure what makes riders like me statistically safer though - is it the higher visibility of the gear itself, or simply a selection bias from the willingness of risk averse riders to wear ugly gear because we think it makes us safer? I'm guessing it's a little of both, and that it is sadly impossible to find out what is what by looking at statistics. FWIW, I've learned that the archetypal lethal moto accident in Sweden is caused by a something categorised as either a "sports bike" or "cruiser" running off the road in a rural area in a turn where the speed limit is 70 km/h, no other vehicles involved, in daylight and good weather, little traffic, usually early morning or late evening on a weekend. My takeaway is "don't get too frisky in good conditions".
|
|
#
?
Jun 3, 2023 14:19
|
|
|
Yea I can't imagine that the entire effect is caused by one of those factors alone. The study's not randomized (and a purely randomized study wouldn't pass the ethics review board) so it's tough to know causation like that. But we can establish whether there's a correlation, and if there's no correlation there's definitely no causation. In my book if there's a correlation then there's likely to be a causation as well, worth enough to spend the money. I should add that note to the OP. However when we start looking at "will X equipment reduce injuries" then I think we can start assuming that most of the causation is from the equipment and not from skill of the driver. AFTERMARKET FRONT TURN SIGNALS COMPARISON Next question I want to ask is "will buying the Admore light bar (review) and turn signals make me less likely to get into a crash?" The problem is that the study didn't track aftermarket modifications to rear lighting, only front turn signals. I don't actually expect front turn signals to reduce crashes much, especially because most replacement front turn signals aren't that much brighter than the standard ones. code:* p is too high for me to be confident in the analysis. I'll call this inconclusive. * I'll probably still buy that light bar and turn signals, but only if I'm planning on doing a lot of riding at night at some point. At the moment I'll only be returning from the occasional outing with friends. SADDLE BAGS COMPARISON I intend to get saddle bags so that I can go to the grocery store or whatever and not have to carry a backpack. Will buying saddle bags make me less likely to get into a crash? I actually don't think so, because why would it? If my code is correct I expect this to turn out to have negligible effect. code:* I'm somewhat surprised. There are only two causes I can think of: 1. saddlebags make the motorcycle appear bigger and cars see it more often 2. saddlebags are typically installed by people who aren't interested in reckless behavior. I'm getting saddlebags either way I suppose, I'm just skeptical that #1 is a significant part of the effect. edit: I realized that for this one it doesn't matter whether the saddle bags are aftermarket or factory, and the data was missing a lot of values in the aftermarket column because the data collector can't always tell whether the saddle bags are aftermarket. Anyway I used a different column but the results are basically the same. Vino fucked around with this message at 22:28 on Jun 3, 2023 |
|
#
?
Jun 3, 2023 19:43
|
|
|
Actually now that I think about it, if you look at the crash configurations chart, less than 10% of crashes are in a configuration that the other vehicle has any opportunity to see the rear of the motorcycle. The light bar may reduce the instance of being rear ended at stop lights, but that's relatively few of the crashes. The hand guard lights would help in a lot more cases.
|
|
#
?
Jun 3, 2023 20:57
|
|
|
Percentage of motorcycle crashes by brand, and bike type(cruiser, standard, sports, adv, dual sport enduro) Do they break down by specific model? Ooh, what about mechanical failure leading to accident? SSH IT ZOMBIE fucked around with this message at 05:18 on Jun 6, 2023 |
|
#
?
Jun 6, 2023 04:38
|
|
|
The problem with crashes by brand, bike type, or model is that it's not randomized, meaning there's a lot of room for rider bias. For example if you discover that harleys get into crashes more often then you still don't know whether that's because harleys are safer or whether safer drivers tend to buy harleys. But here's the data, I only did it for brands that had at least five crashes:code:With bike type you would have the same bias problem but I would have the problem of that the data doesn't classify bike type so I would have to do all that myself. Not sure it's worth it given the bias problem still prevents you from getting much useful data out of it unless the effect is huge. With specific model same bias problem and also your p scores will be even higher since a specific model is even more rare than a brand. Doesn't look worth it for me to do that work either. Mechanical failure I believe is tracked in the data I have and sounds very useful to know!! But it's not something I have code ready for right now, so I'll put it on the list.
|
|
#
?
Jun 6, 2023 18:30
|
|
|
I think (not sure?) that that data is basically just stating that there are a lot more Harleys, Hondas, Kawis, Suzukis and Yamahas on the road than Ducatis and Triumphs? Or is that wrecks per-capita, so to speak?
|
|
#
?
Jun 6, 2023 21:28
|
|
|
No. For each crash, the survey required that two control data points be taken at the same place and time of day. So they went to the place a crash took place at the same time of day and put up a sign that said "are you on a motorcycle? stop and take a survey for some quick cash" and they took two control points for each crash. My analysis does a statistical comparison of the percentage of control data as a group to the crash data. That doesn't give you the number of harleys, hondas, etc on the road in general, it only gives you the number of each brand on the road at the times of day and locations where crashes are likely to occur (it's prob pretty close though.) So as a group, the control data was 1.99% Ducatis and the crash data was 1.42% Ducatis. There aren't a lot of Ducatis in general so it doesn't tell you much, (as indicated by the high p score,) but there are slightly fewer Ducatis in crashes than in the control.
|
|
#
?
Jun 6, 2023 22:06
|
|
|
You can weigh the accident rates against the percentage of market share. https://www.statista.com/statistics/252210/market-share-of-major-motorcycle-manufacturers-in-the-us/ Interestingly, Honda and Harley have a large US Market share and are represented similarly in crash rates. Yamaha looks like an anomaly whereas Kawasaki has similar market share to Kawasaki. That actually is sort of interesting. Would be cool to drill into models. Ie, are the crashes happening on bikes like the MT07 where there's fewer safety features like a slipper clutch, or on R1, R6s. Very interested about percentage by mechanical failure, brand. Edit: Ah, you said there is a percentage tracked of overall ownership in the survey as well SSH IT ZOMBIE fucked around with this message at 01:04 on Jun 7, 2023 |
|
#
?
Jun 6, 2023 22:49
|
|
|
It says I need to register so I did but it still doesn't show the data to me. Rough. But yea I have mechanical failure on my list of things to dig into. I put the retro-reflective strips on the bike and the helmet and hauled it out at night to see how much it affects visibility. Here it is with my helmet on the back, without the camera flash.   Yea I can see how that works. Def gonna put on the four additional strips I have.
|
|
#
?
Jun 7, 2023 07:58
|
|
|
Vino posted:It says I need to register so I did but it still doesn't show the data to me. Rough. But yea I have mechanical failure on my list of things to dig into. It's more noticeable...I don't know where you live but riding at night introduces a lot of variables. Wildlife and you don't see obstacles in the road as easy. MSF I know discourages night riding. I will ride well lit highways at night that I am familiar with, or a well lit city. I generally avoid residential areas and non-highway main roads. I don't know if road classification or time of day is in your data. Dark country roads feel outright suicidal. Also is that a Versys X 300 I spy? We have the same bike 😂 SSH IT ZOMBIE fucked around with this message at 04:13 on Jun 8, 2023 |
|
#
?
Jun 8, 2023 04:06
|
|
|
Yes it is!! Yes the only rising at night I plan to do is the on the way back from the hanging with friends sort of thing. Also I realized those strips are exactly where my legs go so I put two more in each side.
|
|
#
?
Jun 8, 2023 05:14
|
|
|
On the e-bikes I use for year round commuting (and it's pretty much always dark when I commute in mid winter) I put a bunch of retroreflector tape strips in the front wheel. Solid plastic ones reflect better than the tape but these always en up breaking, the tape lasts longer and never rattles when loose. Anyways it's the part of the bike that pokes out in intersections first (esp. on the cargobike where the wheel is way out in front) and as a driver I find that the motion of reflectors in a turning wheel is really attention grabbing. Also when placed on or near the rim there's always reflectors low to the ground where they're the most likely to be hit by photons from a car low beam.
|
|
#
?
Jun 8, 2023 05:39
|
|
|
I thought this article that came out the other day was really good and recommend reading all of it. While it isn't specific to motorcycles, it does a good job of chronicling all the ways drivers have turned to poo poo in the US. It also validates our hunch that drivers actually have gotten worse during Covid and more emotionally unstable. It's pretty dangerous out there, friends. https://www.nytimes.com/2024/01/10/magazine/dangerous-driving.html quote:The bigger the vehicle, the less visibility it affords, and the more destruction it can wreak. In a report published in November, the Insurance Institute for Highway Safety, a nonprofit, concluded that S.U.V.s or vans with a hood height greater than 40 inches — standard-issue specs for an American truck in 2023 — are 45 percent more likely to kill pedestrians than smaller cars. quote:The problem today, in the United States, may be that we’re all baseline angry and anxious — and we’re all in a car, all the time. Or what feels like it, anyway. Outside the few fully walkable cities or the ones with reliable public transportation, most communities require a car to traverse. And traverse slowly: Traffic is up across the country, as is the duration of our average one-way commutes, which recently topped 27 minutes, the longest in our history. quote:More than 32,000 Americans were killed in wrecks caused by distracted driving between 2012 and 2021, with a roughly 11 percent increase in the number of deaths from 2020 to 2021.
|
|
#
?
Jan 12, 2024 04:50
|
|
|
Do these numbers reflect the recent influx of scooters and small electric vehicles? Big cities became completely overrun with delivery bikes, with riders who are distracted to a ridiculous degree. Delivery dudes aside, riding a motorcycle (sanely) still feels safer than driving a car, just by virtue of the fact that it requires more concentration. Also I wired up lots of running LEDs on my bike instead of retroreflectors. Hopefully they're just as effective.
|
|
#
?
Jan 12, 2024 05:10
|
|
|
Ride dirtbikes instead
|
|
#
?
Jan 12, 2024 14:27
|
|
|
I didn’t dig into your links enough to see the date ranges on your data, but I sometimes wonder what happens when you eliminate accidents before, say, 1990 or 2000 as ABS and rider aids started becoming more standard and more advanced. Not to mention safety gear, basic bike geometry (no more Dyna death wobbles, etc.), and other factors that have made things safer over the years. Or do those things lead to an overconfidence that balances it all out? I’m not sure what the crash that killed Lawrence of Arabia can teach me about touring on my Bonneville, y’know?
|
|
#
?
Jan 15, 2024 00:49
|
|
|
Geekboy posted:I’m not sure what the crash that killed Lawrence of Arabia can teach me about touring on my Bonneville, y’know? Wear a helmet and don't outrun your sight lines, for starters.
|
|
#
?
Jan 15, 2024 02:53
|
|
|
|
| # ? May 3, 2024 07:58 |
|
|
Don't do 30 over the speed limit through busy intersections. That's a top tip for free.
|
|
#
?
Jan 15, 2024 03:40
|
|