A Look at Aerobatics Accident Statistics

Note... this has been UPDATED with more data. See Post #5.
------------------------------------------------------------------------------

For those who don't know me, I do a lot of statistical analysis of homebuilt aircraft accidents. Wes Liu suggested that I take a look at aerobatics-related accidents, specifically those that occur in airplanes not approved for aerobatics, and to low-time pilots trying to teach themselves aerobatics.

Ran some analyses over the weekend, and post addresses the results. But I'm a Systems Engineer by trade, so I have to document the methodology, first.

I used my copy of the NTSB accident database, and the accidents I examined ran from 1982 through 2017. I ran a search for "Aerobat*" in both the Probable Cause and Narrative sections. In my experience examining homebuilt accidents, the investigators would use terms like "Loop" or "Roll", but would also use the word "Aerobatics" at some point in the description. I included planes with both Standard and Special airworthiness.

This produced 795 "hits." However, the use of the word "Aerobatic" in the description does not mean the plane had an accident while performing aerobatics... often, it was a mere description of the aircraft type ("...the aerobatic aircraft had been purchased two years earlier...."). These accidents were eliminated.

The next step was to eliminate cases where the fact that aerobatics were involved was not a major factor in the accident. This eliminated instances like landing accidents that occurred after practicing aerobatics, etc. I made a judgement call in cases where mechanical failure manifested itself during the performance of aerobatics. If the failure was powerplant related, I eliminated it, but kept cases of structural failure. This last was for obvious reasons, but in a few cases, the root cause of the structural failure was a construction issue...these were eliminated as well.

I also eliminated accidents that occurred at air shows.

With this winnowing-out completed, I had 486 accidents left. The next stage was to identify those which were approved for acrobatics, and those that were not. This was not all that simple. Technically, most homebuilts can legally perform aerobatics (there are a few, like most Fly Babies, that have aerobatics prohibited in the operating limitations). But there are some that really aren't intended for acro. So I flagged these as "non-acrobatic aircraft."

Similarly, many older production airplanes can legally perform aerobatics simply because the certification standard when they were introduced didn't actually ban acro. On these, too, I generally made a judgement call... considering a Waco biplane an aerobatic aircraft, while flagging the J-3 Cub otherwise (I know there are clipped-wing Cubs, etc.) I did have a couple of exceptions to this, considering the Luscombe and the Globe Swift as aerobatic airplanes.

Plenty of room for argument here, for sure, but that was my original cut. There were several types that I didn't recognize, these weren't included in the comparison.

Anyway, here are the basic results:

	Non-Acro	Acro
Total Accidents	114	357
Median Total Pilot Time	1117.5	1700
Median Time in Type	112	95
Maximum Total Time	36500	29500
Minimum Total Time	16	75
Maximum Time in Type	6727	4506
Minimum Time in Type	1	1
Fatalities	102	296
% Fatal	89%	83%

The first thing that might catch your eye is the comparison between the median total pilot time. The pilots involved in accidents with aerobatic aircraft had about 50% more flying time than those who crashed flying non-acro planes. Oddly enough, the time-in-type relationship is reversed, with the pilots of non-acro planes having a median 112 hours vs. 95. This may just be glitch in the data, or, potentially, it might indicate pilots getting bored with their non-acro airplanes and start playing around.

89% of all acrobatics-related accidents in non-acro airplanes are fatal, vs. 83% of the aerobatics-airplane crashes. A lot of these accidents occur at low level, precluding any emergency exit. But in the higher-altitude cases there were several instances of pilots using a parachute to abandon damaged aircraft. Someone performing aerobatics in a plane meant for it is probably more likely to be wearing a chute...and an aerobatic airplane usually gives the pilot better egress in an emergency.

The most interesting result came when I plotted the pilot total time as a histogram:
http://www.wanttaja.com/acro.jpg
This figure plots the percentage of the total accidents for each type vs. the pilot's total flight time. Note how relatively few accidents occur to lower-time pilots flying aerobatic airplanes. A lot of this is probably driven by the fact that few 80-hour private pilots buy Pitts Specials! But it does indicate that a lot of accidents in non-acrobatic airplanes involve relatively low-time pilots.

But of course, experience is no assurance that problems may not arise. Nor is experience, nor is even the size of the airplane involved....

"THE COCKPIT VOICE RECORDER RECORDED CONVERSATION OF BOTH PILOTS DISCUSSING AEROBATICS. THE PILOT IN THE LEFT SEAT STARTED TO DESCRIBE THE PERFORMANCE OF A PROHIBITED MANEUVER (APPARENTLY A BARREL ROLL). THE RECORDING ENDED BEFORE DESCRIPTION OF THE MANEUVER WAS COMPLETED. THE AIRPLANE IMPACTED IN A NOSE LEVEL ATTITUDE WITH THE LEFT WING SLIGHTLY DOWN. THERE WAS EVIDENCE THE MANEUVER WAS PERFORMED AT AN ALTITUDE INSUFFICIENT TO ASSURE RECOVERY OF THE AIRPLANE. "

This was in a Beech 99 airliner, basically a cross between a King Air and a Queen Air....

Ron Wanttaja

Nice work as always Ron, even if the data doesn't present any real firm conclusions.

Did you also search "acrobat" in the first pass? Could well be that the person writing a report used "acrobatic" instead of "aerobatic", since the former is, after all, the word used in the regulations.

My Hatz had operating limitations prohibiting aerobatics. Not sure why, I don't know what the policy was in 1987 when it was built, but current policy is for the op lims to say "only those aerobatic maneuvers performed and logged during phase I" may be performed during phase II. No matter that the previous owners had all been doing loops and rolls in it. I applied for and got new limitations for my plane, put it back into phase I for the required 5 hours, and made the required entry listing the maneuvers flown, and am now legal for anything I'm likely to want to do.

There are many acro accidents, almost always fatal and they are for the most part not done by novices, rather it is usually some of our experienced pilots that do them close to the ground and make a mistake. If you are low a small mistake can be large, too much to overcome.

Please, PLEASE don't do acro down low with a passenger. Not wives or girlfriends or boy friends or pet cats, and most of all don't do it with kids. I have spun safely, and recovered no problem, but I started at 15,500 over Fon du Lac.
By the way a stock J3 Cub can spin very nicely and recover.

Quote:

---

Originally Posted by Dana

Did you also search "acrobat" in the first pass? Could well be that the person writing a report used "acrobatic" instead of "aerobatic", since the former is, after all, the word used in the regulations.

---

Cool, I DID get another 61 hits. I'll add those and edit the original table.

Ron Wanttaja

Based on Dana's comments, I added "Acrobat*" to my search term and got additional data. This is updated, vs. the original.
----------------------------------------------------------------------------------------------------
For those who don't know me, I do a lot of statistical analysis of homebuilt aircraft accidents. Wes Liu suggested that I take a look at aerobatics-related accidents, specifically those that occur in airplanes not approved for aerobatics, and to low-time pilots trying to teach themselves aerobatics.

Ran some analyses over the weekend, and post addresses the results. But I'm a Systems Engineer by trade, so I have to document the methodology, first.

I used my copy of the NTSB accident database, and the accidents I examined ran from 1982 through 2017. I ran a search for "Aerobat*" or "Acrobat*" in both the Probable Cause and Narrative sections. In my experience examining homebuilt accidents, the investigators would use terms like "Loop" or "Roll", but would also use the word either of those terms at some point in the description. I included planes with both Standard and Special airworthiness.

This produced about 870 "hits." However, the use of the word "Aerobatic" or "Acrobatic" in the description does not mean the plane had an accident while performing aerobatics... often, it was a mere description of the aircraft type ("...the aerobatic aircraft had been purchased two years earlier...."). These accidents were eliminated.

The next step was to eliminate cases where the fact that aerobatics were involved was not a major factor in the accident. This eliminated instances like landing accidents that occurred after practicing aerobatics, etc. I made a judgement call in cases where mechanical failure manifested itself during the performance of aerobatics. If the failure was powerplant related, I eliminated it, but kept cases of structural failure. This last was for obvious reasons, but in a few cases, the root cause of the structural failure was a construction issue...these were eliminated as well.

I also eliminated accidents that occurred at air shows.

With this winnowing-out completed, I had526 accidents left. The next stage was to identify those which were approved for acrobatics, and those that were not. This was not all that simple. Technically, most homebuilts can legally perform aerobatics (there are a few, like most Fly Babies, that have aerobatics prohibited in the operating limitations). But there are some that really aren't intended for acro. So I flagged these as "non-acrobatic aircraft."

Similarly, many older production airplanes can legally perform aerobatics simply because the certification standard when they were introduced didn't actually ban acro. On these, too, I generally made a judgement call... considering a Waco biplane an aerobatic aircraft, while flagging the J-3 Cub otherwise (I know there are clipped-wing Cubs, etc.) I did have a couple of exceptions to this, considering the Luscombe and the Globe Swift as aerobatic airplanes.

Plenty of room for argument here, for sure, but that was my original cut. There were several types that I didn't recognize, these weren't included in the comparison.

Anyway, here are the updated result (updated from Post #1)

	Non-Acro	Acro
Total Accidents	128	383
Median Total Pilot Time	991	2012.5
Median Time in Type	98.5	97.5
Maximum Total Time	36500	32316
Minimum Total Time	16	75
Maximum Time in Type	6727	4506
Minimum Time in Type	1	1
Fatalities	113	317
% Fatal	88%	83%

The first thing that might catch your eye is the comparison between the median total pilot time. The pilots involved in accidents with aerobatic aircraft had about twice as much flying time than those who crashed flying non-acro planes (note that this difference is SIGNIFICANTLY greater, vs. the original in Post #1).

This update also now shows the time-in-type for both classes to be about the same.

88% of all acrobatics-related accidents in non-acro airplanes are fatal, vs. 83% of the aerobatics-airplane crashes. A lot of these accidents occur at low level, precluding any emergency exit. But in the higher-altitude cases there were several instances of pilots using a parachute to abandon damaged aircraft. Someone performing aerobatics in a plane meant for it is probably more likely to be wearing a chute...and an aerobatic airplane usually gives the pilot better egress in an emergency.

The most interesting result came when I plotted the pilot total time as a histogram:
http://www.wanttaja.com/acro.jpg
(you might have to hit "reload" to get the updated version)

This figure plots the percentage of the total accidents for each type vs. the pilot's total flight time. Note how relatively few accidents occur to lower-time pilots flying aerobatic airplanes. A lot of this is probably driven by the fact that few 80-hour private pilots buy Pitts Specials! But it does indicate that a lot of accidents in non-acrobatic airplanes involve relatively low-time pilots.

Ron Wanttaja

That's really great info! Thanks a lot! There is definitely a Sport Aerobatics article in that data.

Great stuff.

Wes

I'll repeat the great data! You're work is always wonderful and greatly appreciated.

Ironically, I am interested in the data that you threw away. In other words, how many aerobatic airplanes/pilots have fatal accidents while not intentionally performing aerobatics? Again, such as in the pattern (takeoff, approach and landing) and buzzing friends?

Thanks, the not-so famous Ron.

Quote:

---

Originally Posted by Ron Blum

I'll repeat the great data! You're work is always wonderful and greatly appreciated.

Ironically, I am interested in the data that you threw away. In other words, how many aerobatic airplanes/pilots have fatal accidents while not intentionally performing aerobatics? Again, such as in the pattern (takeoff, approach and landing) and buzzing friends?

---

Well, it's always hard making such a comparison; I've got about 4400 overall homebuilt accidents in my 1998-2018 database, but only a few dozen of each of several homebuilt types. But here's how the percentages come out for fatal accidents.

Fatal Accidents	All	Pitts	Skybolt	Christen
Accidents	1059	16	13	13
Pilot Miscontrol	30.5%	31.3%	23.1%	15.4%
Builder Error	4.5%	0.0%	0.0%	0.0%
Maintenance Error	4.0%	0.0%	0.0%	0.0%
Undetermined Loss of Power	8.8%	0.0%	0.0%	0.0%
Engine Mechanical	3.1%	0.0%	0.0%	0.0%
Exhaust or Turbo	0.0%	0.0%	0.0%	0.0%
Engine Controller/Electronic Ignition	0.3%	0.0%	0.0%	0.0%
Fuel System	1.4%	0.0%	0.0%	0.0%
Landing Gear/Brakes	0.2%	0.0%	0.0%	0.0%
Other Mechanical	4.2%	0.0%	0.0%	0.0%
Fuel Exhaustion	2.9%	0.0%	0.0%	0.0%
Fuel Starvation	1.2%	0.0%	7.7%	0.0%
Carb Ice	0.8%	0.0%	0.0%	0.0%
CG or Weight	1.0%	0.0%	0.0%	15.4%
VFR to IFR	3.7%	6.3%	0.0%	0.0%
Disorientation	0.7%	0.0%	0.0%	0.0%
Manuevering at low alt	11.6%	37.5%	38.5%	38.5%
Inadequate Preflight	0.7%	0.0%	0.0%	0.0%
Fuel Contamination	0.8%	0.0%	0.0%	0.0%
Midair Collision	2.5%	0.0%	0.0%	0.0%
Turbulence/Winds	0.6%	0.0%	0.0%	0.0%
Loss of Control (Unknown)	4.4%	6.3%	0.0%	7.7%
Taxi Accident	0.3%	0.0%	0.0%	0.0%
Airframe Icing	0.2%	0.0%	0.0%	0.0%
Undetermined	4.7%	0.0%	7.7%	0.0%
Other	6.9%	18.8%	23.1%	23.1%

Obviously, maneuvering at low altitude is a major factor on these aircraft. However, certain other aircraft types have elevated levels here, too. This is a plot of all accidents, not just fatal ones.
Attachment 8848

Quote:

---

Originally Posted by Ron Blum

Thanks, the not-so famous Ron.

---

Maybe change that to the non-INfamous Ron... :-)

Ron Wanttaja
Ron

Nice work, as always, Ron! An interesting look into the nooks and crannies!

Jim

Why did you exclude "engine failure." Have you never heard of "Tillie" from WW-II and her diaphram for Spitfires to fly negative "g's"? Or an inverted flight tank? Saber jets, barrel roll, Phantoms oblique loop into chandelle. Don't neglect "inverted oiling system" as a missing element in aerobatic capability. This is WW-I stuff. SPADS and Nieuports? And of course can we clump oil and fuel tanks together for two strokes? Gas Turbines are beyond Brinnell with ETOPS and run dry for endurance right side up.

Quote:

---

Originally Posted by 2ndsegment

Why did you exclude "engine failure." Have you never heard of "Tillie" from WW-II and her diaphram for Spitfires to fly negative "g's"? Or an inverted flight tank? Saber jets, barrel roll, Phantoms oblique loop into chandelle. Don't neglect "inverted oiling system" as a missing element in aerobatic capability. This is WW-I stuff. SPADS and Nieuports? And of course can we clump oil and fuel tanks together for two strokes? Gas Turbines are beyond Brinnell with ETOPS and run dry for endurance right side up.

---

Because it was not relevant to the question I was asked. To recap: "Wes Liu suggested that I take a look at aerobatics-related accidents, specifically those that occur in airplanes not approved for aerobatics, and to low-time pilots trying to teach themselves aerobatics." Mechanical failure accidents were thus excluded. Cases where the power was lost due to fuel feed issues, etc. would be included as the NTSB would still rule the cause as being related to the aerobatics.

Ron Wanttaja

My initial point of view comes from being asked to determine if the A-1 (AD) production line could be reopened. This was in 1965. As a first step I was asked to look at Navy Safety Center San Bruno data which included aspect and angle off for ballistic hits. You might think after the decision was made not to reopen that I would never be asked again about Survivability. After all a former VX-5 pilot took that on as his specialty for the next 13 years. Typically, in wartime about half of the losses are "operational" and not combat related. My experience could only be comparable to aerobatics if I saw the types of trajectories created as related to trajectories in combat. Guess what? I'm not an engineer of any kind. I'm a physicist. So when I was asked to go out to China Lake to see what the wife of the AIM-9L Program Manager was doing with Army BRL ballistics data to evaluate the survivability of a CH-53E helicopter ,I drove out to Mike Lab to see her work. This was more than a decade later. I will examine Wes Liu's suggestion. He was basically presenting a problem about the expectations of raw beginners. Something simple and cheap being pushed beyond it's safe use. There was a lot more data available to her and she did a much more detailed and well summarized job. Maybe you will have a similar expansion of your responsibility. I will you well.

Ron:

Do you have any data that indicates the aerobatic time / experience of the pilots, as opposed to the total time?

BJC

Quote:

---

Originally Posted by BJC

Ron:

Do you have any data that indicates the aerobatic time / experience of the pilots, as opposed to the total time?

---

Never seen any reference to aerobatic time, other than statements like, "an experienced acrobatic pilot"....

Ron Wanttaja

The sample size is gratifyingly small but that makes the percentages hard to apply. We have a wide spectrum of pilots who own the Pitts/Skybolts/etc. Skybolts, Stardusters, EAA Biplanes and similar ships appear to be owned and flown by pilots who are very casual aerobats. We see them do the occasional low pass at a fly in. I will note that many of these airplanes are flown by second or third owners, not the original builder. You have likely read the observation that "there are builders and flyers". My observations match that. The number of individuals out there who build an airplane specifically to fly a lot of aerobatics is getting fewer and fewer.

Many Eagles are on their second and third owners. We see many at IAC competitions but there is an equal or greater number that are 99% flown to fly-ins and the $100 hamburger. They are casual aerobats.

The challenge and concern with all of the casual aerobats is knowledge and training. The easiest aerobatic figure to have you experience your first unintended inverted spin is a hammerhead with an early kick. Those of us who do a lot of aerobatics want anyone who aspires to fly a loop or roll to have some spin training that includes both upright and inverted rotations. The good news is that Ron's stats report very small numbers.

Pitts owners tend to fit a different personality profile than Skybolt or even Eagle owners and pilots. The airplane's reputation as one of the most difficult airplanes on the planet to land means that you have to want to fly a Pitts. More motivation is required than for the other aerobatic airplanes. Stories abound, and I know of several, that someone buys a Pitts, flys one hop, and parks it. For sale again. Many landing accidents that only bend a wing are not reported, but Ron's stats on landing accidents in the different aerobatic airplanes likely shows significantly more Pitts accidents.

Pitts generally fly a lot more aerobatics than Skybolts, Stardusters, and the other ships. And they fly more complex aerobatic figures at lower altitudes. The floor for the Intermediate category is 1200' and for Advanced it is 660'. My FAA authorized practice box has a floor of 1000' AGL. This provides the opportunity for mistakes to show up in Ron's statistics but all of the participants practice at much higher altitudes before venturing to the bottom of the box where the consequences for mistakes go up in severity. I find it gratifying that although we do not record flight hours in all of the aerobatic boxes across the country, given the number of flight hours that we accumulate, Ron's numbers are low.

My original question was motivated by interest in whether our friends who buy an Eagle or a Skybolt and, without any formal instruction, go out to do their first loops and rolls, are having a large enough number of accidents that we should be really really concerned. In other countries you are not allowed to to exceed 60 degrees of bank and 30 degrees of pitch without formal training and an aerobatic endorsement on your pilot certificate. Not in the US. And Ron's numbers do not suggest that we need that in the US. Yes some of our friends will be stupid, but it is very hard to regulate stupidity and it appears that 99% of our friends are doing "good enough".

Thanks to Ron for slicing-and-dice-ing the data.

Wes
1000 hour Pitts S-2A owner