Continuing on with “How many rounds would you say make up a legitimate ‘reliability test’ for a pistol?”
This is a comment to the 2023 post that echoed a comment I failed to notice and answer on the 2017 post.
“ ‘One of the mathematical analyses presented in the original Facebook discussion was that 5 malfunctions per 1000 meant more than one malfunction in a 17 round magazine (8.72%).’
You need to link to this math because it isn’t correct if you assume independence.”
Here is the person’s explanation, verbatim.
“Claude – 5 in 1000 is a 0.5% failure rate, or a 99.5% success rate per shot. The odds of shooting twice in a row is 99.5% X 99.5%, or 99.5% to the 2nd power. The odds of shooting 17 rounds with no problems is 99.5% to the 17th power, or 91.8%, leaving an 8.2% chance of failure.”
The problem is that the assumed 0.5% failure rate was the result of shooting twice. So double counting the failure and then raising it to the 2nd power is an inherent structural flaw in the analysis. I don’t claim to be a statistician but occasionally a structural problem in constructing a mathematical analysis will be immediately obvious to me. If the output of a mathematical analysis is obviously divergent from reality (Glock 17s do not average more than one malfunction per magazine), it means the math is flawed in one way or another.
What is the reliability of a Glock 17? During one class I was teaching at the elite Rogers Shooting School, I had two failures to go into battery with the Gen 2 Glock 17 I used as my School gun. I pushed the slide closed with my thumb (OMG, a diagnostic malfunction clearance) and went on with the drill. The malfunctions seemed odd to me because they were out of the ordinary. Upon further reflection, I realized the teaching season was nearly over (October) and I had neither cleaned nor lubricated the gun all year. It had fired, with various types of ball ammo, somewhere upward of 15,000 rounds without a malfunction. The other School guns performed similarly.
An analysis that determined a Glock 17 had a probability of one malfunction per magazine flew in the face of my experience. That’s the kind of thing I keep an eye open for. If an analysis doesn’t match up with a broad segment of reality, there’s probably something wrong with the analysis.
Having built a model rocket does not make you a rocket scientist.
tacticalprofessor 
Yes that analysis is called the “survivor rule” which I mentioned in my comment. And we came to the same numbers. Calculating the various ways the gun could fail is difficult, so with the survivor rule you calculate the rate it runs perfectly and then invert it to see where it has at least one problem..
In my experience, Shooting a new carry pistol in the neighborhood of 400 rounds is sufficient to establish basic reliability. However, if a malfunction of any kind shows up, initiate a fix, then start over. Since no pistol can be considered infallible, carry two. Also, in a gunfight, getting shot in the hand is common, be equally proficient with each hand, and have equal access to both carry pistols. This recommendation is for Warriors. For worriers, carry a revolver also. Either way have two fighting folders that can be deployed with either hand. Following this method my involvement with statistical mathematics is minimized.