|
|
|
Topic: The odds that the next QA championship will be cancelled in progress.
| |
|
|
Author
| Message |
|---|
|
|
All that we actually know is that the measured mean time between failure for QA is on the order of 8 years, based on a single data point. (I'm referring to catastrophic failure with severe data loss, as opposed to simple crashes.) That's not sufficient to draw a conclusion, or even make an estimate, of a 12.5% chance per year of failure. You would have to know the probability distribution curve of failures to justify that, and it would have to be a flat distribution. But for complex systems in real life, that's very unlikely. A bell-shaped curve would be far more likely, and that's for an unmaintained system.
In this case, a much better guess than a flat distribution would be a bell curve spread over 16 years (not 8!), with the peak at 8 years. Given that we have just one data point, though. With just one data point, though, that's still a really really bad guess.
Since coin flips have been used as an example in this thread, I also want to point out the importance of the word "unmaintained" that I slipped in a couple of paragraphs ago. A series of coin flips is an example of independent trials. In each flip the probability of heads or tails is exactly the same as it was in the previous flip. You cannot demonstrate that this is the case for QA, or for any other software system that is being actively maintained. The probability of failure is changing with every revision to the system!
A coin can't learn from experience, but a software system (through it's programmers and operators) can. As a professional software engineer and systems consultant with 29 year so of experience, I can certainly attest to the fact that there is no such thing as bug-free software or risk-free operating and recovery processes. I can also attest to the fact that software systems don't always get better with time. The probability of failure will often increase before it decreases. Sad but true. If Miguel flies into a panic and starts making frequent changes to the QA code, I would be worried. I doubt that's going to be the case.
I think it is fair to assume that Miguel is a really smart guy who knows a lot about the pitfalls of software development and systems operation. The quality of this site stands as strong evidence of that. The reason I mention this is not to flatter him, or to reassure you that his code is ever going to be perfect. But I believe in my heart that Miguel is the kind of guy who is sure to have learned from this failure, and I believe he has undoubtedly learned far more than just the immediate cause and consequences of the flaw that brought the system down and destroyed the integrity of the database.
I am confident that Miguel has learned big lessons about the importance of implementing data consistency checks on the live database and on backups, and of implementing roll-forward recovery so that as much good up-to-date data as possible can be salvaged in the event that corruption is detected. If I were Miguel, I know I would not be sitting still. I would be writing and testing new code based on the lessons learned from this failure, with my highest priority being improved recovery for any future problems.
|
|
|
rhsatrhs
You state that my calculation is flawed. You not only dispute my guess that the chance that tournaments will be cancelled at some point in the next year is 12.5%. You seem to also dispute my formula. If the was proof that my 12.5% was correct you still disagree with my formula for calcu;lating the chance of failure over the 2.4 years.
You don't offer an alternative estimate or calculation. I dont mind that you say I am wrong but I want to see your method of doing this calculation.
|
|
|
Blutigeroo,
If you flip a coin 1000 times the chance that you will get 500 heds is 0.02523 or 2.523%
The chance that the number of heads will be greater than 484 but less than 516 is 67.3%
|
|
|
I sit corrected, Phys ...
(1000 choose 500) / (2^1000) = 0.0252250182
Regarding an alternative, there is none. Your calculation is the best but it ain't much! 
|
|
|
Bluetigeroo
I think we agree that the odds that next Anniversary tournament will becancelled in progress is less than 50%. The tricky part is less than 50% but more than ?%. I think it is greater than 1%. Is it greater than 10%? I think that is an important question to many members. I think most us can handle 1%. QA helps me play better in over the board tournaments, I will continue to play even if there is a 40% chance of future tournamnets being cancelled.
also in this thread I wanted to provoke a math discussion.
|
| Previous 1 2 3 4 Next |
|
|
|
