Awfully Slow Warfare Yes or No: Probability and Confidence in ASW By LCDR Tom Phillips, USN (Ret.)
ASW is tough. It is tougher if you are stupid. It is IMPOSSIBLE if you are stupid and don’t know it. Bishop’s Law #5 – You can’t deal with uncertainty by ignoring it.
A
SW is a game of probabilities. We use probabilistic terms all the time but perhaps do not think of them as uncertainties. Modern ASW seems to be steadily trending toward more and more binary: zero or one… yes or no. Absolute. Confidence Levels Right on the face of it, we have four confidence levels, NONSUB, POSSUB, PROBSUB and CERTSUB. POSSUB and PROSUB possible and probable, clearly probabilistic terms. Confidence levels CERTSUB and NONSUB sound very declarative, very binary, and don’t sound as probabilistic as POSSUB or PROBSUB. But look closer. CERTSUB might not be so CERT, as breaking waves, pods of dolphins, flotsam and jetsam, spurious radar contacts, have ALL-too-often been declared CERTSUB during real ASW. The less experienced against REAL submarines that ASW forces are, the less real ASW they have seen, and when the shooting starts, the less actual combat they have had, the more nervous they are, the more “things” will be declared to be submarine contacts, when the danger is real. (I can vividly still recall my old man telling me about sailing out of the Chesapeake Bay into Torpedo Junction in 1942, and the blizzard of “periscopes” detected: every breaking wave was a feather. U-boats had sunk 397 ships there in six months.) If you have never seen the threat submarine, when can you declare it to be CERTSUB? When you see floating tennis shoes and nuclear reactors? Even an oil slick after an attack can be a deception - such tricks almost as old as ASW itself. NONSUB. You are declaring that what somebody thought might be a submarine, something hard to find, something you can not even see, something YOU DID NOT FIND, is, therefore NOT THERE….. based on your extensive and exhaustive investigation….. You are declaring that you have proved a negative. There is no submarine here. Are you saying YES, you have stopped beating your wife? Or NO, you have not stopped beating your wife? Traditionally, NONSUB could only be declared if you found something there not a submarine which explained the alarm. Say for example, sighting Moby, with a 50-Hz ac-powered prosthetic flipper provided by the early Greenpeace folks after that unfortunate incident with the Pequod.
Rotor Review #153 Summer '21
Probability Terms Like the forward pass, where there are three outcomes relative to it (disregarding flag on the play); completion, an incompletion, or an interception, there are three outcomes which are relevant to the business of searching a Datum: Detection: The obvious. Missed Detection: A sub was there but you missed him; False Alarm: A sub was NOT there but you thought you had one. They are all probabilities. Probability of Detection versus Probability of Missed Detection They sum to 100%. Probability of missed detection is 1 minus probability of detection. If, given the submarine is really there, and probability of detection is, say, 75% (0.75) then probability of missed detection must be 25% (100% minus 75%, or 1.0 - 0.25). Probability of Missed Detection and Probability of False Alarm If you do not detect, there are the two other outcomes, missed detection and false alarm. They do NOT sum to 100%. They are in opposition to each other, but in a complex mathematical relationship (not simple addition or subtraction summing to 100% In fact, they will not sum to 100%), unlike the simple summing-to-100% relationship of detection and missed detection. RELAX! we are NOT going to the mathematical proof here, or elsewhere if I can help it. Both missed detections and false alarms are undesirable, but they are inescapable in real life ASW. An ASW Conundrum: Minimize False Alarms Versus Minimize Missed Detections You can minimize false alarms by raising the threshold of what is to be defined as “detection.” No calls unless Helen Keller can make the call. Every call will be a valid submarine contact. You’ll never be wrong that a called detection is not a submarine, since your criteria is so high that there are virtually NO FALSE ALARMS, but you will miss a lot of submarines, and that can be fatal to ships, people, mission, strategy, objectives, etc. A BGO. It is an ASW conundrum. In theory, at the extreme, you can eliminate false alarms, but logically, you can see that there has to be consequently more missed detections. You can minimize missed detections. To be damn sure you don’t miss the sub, you must call POSSUB at anything which COULD remotely be a whiff of a submarine. In theory, you will not miss a valid detection, but clearly, many will later be resolved (or declared) to be NONSUB, and technically fall 76