[TEC] Dud torpedo chance mod?

[Sniper297]

Got an idea from another thread, wondering what would be a good compromise for dud torpedoes? 10% chance, 5% chance?

Main problem I found with realism in the stock 1.4 game;



These are the specs for the Mark 23, which was designed with the mod 13 contact exploder AFTER the problems with the early versions of the original mark 6 mod 1 were fixed. Yet it has the same failure specs as the earliest mark 14 with the mark 6 mod 1 exploder, and same chance of premature detonation with no magnetic exploder at all.

From what I've read 99% seems a bit excessive anyway, but leaving reality aside for the purpose of gameplay, what would be a good failure percentage? 5% would mean 1 out of 20, 10% 1 out of 10. Mark 10 had an older mark 3 contact exploder so historically had far fewer problems than the untested mark 14, so my thinking would be 5% on the mark 10, 10% on the mark 14 would be reasonable, providing just enough chance of failure to make the game interesting without making it frustrating to the point where the average player gives up in disgust and disables the duds altogether.

For the Mark 14 premature chances are 60% for a 10+ meter high wave, 30% for 5 to 10, and 2% for 0 to 5. My thinking is to cut that back to 20, 10 and 2, which in average conditions would leave the player in a GATO with 18-20 reliable fish out of 24. Sort of reliable, still a 0.5 chance of a circle runner (which should probably be INCREASED at least a little, that's only 1 out of 200 fish) and a 70% chance of 10 foot depth variation which might be better eliminated since we're all aware of it and set the depth shallower anyway.

So fire away, flamewars and death threats welcome, can't guarantee not to toss all countering opinions overboard and just use my own personal preferences, but a little feedback might influence my decisions.

[TorpX]

... and a 70% chance of 10 foot depth variation which might be better eliminated since we're all aware of it and set the depth shallower anyway.

This, for me, is the problem with a Torpedo mod, any torpedo mod. If we substitute numbers that we consider appropriate/better/more pleasing, we know what the chances are, and this makes the whole deal rather artificial. The real skippers had to figure it all out by trial and error, wasting fish over the course of several patrols, while we know before we launch the first torp.

That said, a 99% chance of a dud, seems insanely high under any circumstances.

[Sniper297]

Apparently not. While I was playing around I found the speed settings and default depth, "wonder what happens if I do THIS? FIRE IN THE HOLE!!"

So I reset the depth from 1.8 to 6, since the average depth setting is 20 feet I can just leave it instead of having to change it every time. Then I swapped the settings for speed 1 and speed 0 so it would default to high speed - that didn't work, it disabled the speed setting switch for some reason.

So I changed it back the way it was.

Settings for the mark 14;
Dud chances (3);
0-35 - 1
35-70 - 10
70-90 - 10

So anywhere from 35 degrees to 90 degrees (majority of shots) I have a 10% chance of a dud. Out of 20 shots I should have 2 duds. 20 shots, 20 hits, no duds. 2% on prematures for all wave sizes, no prematures but out of 50 shots you would expect 1, 20 shots 0.4, so possibly that should be increased.

Next patrol shooting long range shots at a BB, I had set depth deviation chance to zero, circle probability and gyro problems I left at default - 2% chance of a circle and 2% chance of a screwy gyro with a max deviation of 50 degrees. Fired 3 fish at the BB on slow speed at 8000 yards - all three turned left 50 degrees and went off to la-la land. Fired two more from 5000 yards, both turned left 50 degrees. Fired two more on high speed from 3000 yards, both turned left 50 degrees. Two more at 1500 yards, can you guess what happened next?

Okay, so reloaded the in base save and tried again, same scenario - all shots went hot straight and normal, all hit the BB from various ranges. Seems like the dice are loaded to "when it rains it pours", whatever the chance percentage is, it's gonna use it all on the same patrol. Don't look to me like the random part of the programming is working very well, it seems to be all or nothing.

[TorpX]

Interesting...

If only we could look at the source code - what would we find?

[Sniper297]

Looked over the EXE with a hex editor, at the bottom;

ALL.PROGRAMMERS
NOTE.WRAP.IT.UP
SUITS.SAY
READY.OR.NOT
SHIPPING.NEXT.WE&%6
π^dddddd.xxxx???>??
FATAL.EXCEPTION.ERROR
DIVIDE.BY.ZERO#@?!!!!

[Sniper297]

Alrighty then, reset for dud chance 15 (3.3 out of 22 fish) and premature 10 (2.2 out of 22). 14 shots, 13 hits, 1 miss (snap shot at a turning destroyer, what did you expect?). Then one dud followed by two prematures, 5 for 5 hits after that. So the prematures were just about right although the random chance factor makes the 2 in a row rather unlikely, two in a row following the single dud is even more unlikely. The miss could have been a dud, impossible to tell if it doesn't hit something. One dud out of 22 fish on one patrol followed by 5 duds out of 22 on the next patrol is withing the realm of possibility and would average out the odds while making the skipper tear his hair out. Not sure that's the effect I'm looking for, possibly I should tone it down a bit and keep track over several patrols in a career.

One problem with testing this, a dud hits and bounces off and you get the announcement "it's a dud", but unless you're in external free camera view tracking the torpedo and see the splash there's no other indication of a premature - "Firing tube 1!" followed by nothing at all, check the attack map and the fish is gone. Other problem of course is misses, if I fire and miss there's no way to tell if the miss would have been a dud or a boom, so no way to keep track of the actual average unless I never miss. Possible but tedious since I'll have to avoid any kind of chancy shot during the entire test, that one miss out of 22 shots in the first patrol made the dud percentage test meaningless.

[TorpX]

I can't explain the odd results, but here are some of my thoughts.

I would try to do the tests in a simple mission, where you could readily observe the results, and repeat as often as needed. Maybe only vary one element at a time; at least at first.

Another thing, we ordinarily think the game rolls the dice for our torps at launch (or impact) because that is the way we would do it in a game. But, perhaps the game determines duds and malfunctions when we load the torps at base, so that changing files and replaying a save, might have no effect.

About the gyro problems; maybe the TDC itself was acting up? Otherwise, it is hard to fathom your results.

[deicide]

Actually I think the thing your not taking into consideration is that the perentage chance is for EACH seperate torpedo fired. It is not cumalative, so with a 10% chance for each torpedo to be a dud it does not mean that after firing 9 you now have a 90% chance for the next one to be a dud. EACH torpedo percentage chance is calculated seperately so that 10th torpedo still only has a 10% chance of being a dud.

Hope this makes sense!

-deicide

[Sailor Steve]

Or, as a friend of mine used to say, "Dice have no memory".

Great answer, deicide, and welcome aboard!

[Sniper297]

"But, perhaps the game determines duds and malfunctions when we load the torps at base, so that changing files and replaying a save, might have no effect."

Probably. At first I was reloading the automatic save in base, that was the one where I got all the fish turning left 50 degrees off course. TDC problem unlikely, I was shooting at the KONGO that's loafing along at 1 knot in my "Gone Asiatic" mod - BB is steaming from left to right, so the usual TDC bug would have deviated the fish to the right rather than the left. Auto targeting is ON for all these tests. Now I'm deleting all saves and starting a new career after changing the data in the Torpedoes_US.sim file, but the suggestion about a single mission is probably the way to go - half a dozen YAMATO BBs at anchor maybe. My only concern there, does the game roll the dice the same way for careers as it does for single missions?

"Dice have no memory", unless they're loaded, of course!

deicide, welcome aboard, Abandon Hope All Ye Who Enter Here. I've been to Las Vegas a few times in my misspent youth (even got married there once godhelpus) so I'm familiar with the science - flip a coin and it has a 50% chance of heads or tails, get heads 20 times in a row the chances of the next flip are still 50-50 it will be heads or tails. The law of averages still states that out of 1000 flips approximately half will be heads and half will be tails. The law of random chance means that there's always a possibility that the first 20 flips will all be heads, but since I'm from Chicago I want to examine that coin carefully to see if someone is cheating. Problem with a game is we don't have any way to check if the dice are loaded or the coin is weighted on one side, so we're stuck with guesswork.

I suspect I'll have to fire 1000 mark 14s and see what happens, all I know is in the past having duds on in a stock game career means an unreasonable number of failures. Which again, is historically accurate, but doesn't make for a fun game.

[Sniper297]

After editing a test mission to place half a dozen anchored Yamatos, ran the mission a few times and noted more duds from the stern tubes than from the bow tubes, dunno if that was random chance or not. Changed the circular run probability to 99%, set depth 40 feet, sub at 60 feet running at 2 knots - all fours shots (SARGO class) ran circular, two left and two right, in very tight circles. They ran around twice before the first two exploded over the sub on the third circle, the next two made one more circle for a total of four before getting close enough to wreck the sub.

Objections so far;

1. In real life WWII fleet subs had two different sonar sets with three different sound heads, one set was always designated to track the torpedoes "hot straight and normal" or otherwise, in game there's no warning from sonar that a fish is deviating from course or starting a circular run. First warning you have of a circular run in game is when the fish blows you up.

2. Same thing for a premature, in real life sonar would report the explosion of a premature if it was more than a few seconds from the estimated impact time. In game unless you are in external camera and see the splash, you have no idea what happened to the fish.

Duds are reported, but other failures are not, for any kind of realism that situation needs some kind of fix. There's also no variation in the gyro problems, a circular run is always the same diameter circle and a deviation from course is always exactly 50 degrees.

Summary, I'm not sure the dud programming is even worth trying to fix, whatever I do it will still be too lame for prime time.

[TorpX]

Originally Posted by Sniper297

There's also no variation in the gyro problems, a circular run is always the same diameter circle...

I think this makes sense. The problem was likely caused by the rudder being jammed over hard, so the torpedo would make a tight circle; hence the danger - there was little time to react.

[Sniper297]

Problem in game is there's plenty of time to react, the circle is so tight it passes ahead of the sub twice before the sub forward movement puts it in the path of the torpedo on the third circuit. But unless you're in external cam or the attack map there's no warning to react to. This is yet another part of the failure of the crew AI to simulate what a real crew would do, notice and report things that a captain would need to know.

[TorpX]

Well, O'Kane's boat suffered that fate, in spite of the fact that it made a couple circles before it hit.

I don't agree that it is certain that the crew would know about a circle runner. Your sound man could be dutifully tracking the first two torpedoes you fire, oblivious to the third looping around and coming back.

[Sniper297]

Can't remember if it was THUNDER BELOW or some other book, but the author specifically said that each torpedo was tracked as it was fired, and any deviation in bearing was immediately reported.

"Your sound man could be dutifully tracking the first two torpedoes you fire, oblivious to the third looping around and coming back"

Doesn't work like that. He'd be sweeping back and forth across the bearing and would hear the first or second deviating even though they would be further away than the third, and it's hard to imagine a situation where the closest torpedo sound would be drowned out by ones further away.

It's a failure in the AI programming, similar to the squawking parrot repetition of reporting the same wing escort as a new target every time they lose sight for a second and reacquire it, but don't think it's important to mention when the previously reported escort suddenly turns toward you and increases to 38 knots. Or even bother to report he's opened fire until you actually get hit.