Suggestion: improved spotting system

neistridlar · 04-22-20, 03:57 PM

As is my custom, I have spent way too much time over the last few days thinking about how one could make a spotting system for wolfpack that would feel realistic, and reward playing as is described in the literature from WWII. It is based on my own experience as a lookout on recreational vessels at sea, together with reading a translation of the U-boat Commanders handbook. Also some Ideas are lent from other games, like KSP and Cold waters.

So, here is what I envision: Each ship will have a number of lookout points. These lookout points have a sector in which they can look. They also have a scan period/frequency, a detection threshold, and a spotting score. The scan period is in part for performance reasons, and in part to emulate how a lookout may sweep back and forth along its sector. Detection threshold and spotting score will make more sense later.

On the U-boat there will be 3 checkpoints, one at the top of the bow, one on the top of the stern, and one at the highest point of the boat with all equipment extended (probably on top of the attack periscope).
1. First a simple check is made, whether any of these points are above a min depth plane. This plane would reflect the visibility of the water itself. In my experience this would typically be around 5m, but may vary between 1m and 20m. If no points are above the plane, visibility score is 0, and no further checks are needed.
2. Next the distance to the points are checked. If they are not within the visible range, same as before, no further checks are needed, visibility score is 0.
3. Next we check for obstructions. All ships are replaced with a very simple model, like a box, for performance reasons. A ray trace is done to check for obstructions. Again, if no points pass this check, visibility score is 0, and no further checks are needed.
This is where the real “meat” of this method is. In order to find a score we do a simple render. To keep the performance impact down the quality of this render should be kept as low as possible, think 1990s levels of graphics. Also, a low resolution and a narrow field of view should be used (similar to the field of view in the binoculars).
1. The u-boat is rendered diffuse light gray. Except for the glass on the periscopes, which are specular, in order to make them glint in the sun.
2. The sea is rendered as black fog. The thickness of the fogg is according to visibility, like with the depth plane.
3. The surface of the ocean is rendered so that it is transparent when viewed head on, and black at oblique angles. That way you can “hide” behind waves, but the boat may still be visible, even submerged at close range, and fully submerged in clear water. Also this mimics the way water reflects the sky at oblique angles more than right angles.
4. Ships render black.
5. Black fogg, according to visibility conditions.
6. Black “glare” post processing according to sun-direction.
7. Other effects, like spray from the periscope or wakes at high speed could also be rendered light gray.
Visibility score is calculated simply by summing up the values of the pixels in the render (black being 0, and white being 255). So if more of the boat is visible, either because it is closer, or for instance surfaced, and showing its broad side, the higher the visibility score is. While attacking out of the sun, while low on the sky, will blind the enemy, and render the boat almost invisible.
Here is where my own experience, U-boat Commanders Handbook, and Cold Waters come into the picture. Every lookout has a spotting score (from 0-100), that works similar to the “solution” in Cold Waters.
1. Visibility score is compared with the detection threshold.
2. If the visibility score is below half the spotting threshold, spotting score is reduced by 50 points.
3. From visibility score between half the spotting threshold and the spotting threshold, the spotting score is reduced between 0 and 50 points, proportional to how much lower it is.
4. Similarly, above the spotting threshold, up to 200% of the spotting threshold, spotting score is increased between 0 and 100 points, proportionally to how much larger the visibility score is. So at close range, you can be detected immediately, but at longer ranges detection takes longer.
Spotting scores above 0 may trigger warships to get suspicions, and act accordingly. The chance would increase as the spotting score increases.
At 100 points, you are definitively spotted. Great speed ahead

The reasoning behind the visibility score, is from my own experience spotting things at sea. Your eyes will frequently play tricks on you, leaving you to think you might have seen something, but often you can not find it again. Other times they will appear again, when you have closed distance, or it appears from behind a wave. This is especially true for fishing equipment, which like a periscope is just a small speck dipping in the waves. Also, as is written several times in the U-boat Commanders Handbook, at close range the periscope should be used sparingly, but frequently, and be kept as low as possible, with the ocean washing over it constantly. Using the periscope this way would then make the spotting score raise more slowly, or even be completely arrested by alternating between increasing and decreasing.

Oh boy does it feel good to get all that out. Now I invite you to tear it apart, and figure out why it would be good/bad, or how it could be better.

Onkel Neal · 04-23-20, 07:00 AM

Interesting. Yes, I agree that human lookouts are subject to stress when scanning the never-static ocean. You have suggested that those stresses and fatigue should be factored in, I agree.

Quote:

Spotting scores above 0 may trigger warships to get suspicions, and act accordingly.

Can you explain what behaviors a suspicious warship would exhibit?

We have debated exactly what an alerted escort would do to a torpedo explosion. Of course, it would initiate more aggressive search patterns but since they are already searching while escorting the convoy before alerted, the difference would not be massive. I'm a fan of Fidd's idea, bumping up the escort's abilities after the first hit. Once alerted, they would be a bit more "alert", more capable.

Also discussed: when a ship blows up, we assume the escorts would race to a search box that is 3000m around the attacked ship and this would be realistic and pose the greatest challenge for the players. We have tweaked the night surface detection ranges to allow surface night attacks inside the convoy as the famed Kretschmer managed; but if the first attack produces flares and your surfaced sub is surrounded by ships 800m away, you are bound to be detected. Maybe our flares are currently too effective? Maybe our ships are spaced too closely together... but we used historical info for metrics such as spacing and range. We still are working on zig-zag behavior, as evident in patch .23 and we also want to have better damage/sinking behavior in the future. One hit should not equal an automatic sinking, and damaged ships should get left behind or struggle to stay in formation. Stuff we are going to work on.

Oscar has built an impressive visibility and detection system, it has been revised and balanced over the last year. It's the secret sauce of the AI. In every situation we ask, what would the enemy do in real life? How would the escort's systems fit into our game? What range can they detect you at, in any given conditions (clear, day, foggy, night, full moon, etc)? How often would they slow to listen? How do they handle search patterns? Likely areas to search? What range should flares illuminate? We still need to upgrade the periscope detection element, but that's on the list of to-do items. We've seen areas that need to be strengthened and received player feedback, and where necessary, tuning and improvements have been made. The goal: realistic AI so you can assume the enemy acts realistic, behaves in ways that makes sense so you can plan tactics accordingly. All fit in the scope of a real-time computer game that balances difficulty with fun.

neistridlar · 04-23-20, 10:15 AM

Quote:

Can you explain what behaviors a suspicious warship would exhibit?

I left that open because I don't have any historical sources to go off. But The first thing that comes to mind is to change course and speed more often. Possibly point the ship towards the suspected location for better sonar performance, maybe send off a ping or two. I'm guessing the escorts would not stray too far from the convoy to investigate though without a firm contact.

blackswan40 · 04-23-20, 11:05 AM

WWII Convoy Course Canges and Escort Tactics from Aces of the Deep Manual when Naval sims had Manuals the size of the Yellow Pages full of source data of substance no pun intended and other good stuff.

ElJammos · 04-23-20, 11:42 AM

From my own experience playing the game, it's very hard to position yourself in between lines if convoys without getting spotted (and by that I mean that the merchant ships pick up on the periscope, even when raised for a very short amount of time). Now, most times I play in daytime, so that might contribute to that as well. I don't think wider spacing is the issue there - as it's also quite fun to position there - but mostly the visibility threshold of the merchant ships.
Do the escorts, for their search box, also take into account on which side the torpedo impacted?
At any rate, very sensible stuff indeed. I'm looking forward to that update indeed!

blackswan40 · 04-23-20, 01:33 PM

Hi Eljammos When I play Wolfpack Setting up a Night Mission date 25th December 1941 Time 22:50 as the Mission starts i go on the Bridge search with the Uzo I see a Corvet on my 190º-210º so i turn toward them from my base course of 300º i increase my speed to flank.
Then as i get closer and the rest of the convoy comes into view if theres a large enough gap between the lead side escort and the very front lead escort.
I point my bow between them when a get to 4500-5500 away from the Convoy i dive to periscope depth lower my speed to slow as i and the Convoy close on each other i set my speed to dead slow or may even stop if they are now heading traight for me or approaching at a 45º angle i get the speed of the Convoy from one of the ships ive identified then when they are at 2200-3500 meters away i fire all four bow tubes at different ships if i am inside the convoy even better i can fire my stern tube as well sometimes i will cross the wake of a nearby merchant to throw the escorts off.
Also i fire at the furthest ship first then then next furthest away ship and nearest one last so i might have a 2-3 minuet gap between eels so they all hit within 6-15 seconds confusing the escorts making them think theres more than one Uboat attacking the Convoy.

I use Gutted's Solution Solver Great piece of Kit Link Below

https://www.subsim.com/radioroom/dow...o=file&id=5118

Onkel Neal · 04-24-20, 09:01 AM

Quote:

Originally Posted by ElJammos

Do the escorts, for their search box, also take into account on which side the torpedo impacted?
At any rate, very sensible stuff indeed. I'm looking forward to that update indeed!

Very good question, I don't know for sure but I would guess, no. In 1941, one would suppose that any attack would most likely becoming from outside the convoy. Sure, the escort could search inside the convoy but there are a lot of limitations imposed by surrounding ships and their screw sounds, and not running into them. Plus, you would figure the odds are much greater that a U-boat would find it easier to get into position 1500m outside the convoy perimeter than holding fire until working into the convoy.

Of course, we know some U-boat aces did manage to get inside the convoys, and that seemed to be a celebrated achievement, not a simple (or safe) task.

One other thing; currently when the ships fire flares it basically extends their visible detection range. Ideally, there should be less certainty that a flare will make a sub visible.

Imo, the trick for a good game is to have a variety of factors: day/night, weather, moon/no moon, alerted/not alerted, sea state, and a random experience/crew quality factor for each ship. Combine these and the gameplay is realistic and improved. It will keep the players on their toes in honest fashion.

Thanks, Andy, for posting that excerpt!

Fidd · 04-30-20, 08:40 AM

Quote:

Originally Posted by Onkel Neal

Interesting. ...I'm a fan of Fidd's idea, bumping up the escort's abilities after the first hit. Once alerted, they would be a bit more "alert", more capable.

I think it'd be worthwhile differentiating between AI behaviours of the convoy being "on alert" ie course-changes, flares, searchlights etc, and the "lookout alert" - as described elsewhere, where the AI lookouts become both quicker at acquiring a U-boat or periscope thereof, and also do so at increased ranges.

I would suggest that the lookout-alert phase be much longer lasting, in the order of 3 hours, very gradually decreasing over that time to the un-engaged value, and resetting to the lookouts alerted values on each occasion a U-boat is acquired or a torpedo hits. As most torpedoes in this period were steam-driven, consideration should be given to load-outs for torpedo captains, and if steam driven torps are used, then the effect of a near-miss with a torpedo leaving a visible wake should likewise create both a lookout alert and a convoy alert. Consideration should be given to the nearest AI escort heading back up the seen torpedo track.

I think the lookout alert phase should be quite a subtle effect to start with, with the precise values and rates of change related to light-level and gameplay feedback. The main point of it is to create uncertainty in the U-boat captains mind as to whether or not he's able to be seen at a given range. Feedback from the hydrophone operator will be important here, as the sound of DC attacks, torpedo hits or rapid-pulse Asdic will all indicate a higher than normal likely lookout alert status; in much the same way as visible flares/searchlights indicate a convoy alert.

John Pancoast · 05-02-20, 03:32 AM

Quote:

Originally Posted by Onkel Neal

Very good question, I don't know for sure but I would guess, no. In 1941, one would suppose that any attack would most likely becoming from outside the convoy. Sure, the escort could search inside the convoy but there are a lot of limitations imposed by surrounding ships and their screw sounds, and not running into them. Plus, you would figure the odds are much greater that a U-boat would find it easier to get into position 1500m outside the convoy perimeter than holding fire until working into the convoy.

Of course, we know some U-boat aces did manage to get inside the convoys, and that seemed to be a celebrated achievement, not a simple (or safe) task.

One other thing; currently when the ships fire flares it basically extends their visible detection range. Ideally, there should be less certainty that a flare will make a sub visible.

Imo, the trick for a good game is to have a variety of factors: day/night, weather, moon/no moon, alerted/not alerted, sea state, and a random experience/crew quality factor for each ship. Combine these and the gameplay is realistic and improved. It will keep the players on their toes in honest fashion.

Thanks, Andy, for posting that excerpt!

I'm not so sure all those different factors are a good thing. SH3 has, or tries, to have most of them and it seems to cause more problems than solve with one factor seeming to clash with another at times.

Case of "to many cooks in the kitchen" type of thing.

04-22-20, 03:57 PM	#1
neistridlar Swabbie Join Date: Mar 2020 Location: Norway Posts: 11 Downloads: 3 Uploads: 0	Suggestion: improved spotting system As is my custom, I have spent way too much time over the last few days thinking about how one could make a spotting system for wolfpack that would feel realistic, and reward playing as is described in the literature from WWII. It is based on my own experience as a lookout on recreational vessels at sea, together with reading a translation of the U-boat Commanders handbook. Also some Ideas are lent from other games, like KSP and Cold waters. So, here is what I envision: Each ship will have a number of lookout points. These lookout points have a sector in which they can look. They also have a scan period/frequency, a detection threshold, and a spotting score. The scan period is in part for performance reasons, and in part to emulate how a lookout may sweep back and forth along its sector. Detection threshold and spotting score will make more sense later. On the U-boat there will be 3 checkpoints, one at the top of the bow, one on the top of the stern, and one at the highest point of the boat with all equipment extended (probably on top of the attack periscope). First a simple check is made, whether any of these points are above a min depth plane. This plane would reflect the visibility of the water itself. In my experience this would typically be around 5m, but may vary between 1m and 20m. If no points are above the plane, visibility score is 0, and no further checks are needed. Next the distance to the points are checked. If they are not within the visible range, same as before, no further checks are needed, visibility score is 0. Next we check for obstructions. All ships are replaced with a very simple model, like a box, for performance reasons. A ray trace is done to check for obstructions. Again, if no points pass this check, visibility score is 0, and no further checks are needed. This is where the real “meat” of this method is. In order to find a score we do a simple render. To keep the performance impact down the quality of this render should be kept as low as possible, think 1990s levels of graphics. Also, a low resolution and a narrow field of view should be used (similar to the field of view in the binoculars). The u-boat is rendered diffuse light gray. Except for the glass on the periscopes, which are specular, in order to make them glint in the sun. The sea is rendered as black fog. The thickness of the fogg is according to visibility, like with the depth plane. The surface of the ocean is rendered so that it is transparent when viewed head on, and black at oblique angles. That way you can “hide” behind waves, but the boat may still be visible, even submerged at close range, and fully submerged in clear water. Also this mimics the way water reflects the sky at oblique angles more than right angles. Ships render black. Black fogg, according to visibility conditions. Black “glare” post processing according to sun-direction. Other effects, like spray from the periscope or wakes at high speed could also be rendered light gray. Visibility score is calculated simply by summing up the values of the pixels in the render (black being 0, and white being 255). So if more of the boat is visible, either because it is closer, or for instance surfaced, and showing its broad side, the higher the visibility score is. While attacking out of the sun, while low on the sky, will blind the enemy, and render the boat almost invisible. Here is where my own experience, U-boat Commanders Handbook, and Cold Waters come into the picture. Every lookout has a spotting score (from 0-100), that works similar to the “solution” in Cold Waters. Visibility score is compared with the detection threshold. If the visibility score is below half the spotting threshold, spotting score is reduced by 50 points. From visibility score between half the spotting threshold and the spotting threshold, the spotting score is reduced between 0 and 50 points, proportional to how much lower it is. Similarly, above the spotting threshold, up to 200% of the spotting threshold, spotting score is increased between 0 and 100 points, proportionally to how much larger the visibility score is. So at close range, you can be detected immediately, but at longer ranges detection takes longer. Spotting scores above 0 may trigger warships to get suspicions, and act accordingly. The chance would increase as the spotting score increases. At 100 points, you are definitively spotted. Great speed ahead The reasoning behind the visibility score, is from my own experience spotting things at sea. Your eyes will frequently play tricks on you, leaving you to think you might have seen something, but often you can not find it again. Other times they will appear again, when you have closed distance, or it appears from behind a wave. This is especially true for fishing equipment, which like a periscope is just a small speck dipping in the waves. Also, as is written several times in the U-boat Commanders Handbook, at close range the periscope should be used sparingly, but frequently, and be kept as low as possible, with the ocean washing over it constantly. Using the periscope this way would then make the spotting score raise more slowly, or even be completely arrested by alternating between increasing and decreasing. Oh boy does it feel good to get all that out. Now I invite you to tear it apart, and figure out why it would be good/bad, or how it could be better.

04-23-20, 11:05 AM	#4
blackswan40 LSH-2022 Liaison Officer Join Date: Mar 2008 Location: IN THE STEEL SHARKS VIRTUAL OCEAN Posts: 2,623 Downloads: 594 Uploads: 30	WWII Convoy Course Canges and Escort Tactics from Aces of the Deep Manual when Naval sims had Manuals the size of the Yellow Pages full of source data of substance no pun intended and other good stuff. __________________ LSH3 DOWNLOAD - LSH3 VIDEOS YOUTUBE - TUTORIALS by THEBERBSTER Last edited by blackswan40; 04-23-20 at 12:29 PM.

04-23-20, 01:33 PM	#6
blackswan40 LSH-2022 Liaison Officer Join Date: Mar 2008 Location: IN THE STEEL SHARKS VIRTUAL OCEAN Posts: 2,623 Downloads: 594 Uploads: 30	Hi Eljammos When I play Wolfpack Setting up a Night Mission date 25th December 1941 Time 22:50 as the Mission starts i go on the Bridge search with the Uzo I see a Corvet on my 190º-210º so i turn toward them from my base course of 300º i increase my speed to flank. Then as i get closer and the rest of the convoy comes into view if theres a large enough gap between the lead side escort and the very front lead escort. I point my bow between them when a get to 4500-5500 away from the Convoy i dive to periscope depth lower my speed to slow as i and the Convoy close on each other i set my speed to dead slow or may even stop if they are now heading traight for me or approaching at a 45º angle i get the speed of the Convoy from one of the ships ive identified then when they are at 2200-3500 meters away i fire all four bow tubes at different ships if i am inside the convoy even better i can fire my stern tube as well sometimes i will cross the wake of a nearby merchant to throw the escorts off. Also i fire at the furthest ship first then then next furthest away ship and nearest one last so i might have a 2-3 minuet gap between eels so they all hit within 6-15 seconds confusing the escorts making them think theres more than one Uboat attacking the Convoy. I use Gutted's Solution Solver Great piece of Kit Link Below https://www.subsim.com/radioroom/dow...o=file&id=5118 __________________ LSH3 DOWNLOAD - LSH3 VIDEOS YOUTUBE - TUTORIALS by THEBERBSTER

04-23-20, 11:42 AM	#5
ElJammos Swabbie Join Date: Dec 2017 Posts: 9 Downloads: 1 Uploads: 0	From my own experience playing the game, it's very hard to position yourself in between lines if convoys without getting spotted (and by that I mean that the merchant ships pick up on the periscope, even when raised for a very short amount of time). Now, most times I play in daytime, so that might contribute to that as well. I don't think wider spacing is the issue there - as it's also quite fun to position there - but mostly the visibility threshold of the merchant ships. Do the escorts, for their search box, also take into account on which side the torpedo impacted? At any rate, very sensible stuff indeed. I'm looking forward to that update indeed!