Reasoning as to why playable escorts are needed.

[Fidd]

This is a niche game, the nature of which is not widely popular in the way others have general appeal to the gameplaying public. It appeals therefore, I believe, to detail-oriented players, who prefer seeing good leadership and teamwork, and precision, and who value playing with a fairly regular but small number of friends as a single crew.

It is fairly usual for players to play around 2 or 4 times a week for 3 hours a session.

This means that variety of outcome, and being faced with new and/or unusual situations is critically important to retain players in the medium to long-term, else the repetitive nature of games ultimately leads to burnout and players turning to other games. We've all seen this, many many times over the years.

There are several ways to create this variety of outcome and game experience. They are:

1. PvP play via players manning escorts, and able to move from escort to escort.
2. AI escorts being directed to perform tasks at positions relative to the convoy, or to absolute positions.
3. Longer DC attacks (with less lethality) causing progressive damage.
4. Damage-control problems for the crew to prioritise and deal with.
5. Ability to operate optional tasks as a crewman at greater difficulty for reward.

I propose to lay out a suggested sequence for introducing these.

(1.) PVP. To start with, I'd create a single allied player, which we'll call "Convoy Commander (henceforwards "CC). He will have a map view of the convoy shewing the positions and type of all escorts, and the merchants, which can be zoomed in and out from within the convoy out to circa 100km radius, defaulting to 20km radius. On this map he can click on an escort, then click on any position within the area covered by his zoomed view of the map. He then may choose "Position" or "Position relative to the convoy" and then selects a speed from "at convoy speed", "at hydrophone speed", "at asdic and hydrophone speed", "at asdic speed", "at half ahead" or "at full-speed". The AI escort will then move to that position at the ordered speed, using as many sensors (visual, asdic and hydrophone) as that speed allows. On reaching the position, the escort then reports in, allowing the CC to set what he wants the AI to do, which might be remaining stationary, listening to hydrophones, conducting an asdic search etc.

So, you then have AI escorts being sent to locations to undertake tasks automatically, more or less as they do now, however, directed by a human mind, meaning that the players in the u-boats now are playing against human decision-making and planning, rather than AI. This makes best use of AI for the dull task, but improves it's tactics and creates more varied outcomes for the U-boat players. If the CC is not efficient in exploiting the escorts different capabilities, then gaps may occur in the escort screen, which u-boat captains may use to close with the convoy. Similarly, if he detaches escorts at high speed, then he may be unable to use hydrophones or asdic - again conferring an advantage to a u-boat. Conversely, if he detaches them at moderate speed, then he may detect a u-boat that was unable to "get out of the way". An AI escort making such a detection would preferentially attack the detected u-boat rather than continue with it's previous order to go to another position.

The 2nd phase of PVP would be to add "playable escorts". Unlike u-boat players, these players would be up to 4 or so per escort. Captain/Helm, asdic operator (also able to listen on hydrophones if no 3rd player), and hydrophones. These players can hop into any escort they please, and can see which escorts have detected a contact, and are notified when such a detection occurs. They then take-over the operation of that AI escort. If they then leave it unmanned, then it reverts to the last order it received from the CC.

The effect of this is that now a human player (CC) is responsible for the distribution of AI escorts, and directs them as he sees fit. Once an AI detection is made, or, a human player operating an asdic or hydrophone makes one, allied players can then take-over the nearest boats. They can see the orders given to the AI - "go here and do this", and can communicate with the CC to amend the AI order if needbe, so that the CC is aware of what's happening, and who may wish to redistribute AI escorts if the player operated escort needs more time to conduct an attack.

It follows from all this that two things need to change. Firstly it should no longer be possible to simply dive to 185m to be safe from detections, either from asdic or from hydrophones. However, if a u-boat goes deep, it will be safer, as it takes DC's more time to descend, the asdic operator will be "blind" at a greater range, and the u-boat can use that delay to move, rendering the DC's much less accurate. The accuracy of the DC attack will also be varied by the skill of the escort crew, rather than that of AI. Accordingly, I think the lethal proximity of a DC needs to come to circa 5m, rather than than 20m, but that DC's between 5m and 20m confer damage to the u-boat in order to erode functionality of the boat and to generate problems for the crew to contend with. Sometimes the nature of such damage may mean the loss of the boat, if the sum of damage warrants it. For example, if the damage causes leaks, the weight of water in the bilge may cause the boat to sink to crush-depth, if there's not a sufficiency of air to empty the bilge enough to arrest the descent...

Summary.

We now have a situation where every AI escort may actually be manned by players, but the u-boat captain cannot tell if it is - or isn't. A human player is - or may be - directing the defence of the convoy, making the AI escorts less predictable in where they go, and what they do when they get to a position. The AI escorts may combine onto a detected u-boat and so have several escorts attacking it - or they may hold their relative positions to help keep others at bay.

Because DC attacks/ASDIC searches are now much more likely to persist for an unknown period, but are likely to be much longer than now, it's also necessary that there's a further change: The way alerts operate needs to change, in order that a prolonged DC attack does not keep the convoy alerted for long periods, as in a multi-boat game this would mean the other boats not having a settled convoy to attack for that long period. I suggest that alerting and de-alerting be handled thus:

A u-boat is detected by an escort (AI or Human). Once the returning ASDIC echo or noise on hydrophones exceeds the value at which it would detect a u-boat, then the ability to drop DC's or other weapons is unlocked. The convoy then alerts for circa 6+ minutes, with a total of 17 minutes or so before it's settled on a new course and de-alerted. When an AI escort loses a contact, it resumes an ASDIC "lost contact" search. If it re-detects it - or detects another u-boat - within 10 minutes, the convoy continues it's de-alerting count-down, as the u-boat is no immediate threat to the convoy. The exceptions to this would be if the u-boat either hits the escort, or, if a torpedo wake is seen, or if the u-boat is seen to be at or near the surface.

So now, with all the changes laid out, any given u-boat being detected only alerts and "muddles" the convoy for a similar period than it is now, but, for the crew of the detected boat, they may be in for a protracted period of asdic searches and DC attacks/ During this, they will suffer non-lethal damage, possibly involving some loss of functionality - eg loss of electric rudder operation, or loss of 'planes (ditto). They may have taken on water, used unusual amounts of air to purge the bilge, lost instruments, and possible been presented with damage-control sequences and/or lost the boat. If they're undetected, they will have to assume that every escort may be human manned, and that therefore may be capable of guile and subterfuge. For example, if a human crew detects a momentary noise on the hydrophones, they may ask the CC to send a 2nd escort slowly towards where they think the u-boat is, "quietly", to that when they turn and rapidly acquire the u-boat on ASDIC, they may fairly quickly have a 2nd AI escort, or a now-manned 2nd Human Escort, to attack the contact with 2 escorts, and really slather it with DC's! - or conduct a "creeping attack"...

[Onkel Neal]

Excellent analysis!

[Fidd]

Thanks! Thoughts on an approach to damage-modelling may be found on idea #98 at https://www.subsim.com/radioroom/sho...=255408&page=8

and further thinking on playable escorts is at #100, same link.

[Fidd]

After reflection, I think there's a problem in implementing this which I overlooked when first thinking about it. In my mind's eye, I'd been considering it in terms of the convoy having been sailing for days or weeks, and the CC simply setting escorts off initially on effectively semi-random bearings and distances, because he could have no knowledge of where the opposing u-boats may be in relation to the convoy, or when they may attack.

This is of course, not actually the case. An astute CC would know that u-boats will be seeking to deliver 3 attacks in the usual 3 hour game, he will know that after spawning in, they will seek to get abeam of a ship with kingposts to acertain convoy heading, he will know that after 15-30 minutes they are likely to try to determine the speed of the convoy, probably by going all-stop, likely on the surface (at night) and likely at circa 5000m range before closing. He will likely bank on the first attack being launched with 4 steams, because that's the default load-out. He may believe, with reasonable likelihood that the u-boats have steams set to 40 knots, because that's the most popular setting. And so on and so forth.

Now, to a certain degree, it is a duel of wits, and over-reliance on anticipated behaviour, is likely to be poor for the convoy. However, some measures are going to be required to constrain the freedom of action of the CC in order that he's unable to exploit assumptions on u-boat behaviour that would not be available to him (such as knowing more or less exactly when an attack is likely to develop).

So, I think the CC and allied players need to spawn in first, with the precise spawn-in time of the "German" players unknown to them. The number of "German" players, and the number of u-boats spawning-in, should likewise be concealed from the allied ones. There should be limits on how far escorts may range from the convoy, under either AI or human control, or a mixture thereof. These limits could possibly be gradually relaxed as the game develops, but to start with, there needs to be an assumption that the u-boats would be able to close with the convoy sufficiently close to gather data, without undue interference, and that this should be reflected in the limits at or near game-start. It might also be worthwhile having players only able to take-over a Flower Class for the 1st 30 minutes of a game, with access to other escorts increasing as time goes on... This needs a lot of thought!

[Tabris]

In "Kriegsmarine: Naval Campaigns" by Wargame Design Studios (RIP John Tiller) the allied player is simply not allowed to command his units at all until either an axis unit is spotted or an allied unit is hit. It's a simple solution, but of course you've already addressed how there'd have to be restrictions in place on how the player's unit can spot things, plus it wouldn't be very fun to have nothing to do.

[Fidd]

That would be one approach, however, we know from current play, that as the detection-range is known, literally to the meter, that this allows u-boats to attack a convoy with complete impunity if the escorts do not move outwards from their base track. It follows from that that playable escorts would be a complete waste of time and effort, if the convoy commander is not able to send at least some of his escorts out wide to expand the area in which surfaced u-boats cannot operate. Conversely, because of the time/forseeable uboat actions problem laid out above, there do need to be some fairly robust protections to prevent those problems. We need BOTH some escorts occasionally moving out to catch the unwary, AND mechanisms to ensure this can't be done both sides and ahead of the convoy at launch plus 15 minutes, which would likely render most u-boats detected of forced-down every single play session!

[THE_MASK]

PVP -1
Better AI for the enemy +1

[Fidd]

Originally Posted by THE_MASK

PVP -1
Better AI for the enemy +1

I'm not advocating for one or the other, but BOTH, using a human player to directing which areas AI escorts move to, and then giving human players the ability to jump into an AI escort and take it over.

The object here is to use AI to do the majority of the boring stuff, and to use PvP for whenever finer judgment or decision-making, and of course the fun bit (asdic/DCing).

If one ONLY used PvP, it would suck, because the human players might spend 3 hours game-play without even detecting a u-boat.
Conversely, if one ONLY used AI, it would (does) suck, because the AI is too predictable, and is incapable of providing a quality opponent.

It's not a case of "either/or", it's BOTH that are needed.

[Fidd]

Further to this, I think as a general principle, wherever possible, AI decision-making, and attacks, should be conducted by human players, because they're capable of artifice, deception and intuitive guesswork - and other necessary qualities for hunting uboats in a way that AI really struggles to do. We call AI "Artificial Intelligence", but I submit it should more properly be called "AS" - Artificial Stupidity", because it is such a blunt-instrument and largely incapable of nuanced decision-making.

Example: Escort has firm pings on a uboat and is prosecuting a DC attack. Another uboat surfaces 8km away, and commences firing it's deck gun, causing the escort to cease prosecuting it's DC attack, and it then seeks to hit the surfaced uboat with gunfire!

Self-evidently therefore, AI is both as dumb as fence-post, but also, and this is borne out of experience in dozens of games, bloody difficult to program so that it behaves similarly to a half-way intelligent human! Another example is AI "gunnery". At range, we do not experience a gradual nearing of shell impacts before we're hit, as one would tend to experience for real. Instead, we have a number of highly inaccurate rounds land, and eventually a single hit is landed, whereupon the gunnery becomes laser-like in it's precision thereafter!

So AI (AS) sucks badly, and anything that can be done to reduce it's idiocy by replacing it with humans is a very good thing!. Conversely, much of the tasks required of an escort crew, are mind-numbingly dull and repetitive. Endless days or weeks of pinging, with no returns from a uboat, and the occasional false return from a hapless whale or wreck on the sea-bed. (virtually ever ww1 and ww2) ship-wreck on the sea-bed around the UK is literally hammered-flat by DC's dropped during the war to ensure that if they were a uboat, they were dead!

AI is therefore required to do the heavy-lifting of these boring tasks, so that AI escorts can still pose a problem for uboat crews, until they are taken over by allied players to do the attack/asdic search or surface gunnery - better (which is not necessarily more accurately!) than an AI would do in the same scenario.

If we consider the "get to 185m and you're safe" is got rid of (or is a lobby-setting so to allow/prevent), then a human player would know to keep a u-boat down until it has fallen well astern of the convoy, and likely would keep it down until dayight, so that it had to spend much longer overtaking the convoy even on the surface. An AI one is incapable of any such nuance. It has to be remembered of course that this has to be kept as a playable game, and few are so hardcore as to wish to play with realism wound up to full, however, if AI and human operated tasks can be intelligently applied so that the game is capable both of more varied outcome, but also increasing difficulty, then this will help retain players.

The same applies to the complexity of operating the boat. At it's most simplistic level, we have the bots. Then we have players some of whom have only been playing a day or so. Better skilled players of a few weeks usually improve over the next year or so, but often inside a year, they're able to dive/use the engines/helm with little or no further scope for advancing their skills, beyond playing in shallower water and dialling up settings such as torpedo reliability. Wouldn't it be great if even the simple-jobs could be made (electively) more difficult to vary the expertise and difficulty of performing a role? Such flexibility in the game would also help player retention.