The Spartan Academy

[vmxa1]

Now that no war is in sight on any front that we will be involved in, the value of holding those knights is not high. You pay 1 gpt for each one. When and if you get MT and SP, you will have to spend money to upgrade them. If you add all that up, is it a good investment? When you have to go on attack overseas, we will need transports for any troops. Will you want to load calvs at that time instead of something else?

[asleepathewheel]

the problem is that we might be without either rubber or oil, thus no tanks. It depends on what role we take (if any) in the gwnd alliance. it might be much better to just build 30 marines and a pack of ships for protection, let the larger gwnd forces tackle the interior, we just siphon units on the perimeter.

sucks the turn order, rp's spying will be woefully innaccurate.

[Cort Haus]

Yeah, our spying would be the best for turn-order. Gives us an excuse if we wanted to hold off on a DoW on Lego.

One scenario we might consider Cav for is if ND attacks GoW within, say 30 turns. We'd want Lego to jump in there and fight ND, but we'd could use the chance to bid for Bob.

Currently the Knights are plugging Eliopolis. If we decided that

1. Cav wouldn't play a role in marine-era defence or assult
2. We're not going back to Bob - even if there's a war there.
3. We don't want to defend Eliopolis

... then we could spoon a possible 13gpt on the economy, plus 2 marketplaces worth of shields. At a high tech rate that could pull it up to 20+ gpt, and bring forward 2 coastal Universities. It's tempting, if we are to commit ourselves to going builder-tastic for 30 turns before re-arming from scratch with modern kit.

[Theseus]

I'd rather have a slew of Arty than just about anything in a worst case scenario.

Does that perhaps mesh with CH's builder-tastic idea, if we slip Cannons in amongst the buildings, timed to sufficiency as of the Marine menace?

Build a Cannon, disband a Cav...

[Cort Haus]

Cavalry have another trick - pecking at passing enemies.
How does a Cav's ZOC mix with a Fort's ZOC - can they work together?

Did you mean 'build a cat, disband a knight?' So melt down all the surplus now, and replace the rest of the knights gradually with kittens who will grow up to be tigers, after RP (Rep Parts not R.Podos )

[asleepathewheel]

gads! No one would ever consider invading us if we were defended by an army of nude RPodians! I like it... as long as I don't have to look at another save.

[DeepO]

For the moment, we don't need extra units. And while I agree with getting another 13 gpt, I wouldn't throw our extra knights away just yet. As mentioned before, in case we don't have oil or rubber, and we somehow acquire SP, we might need cavs.

So, let's hold the knights for the time being, and disband some WCs... we don't have the cash to upgrade these to knights or cavs anyway.

ZOC: I thought every unit could only be hit once per tile it crosses, and every unit (and fort) could only try to hit an anamey unit once. They don't add up, I thought, but perhaps it might increase the odds, I don't know.

DeepO

[vulture]

I've had to do some code re-writing to put the effects of artillery into the stack combat calculator. On the attack, I've assumed that artillery pits its attack strength against the defenders modified defense (the value it would normally use in combat), and fighting a number of rounds equal to the rate of fire. So the actual 'artillery' unit with ROF of 2 fight two rounds of combat with its strength of 12 (vs e.g. an infantry with defense 11.0 unfortified on plains). If there is no lethal bombard for this combat type, results once the defender is on 1 hp are ignored.

On defense, and bombadr unit gets a single shot (ignoring ROF) at those same odds - bombard strength vs attackers defense on the terrain they are attacking from. I'm not convinced this is correct. The first part seems right - I don't recall seeing an attackers losing 2 hp to defensive bombard when attacking a unit stacked with artillery - but the second I'm suspicious about: catapults seem to be far more effective on defensive duties than this calculation would imply. Maybe they use the unmodified unit defense, or something...

EDIT: the code doesn't use bombard optimally (which would take far too much work IMHO). e.g. if all defenders are at 1 hp, bombard units will still be used, when it would be better to keep them spare in case a defender wins and gets a promotion - and can be bombarded back to 1 hp again. I suspect the effect is relatively small except in special cases though.

[vulture]

So, I've done two sets of tests to try and see what patterns emerge, and to test the code.

Firstly, I pitted two test stacks against various numbers of infantry. Stack A was 17 cavalry. Stack B was 8 guerillas (or infantry, same cost) and 8 artillery. Both stack have the same shield cost (1360 shields). Attacks are against 4, 5 or 6 infantry (a pretty narrow range, but enough to show what is going on I think).

no. of             AAA                     BBB            loss ratios
infantry   att dead   def dead     att dead   def dead      A     B
  4         3.68       3.96         2.44       3.75        1.21  1.54
  5         4.52       4.74         3.60       3.91        1.18  1.09
  6         5.22       5.11         4.42       3.52        1.10  0.80

'Loss ratios' are the ratio of shields lost by the attackers and defenders in each case, although the losses in numbers of units aren't significantly different. One thing to notice is that the artillery / infantry (guerilla) combo always kills less than the cavalry force, but does so at less loss to our forces. It's much like having retreat - you give up some of your victories in return for giving up a far greater number of your losses. So against a weak invasion stack, arty and slow movers are better than pure cavalry, since they conserve our forces more efficiently. But with our two sdtacks, you can see that around the 5/6 infantry mark the cavalry still have a decent chance of only leaving one guy alive (or no-one), while the mixed stack is running a pretty good chance of leaving 2 or 3 units alive. As the invasion force gets bigger, the problem gets worse.

IMHO the value of cavalry being able to deal with bigger stacks more than offsets the slightly higher losses incurred in dealing with small stack (which are crushed regardless - the question just being how many losses we take in the process).

The second test was to look at changing the ratio of artillery to guerillas in stack B, against 4 infantry. This yielded no great surprises. With a pure guerilla stack, average enemy dead were 3.95 vs 6.62 of ours. with a 50/50 mix those bcome 3.75 enemy dead vs 2.44 of our - our losses cut by a factor of nearly 3, with very little change to enemy losses. With 12 artillery and 4 guerillas, enemy losses are 2.99, ours 1.01. That's the best 'exchange rate' we get in this scenario, although we would be running a very high 'risk' of 1 or 2 (or even 3) 1 hp infantries hanging around. The ideal artillery / guerilla ratio depends on how much emphasis you put on destroying the enemy right now vs the epmhasis put on conserving strength. For me, the 50/50 mix is reasonable as an all-purpose force. With cavalry instead of guerillas, I'd guess a slighly lower fraction of artillery, since the retreat of cavalry is more or less equivalent in function to bombardment: perhaps 30:70 force mix, although I'd have to run some tests to get a better idea. I'll try and do those tests shortly and put up the numbers for everyone.

[Cort Haus]

Superb work, Vulture

[DeepO]

nice work, Vulture... and it shows your point well. However, this mainly shows the fastmovers vs slowmovers+artillery side of the story... I'm still expecting (or hoping at least) that we have tanks instead of cavs for fastmovers

DeepO

[vulture]

Second round of test done - this time with 10 attackers as a mix of cavalry and artillery, vs varying numbers of defenders (2-10 infantry, unfortified, open ground). In the tables, no. of cavalry is given; no. of artillery is 10 - no. of cav.

Firstly, the average number of friendly / enemy dead in the combat:

                          Infantry
C           2              3              4              6              8              10
a 2   0.369 / 1.603  0.480 / 1.320  0.610 / 0.960  0.699 / 0.623  0.745 / 0.431  0.762 / 0.361
v 4   0.555 / 1.905  1.034 / 2.157  1.275 / 1.699  1.432 / 1.087  1.502 / 0.785  1.516 / 0.715
a 6   0.871 / 1.945  1.576 / 2.446  1.956 / 2.218  2.189 / 1.459  2.260 / 1.142  2.275 / 1.076
l 8   1.336 / 1.961  2.121 / 2.598  2.612 / 2.579  2.940 / 1.841  3.018 / 1.499  3.033 / 1.432
r 10  1.835 / 1.972  2.654 / 2.706  3.237 / 2.878  3.681 / 2.234  3.775 / 1.857  3.785 / 1.792
y

This on it's own isn't much use - it is essentially what we saw before. For a given enemy stack, increasing the fraction of bombard units in your force decreases enemy casualties, but decreases friendly casualties to a greater degree.

Another not-very-illuminating table is the ratio of losses : enemy losses divided by friendly losses:

     2     3     4     6     8    10
2  4.34  2.75  1.57  0.89  0.58  0.47
4  3.43  2.09  1.33  0.76  0.58  0.47
6  2.23  1.55  1.13  0.67  0.51  0.47
8  1.47  1.22  0.99  0.63  0.50  0.47
10 1.07  1.02  0.89  0.61  0.49  0.47

This also tells us the fairly obvious: as we put more artillery in the force (at the expense of cavalry) we do better and better at inflicting more casualties than we receive - although this may be at the expense of inflicting enough casualties to hurt the enemy at all. Note the last column - all values are the same for the simple reason that every cavalry that attacks has to face a completely health opponent, so all fights are at the same odds, and give the same loss ratio.

In reality (well, in the game) we want to achieve both conservation of our own forces, and destruction of the enemy, and find some balance between the two. We don't want to get a great exchange rate, but leave most of the enemy alive. Nor do we want the enemy body count as high as possible if that means throwing away three times as many troops as he loses. One very crude measure of 'combat effectiveness' can be found from multiplying the loss ratio by the total number of enemy killed (which is in effect enemy_dead^2 / friendly_dead). THIS IS A VERY CRUDE MEASURE. It takes no account of the health of the surviving defenders, the other uses of units (especially artillery), ability to survive a counterattack, and it may well be that e.g. enemy_dead^1.93 / friendly_dead^0.632 is a better measure to use - this is just a hurried attempt to see what falls out when we crunch some numbers.

     2     3     4     6     8    10
2  6.96  3.63  1.51  0.56  0.25  0.17
4  6.54  4.50  2.26  0.83  0.46  0.34
6  4.34  3.80  2.52  0.97  0.71  0.51
8  2.88  3.18  2.54  1.15  0.74  0.68
10 2.12  2.75  2.56  1.36  0.91  0.85

How much this means remains to be seen. Taken at face value, it tells us the following. If we are much stronger than the enemy force, we basically need just enough units to kill N 1 hp defenders (which is slightly more than N attackers), and the rest of the force should be artillery. When the enemy is about 1/3 of our numerical strength or more (using infantry, cavalry, artillery only), there is some optimal balance between cavalry and artillery that gives the best results. Any invading force stronger than that, and we're best off with an entirely cavalry force; otherwise we give up too many kills for very little return in saving ourselves.

The point is, in this test it appears that the window where a mix of cavalry and artillery is useful is very narrow. A weak force will be seen off by anything. Even if a force does fall in the narrow window, it is still weak enough to be seen off regardless; the question being at what cost to ourselves. If the force is a genuinely significant invasion force, we need every real attacking unit we can get, and the artillery are more of a hindrance than a help (ignoring their value in defensive bombard).

So I'm going to be very controversial here (and annoy Theseus no end) and say that since we only have resources and time to build a fixed number of units, we should stay away from building any more bombard units, since our enemies aren't now going to invade us unless they are landing a significant force - one that is far out of the window of artillery usefulness.

(Actually, I'm playing devil's advocate to some extent here - the value of artillery isn't going to come out here necessarily. As can be seen from the last two tables, as we put more emphasis on survivability of our units vs kiling the enemy, the window of usefulness widens to the point where artillery is almost the only useful thing to build. So the debate is really about how much value we put on these factors (survivability vs killability) and what kind of size of invasion force we are likely to be up against, and how it stacks up against our ability to manufacture defensive units).

[DeepO]

very interesting analysis, vulture!

Something I didn't know... and of course, artillery has more use then purely offensive (one thing not explored here is that it is one of the ways of damaging ships, something cavs can't do), but still...

As always in civ, it boils down to number of troops. The only real way to wage a war is with far superior numbers (of shields, not of units), even if that happens locally. As such, I think we'll want to have bombard units with us in any invasion. Of course, ships and airplanes can also help. In invasions, the survivability ascends the killability, as you need to ship reinforcements from elsewhere. In defense against invasions, we might think otherwise, and want to deal with invaders fast at all costs.

Other aspects of combined arms forces are of course also the defense of stacks. You can't simply put 10 cavs in the open field within reach of the enemy, and expect them to be there next time you look. You can put 10 guerilla + 10 arts in enemy territory. So we'll always need slowmovers, which means we'll always need arts as well.

But that's beside the point. Nobody is suggesting to go invading with an all cav-army, but you've shown they can have their uses. So we better don't throw them away just yet.

I still think that we would do well to disband 13 of our WCs (maybe keep the elites, but even their use is limited). That's 13 gpt... which we later could use to upgrade some other units. We're short of cash...

DeepO

[vmxa1]

I don't see much difference in a vet WC and an Elite one. Neither will kill anything reliablely, even if at 1HP. IOW I doubt you can get a Leader form an elite WC during a marine invasion.

[DeepO]

Just out of principle, let's disband the vets before disbanding the elites... you never know.

There's one side of Civ warfare that vulture's analysis doesn't do justice, and that is if you have inferiour units, you just need a multiple of what the other guy has. In that sense, all our obsolete WCs will come in handy if we ever find ourselves on the receiving end of a marine invasion.

But, as already been said above by others, there is only one good way to counter an invasion, and that is to destroy any ship which comes too near for our tastes. We need a navy more then we need infantries in our cities. And bomber support. The main tactic I tried in scenarios like the Alamo (MT3 or 4, before the AU games took off) was to bombard the support vessels with bombers until the transports showed, destroying these first with ships. This way, you don't 'waste' destroyers on taking out destroyers, if all you really want are the troops on the transports. Those enemy destroyers can be handled after the transports are sunk.

So, let's disband some WCs or warriors for all I care to get as much gpt as possible, and build our coastal cities so they can produce our navy when the techs become available. We will need a lot of ships in a very small time window if we want to be safe.

DeepO