Topic: maning military buildings

Maybe the solution is to keep the soldiers sorted in the warehouse, maybe in a 2-dimensional data structure. So, we'd have a slot for each summed up soldier level and a list of soldiers in each slot. Then, depending on whether heroes or rookies are preferred, start searching the slots from the front or back for the first non-empty solder list.

Basically, the same type of data structure that you would use for https://en.wikipedia.org/wiki/Bucket_sort

Busy indexing nil values

Top

Quote

kaputtnik wrote:

I couldn't say something to the code, but for a player i would say that if a military building is set to prefer Heroes, the code should send Heroes if Heroes are available (in a warehouse). And i think this was the behavior sometime (but i am not sure).

If you build a Castle (which prefers heroes by default) nearby the frontier to get some resistance to the enemy, it is somehow bad if it get filled with rookies first even if there are some heroes in a warehouse. Of course, currently one couldn't see how many heroes are in a warehouse...

I proposed that soldiers are sorted in warehouses depending on level, at least roughly (recruits, level 0 - roookies, level 1 to 3 - veterans, level 4 to 6 - elite, level 6 to 9 - heroes, level 10, except for atlanteans that only have 9 levels, so for themm it would go level 6 to 8 - level 9). But that would require a lot of rewriting, because as of now, all soldiers are considered as a single ware, with a string attached to show their stats.

However, I thought of something that may allow that kind of change easily: it ccould be easy enough to make separate categories of soldiers in the same way that there are miners and master miners. Both kind of miners can fit into a normal mine, so coding soldiers like that would ensure that there would be different wares named "soldiers" and they all could fit into military buildings. then all those soldiers would still have theur string of attached stats values, just like now. Simply, when one soldier accumulated enough promotions, it get automatically upgraded to veteran soldier, or however we want to call them. A newly built fortress would look for heroes first, then for elites, and so on. it may still exchange a few soldiers afterwards, because all the various categories would still be stored according to the current algorithm, so after it got all the elites, there may still be some elites left in the warehouse with better stats. but it would ensure that the fortress would get at least the highest tier of soldiers available.

No0815 wrote:

Thanks! It's actually nice to see that someone explicitly agrees that the current behavior is suboptimal, to say the least.

Oh, I also agree that the system is suboptimal, but the programmers have limited time, so it is best to spend that time on more important things. Furthermore, we must make sure to not swap a suboptimal algorithm for one equally subotpimal, only with different problems. And third, I see that virtuallly every game I've ever seen - indeed, every piece of sofrtware - is suboptimal in many aspects. It's just that in those cases, we know it's going t stay that way, so we treat the suboptimality as a feature. while in widelands we know stuff can be improved all the time, so we as of it to be improved. It's the same reason the quality of the leaders has much improved in going from monarchy to democracy, yet the citizens of democracies complain about their government much more. People tend to complain on what they can change, and to adapt to what they cannot.

Top

Quote

If we're complaining about soldiers, then we should at least consider rebalancing them. Barbarian 0 -> 2 attack, for example, is vastly cheaper than Barbarian 2 -> 5 attack, yet it also provides a bigger increase (0 -> 2 is +100% attack, while 2 -> 5 is only +75% attack). If you want, I can do the math to get better stats (I really am a math person, but it depends on some questions for the staff:

* Do you want higher levels to give the same total increase? (so if 0 -> 1 is +20% attack, should 3 -> 4 also be +20% attack?)
* Do you want higher levels to give a same increase per ware? ((so if 0 -> 1 is +20% attack and 3 -> 4 costs twice as many wares as 0 -> 1, should 3 -> 4 be +40% attack?)
* Do you want all four categories to be roughly equal? (so a 2/0/0/0 soldier and a 0/0/2/0 soldiers are equally matched)
* Should evade training also increase accuracy a bit? (I'm thinking bases of 100% for level 0, 110% for level 1, and 120% for level 2, so 0/0/2/0 vs 0/0/2/0 doesn't take forever).
* Attack levels vary a lot at lower levels (12-16 for barbarians and atlanteans, 13-15 for empire), but very little at high levels (47-51 for barbarians, 45-47 for empire, 44-48 for atlanteans). Should the higher levels have larger ranges?
* Level 0 defense is near-worthless (Barbarian level 5 attack always does 45-49 to barbarians and 44-48 to empire). Should it be changed (either 0 or somewhat respectable values)?

And, of course, it also depends on the staff wanting the numbers changed -- if they're fine with the current numbers, then we can continue to use them. ;)

I should add that the stats in my examples were taken from https://wl.widelands.org/wiki/SoldierLevels/ -- those stats are probably still current, but if they have changed, I apologize.

Top

Quote

DragonAtma wrote:

If we're complaining about soldiers, then we should at least consider rebalancing them. Barbarian 0 -> 2 attack, for example, is vastly cheaper than Barbarian 2 -> 5 attack, yet it also provides a bigger increase (0 -> 2 is +100% attack, while 2 -> 5 is only +75% attack). If you want, I can do the math to get better stats (I really am a math person, but it depends on some questions for the staff:

someone already mentioned that in the past (maybe it was you tht time too?), and i replied that while it's true, a stronger soldier can kill a weaker soldier, then go heal, therefore it's still more effective to promote soldiers all the way. there had been a general consensus to leave things as they were. on the other hand, attack promotions are much more useful than defence promotions, and looking at my calculations, i see that in a fight agins fully promoted soldiers, it would probably be better in some cases for the empire to not build armor at all, or to only build level 1 armor - takes 4 hits to kill instead of 5, 16% more chances to lose for that, but some 30% cheaper.

anyway, you should open another topic if you want to try and change those values.

I should add that the stats in my examples were taken from https://wl.widelands.org/wiki/SoldierLevels/ -- those stats are probably still current, but if they have changed, I apologize.

those stats are outdated. a few months ago we agreed to give the barbarians an extra +1.5 damage per level, since a fully promoted barbarian soldier only had a 36% chance to win against an atlantean, while being more expensive; barbarians pay their stronger early game with a weaker late game, but they should not become completey useless. They already have the most expensive soldiers to train and the most expensive mines to operate. So, someone should update that page.

Top

Quote

No0815 wrote:

Well, rating importance can sometimes be a delicate issue. But if a problem can affect the game in such a deep and potentially game-changing way, as the example of kaputtnik should illustrate, I would hardly consider it as unimportant.

Furthermore, we must make sure to not swap a suboptimal algorithm for one equally subotpimal, only with different problems.

Now that's definitely fair enough, as long as it doesn't implicitly means "Nah, changing would require effort, do not want." Though I can hardly imagine that this change could have unwanted side effects, unless the changed code itself is defective. Though I still could be wrong, I don't know the source code, so if someone has substantiated arguments ...

Actually, changing that algorithm has rather profund game-changing effects, which will affect balance, so we must be very careful to consider them. For see, you consider the fact that strong soldiers will take time to man a new building at the border to be a flaw, and it certainly is such when your soldiers are doing it. But what when you are on the defensive?

In my first game in the torunament, I was the only one who knew the micromanaging to create level 10 soldiers without a strong economy. I had only a handful of them, but they killed weaker soldiers and then retreated to get healed. I could slaughter the enemy army without taking a single casualty. But human opponents can adapt. They dismantled their border buildings, retreating. Now I had to build forward military sites to get to them. and when i did, those sites got filled with weak soldiers, and got overrun by the enemy. I found myself unable to advance. Eventually I found a way, but I needed one or two hours and several more level 10 soldiers. By the time I managed to break through, my opponents also had level 10 of their own, though not enough to change the outcome.

You see, if the algorithm had worked as you wish, that game would have ended immediately, with no contest whatsoever. Do one deserves to win only because he can micromanage a couple of level 10 soldiers in the first two hours of the game? It turned out, what you considered simply a bug or an unoptimized algorithm actually turned out to be an important factor in attack/defence balance.

So, do we want to give that advantage to the attacker? Letting the stronger soldiers occupy a new building will remove one of the very few viable defensive strategies one has when facing down stronger soldiers. I'm not sure it is wise to alter that mechanic. In fact, the only reason I'm not vehemently opposed to changing it is that one can, again by careful micromanaging, ensure that his new building will be filled with strong soldiers from the start.

(If you are curious: step 1: fill anoter building away from the front with good soldiers. Step 2: empty it, so that all the good soldiers will enter the warehouse and be placed consecutively in the array of soldiers. Make sure to count the soldiers in the warehouse first. Step 3: finish your border building, or connect a road to it if you completed it but cut road to avoid it being manned. Step 4: now the weaker soldiers will leave the warehouse. Let them come out, then cut the road. those soldiers will go back to the warehouse, but will be in the end of the array now. If you had X soldiers in the warehouse before you sent your best soldiers in, then you need to send out X soldiers. Every time weak soldiers try to go into the building, cut the road. Eventually, the good ones will come. You may need to cut roads to your other military buildings, if soldiers keep going in or out of the warehouse they may mess with that)

Widelands is a game of unstable equilibrium. The defender has no bonus for fighting in his land, so maybe we should leave every possible little advantage aa defender can get, unless we want all the games to be decided in the first couple of hours. As I write this, I am less convinced that we should change the mechanism.

People tend to complain on what they can change, and to adapt to what they cannot.

That's a pretty feeble argument.

That wasn't an argument against changing stuff. That was a bit of philosophical though on being more appreciative for something can be improved, instead of scoffing it for its failings; and they often are misled at perceiving something good as something bad because of that, so I preach about it when I can. Yeah, of course it would be a dumb argument for not fixing problems.

Top

Quote