Victoria 3 - Dev Diary #68 - Patch 1.1 Changelog (part 4)

[kolpo]

The change to treaty ports will make the ''Western protectorate" achievement extremely hard, not only do you need a treaty port in those major European powers but you will also need to reduce them below great power and keep them at that rank.

 

[Soranya]

They're not going to change Monday's patch, it's locked in days (if not weeks) in advance. If there's no undocumented fix about buggy general behavior around fronts, earliest we'd see it is 1.1.1. But yes, hopefully it's on their radar, though there are multiple ways to interpret the lack of comments about the issue (they don't know if it was fixed, the fix is not in, etc.).

More rollign releases (CD) would be helpful for the development cycle too.
So many things where the Softwaredevelopment of Paradox seems to lack behind industry standards.

 

[Tempscire]

Setting b=1 does no harm, and we got your conditions. as you did implicitely we can also set p = P/b
0.25 = f*0.5
0.75 = 0.5
and get as a piecewise linear solution
p(y=a,x) = 3/4(x/a) + 1/4 for S < B
p(y=a,x) = 3/2(x/a) - 1/2 for S > B

you could do it the other way around (in fact expanding it probably is easiest to do it like that), but as you said, these functions wouldnt be affine to the one I wrote out. however I am not sure or couldnt follow what would technically be the argument against using those functions.

Yes, maybe my explanations are sometimes a bit fast. But we arrive at the same function for fixed S, as 1+1,5(B-S)/S=3/2(B/S)-1/2 for S>B and 1+0,75(B-S)/S=3/4(B/S)+1/4 for S<B. This hopefully convinces you that there is indeed a unique two piece affine linear solution. In your notation we can do the same for a fixed B as you said. This will be important to understand the problem with the approach. We get:

p(y,x=a)=-3/4(y/a) + 7/4 for S>B
p(y,x=a)=-3/2(y/a) + 5/2 for S<B.

Now the problem with using these functions is the following. They can not be combined into a well-defined function in two variables. To see this consider for example the point S=2, B=3. What is p(2,3)? On one hand you could argue that this should be p(S=2,3), i.e. the function with fixed S. This is
p(S=2,3)=3/4(3/2) + 1/4=11/8.

But on the other hand this should be p(2,B=3), i.e. the function with fixed B. This is
p(2,B=3)=-3/2(2/3) + 5/2=3/2=12/8, which is not equal to 11/8. Therefore, you have to decide whether to use the function in B or in S in order to calculate the values p(S,B) for S not equal to B. The only places where they coincide are the values on the diagonal, i.e. S=B and the other points defined by the conditions, i.e. the lines S=2B and 2S=B.

So every time you want to calculate the price in a market you have to decide which function to use in order to do so and you only stay (affine) pl if you use only one of these functions. If in a market only buy or sell orders change, the choice might still be clear. But if both change, you run into serious problems. And even neglecting trade the markets should usually have both buy and sell orders.

You can think about the currently implemented price formula (up to the cutting at the maximal price range) in the following way. The devs have made a choice and they said p(S,B)=p(S=S,B), if B>S and p(S,B)=p(S,B=B) if B<S. Therefore p is affine linear in B, if B>S and affine linear in S, if B<S. But in the other variable it is no longer affine linear in both cases. You can of course make a different choice, but you have choose at some point. And none of the choices is affine pl in both variables all the time, as the affine pl solutions for one fixed variable are unique and they do not agree, as explained above.

Possible options are to say p(S,B)=p(S=S,B) in all cases. Then it is affine pl in B. But in a market importing goods the price will no longer be an affine pl function, as S changes. The same holds for p(S,B)=p(S,B=B) in all cases.

A different approach would be p(S,B)=p(S,B=B), if B>S and p(S,B)=p(S=S,B) if B<S. I.e. replacing the min(B,S) in the current formula by a max(B,S). This would solve some issues, as for example a trade route from markets A to B, with price p_A in A lower than base price and price P_B in B higher than base price, would increase the buy orders in A and the sell orders in B. At first this is affine linear, because as long as p_A is lower than base price we have Buy_A<Sell_A and therefore the price function is affine linear in Buy_A. Similarly for the market B. But still problems occur when p_A and p_B are both higher (or lower) than base price. Lets assume they are higher, then the price in B is still affine linear in the sell orders. But the price in A is no longer affine linear in the buy orders, as now Buy_A>Sell_A. Therefore, the price difference is an affine linear function minus a non-linear function and this will be non-linear.

I hope this makes it a bit more understandable and at least to some degree explains why there was the impression that the price function was affine pl. It is but only for one of the variables S and B at a time. As it is actually impossible to have a function fulfilling the desired properties which is affine pl in both variables at the same time.

The only functions which are affine pl in both S and B are of the form, x+yB+zS and in order to at least have p=1, if B=S for all B, we need y=-z and x=1.
I.e. the price function looks like p(S,B)=1+c(B-S) where c is some constant. Therefore, it only depends on the difference between B and S. But such a function was probably ruled out by the devs, as it has the property that p(101,1)=p(1000100,1000000), i,e, the price in a market with 1 buy order and 101 sell orders is the same as the price in a market with 1000000 buy orders and 1000100 sell orders. The price can only depend on the difference and not on the ratio between buy and sell orders for such a price function.

 

[8balls]

So right now I'm trying and trying to successfully naval invade dai nam as france without using the general teleportation exploit to figure out how to play in the new patch. It's completely impossible to do. Whenever a general lands it isntantly opens 3 fronts, 2 of which are undefended. No matter how much better my army is the undefended front causes it to lose the territory and withdraw without putting up a fight. There is no way to reinforce the empty front fast enough to actually invade anything

 

[aekkea]

Regarding the second paragraph. If the function was really linear then averaging post-and pre-trade prices would give you exactly the area underneath the price difference function. I.e. the profit function as in Profit = (PD_PostTrade+PD_PreTrade)/2*Volume would be the antiderivative of the price difference function. This implies via the fundamental theorem of calculus that the price difference function is the derivative of the profit function. In particular the profit function can only be maximal, if the price difference is equal to zero. Therefore, you would not have a price inversion problem. Granted you do not differentiate continuously, but in a discrete way and currently do not search exactly for the optimum. Therefore, the volume would probably slightly overshoot the zero price difference mark. But this is not related to the function being strictly convex or linear.

I've set up a game and a desmos project to clear up this confusion. I emulated a luxury furniture (base price B = 60) trade route France -> Russia (w/ free trade, enough convoys, no other route for luxury furniture in both markets). In my game it converged at level 10, with 480 volume (orange line on the plot).

The formulas I used to calculate prices (they turned out to be accurate):

MaxPriceVariability = BasePrice * PRICE_RANGE
MaximumDifferenceFactor = BUY_SELL_DIFF_AT_MAX_FACTOR - 1
BuySellDifferenceRatio = Min( ( Max( Buy, Sell ) / Min( Buy, Sell ) ) - 1, MaximumDifferenceFactor )
PriceAdjustment = MaxPriceVariability * ( BuySellDifferenceRatio / MaximumDifferenceFactor )
FinalPrice = ( Buy > Sell ) ? BasePrice + PriceAdjustment : BasePrice - PriceAdjustment

Assumptions: PRICE_RANGE = 0.75, BUY_SELL_DIFF_AT_MAX_FACTOR = 2.
On the plot, x axis is trade route volume, y axis is money (price or profit in 000's). Lowercase p represents market price. r is BuySellDifferenceRatio. P_f is trade center profit in 1.1, P_m is profit in current formula (avg of pre-trade and post-trade), and purple dots (P_t) are the area under price difference function (the "true" profit I assume). (Have to use dots because integrals take a long time to compute).

A few observations:

1. Market price functions and price difference function are not piecewise-linear. Price difference function can have more than two "pieces" due to price bounds.
2. P_m is closer to true profit than P_f in the area where the trade route I'd expect to end up (where the marginal gain is positive but not enough to employ extra workers). However, using true profit, the route will probably still overshoot
3. Trade route profit under the new formula will start decreasing way before reaching equilibrium (keep in mind this does not take into account increasing trade center wages)

I guess the root of the problem is the way trade routes expand - way beyond maximum trade route profit. This would not be solved by using "true profit" since it's pretty close to the current formula. The new formula seriously underestimates trade route profits, but will make routes converge more realistically, i.e. before price inverse point

Can't link the project so attaching a screenshot

 

[shoebird]

That's why I simply asked "Is this just a temporary solution only for now, or this is how it will stay forever", because I was wondering about intended design of it. If the your quoted part means that it is indeed a temporary solution than I have my answer, thank you

I wouldn't get too exited. I don't think they see too much problem with Gibraltar. It seems to me they haven't cared about this until someone brought the Pondicherry issue, despite the Gibraltar one had been mentioned several times before. The (temporary?) fix doesn't seem to affect Gibraltar either as Spain is not great power at game start. If they cared about Gibraltar they would have made treaty ports to affect only unrecognized countries.

But let’s see what happens.

 

[doubleskulls]

So right now I'm trying and trying to successfully naval invade dai nam as france without using the general teleportation exploit to figure out how to play in the new patch. It's completely impossible to do. Whenever a general lands it isntantly opens 3 fronts, 2 of which are undefended. No matter how much better my army is the undefended front causes it to lose the territory and withdraw without putting up a fight. There is no way to reinforce the empty front fast enough to actually invade anything

I entirely agree. First of all I quite like the concept behind the front system for warfare, but the implementation just isn't good enough, and the exploit was just about the only thing in my opinion that made it tolerable. I'd have no problem with the instant travel exploit being removed if they sort out the allocation of generals to fronts at the same time.

Even when you've got lots of generals assigned to fronts I can't work out what logic is applied to redistribute them when the front splits, or when fronts complete the logic for reallocating seems weird. Alaska is a good example of a state with these problems. Invade it, see multiple fronts form. Even if you assign one general to each, when they complete their front they don't automatically re-assign to an unmanned front, but I think just to the closest one. So you can have 4 or 5 fronts open, and all your generals attacking the same one.

Whole wars can end up being decided because the general gets allocated to attack/defend the wrong front, leading to the loss of vital territory. If the other issues with general allocation aren't fixed I'm probably going to just play something else until 1.2 comes out. Or find a mod with 0 travel time.

 

[Sarmatian]

Has late game slow down been addressed at least to a degree?

 

[doubleskulls]

Has late game slow down been addressed at least to a degree?

Yes, 1.06 made it a LOT better.

 

[RELee]

But you no longer get to play it. That time has passed for you.

 

[Sovetskysoyuz]

I entirely agree. First of all I quite like the concept behind the front system for warfare, but the implementation just isn't good enough, and the exploit was just about the only thing in my opinion that made it tolerable. I'd have no problem with the instant travel exploit being removed if they sort out the allocation of generals to fronts at the same time.

Even when you've got lots of generals assigned to fronts I can't work out what logic is applied to redistribute them when the front splits, or when fronts complete the logic for reallocating seems weird. Alaska is a good example of a state with these problems. Invade it, see multiple fronts form. Even if you assign one general to each, when they complete their front they don't automatically re-assign to an unmanned front, but I think just to the closest one. So you can have 4 or 5 fronts open, and all your generals attacking the same one.

Whole wars can end up being decided because the general gets allocated to attack/defend the wrong front, leading to the loss of vital territory. If the other issues with general allocation aren't fixed I'm probably going to just play something else until 1.2 comes out. Or find a mod with 0 travel time.

At least the combat changes in 1.1 will allegedly make the individual battles shorter, so hopefully generals won't get stunlocked for a whole month because three battalions are fighting.

 

[ifg838]

They said in the notes that they fixed that. Did you read the notes?

No, I didn’t read it, it was easier for me to ask

 

[Catoscar]

Are there any current ideas for changes to supply chain logistics in buildings based on distance from their inputs? It feels really weird to have something like iron mined in Australia to be just as efficient for a factory in Britain to use rather than iron mined either in the same province or nearby.

 

[MrRazza]

Hardly game breaking but I do think it really stands out as an error - is the "North Sea" strategic region going to be corrected to "North Atlantic"? It has almost nothing to do with the actual North Sea (circled in purple). I flagged it on the map feedback thread but as far as I can see there's been no response to anything posted there. I know there's bigger adjustments to be made that will be attracting attention right now but it is only a name change for an error in very big letters!

 

[Shevvek]

Are there any current ideas for changes to supply chain logistics in buildings based on distance from their inputs? It feels really weird to have something like iron mined in Australia to be just as efficient for a factory in Britain to use rather than iron mined either in the same province or nearby.

Especially so for electricity.

 

[veryalien]

Are there any current ideas for changes to supply chain logistics in buildings based on distance from their inputs? It feels really weird to have something like iron mined in Australia to be just as efficient for a factory in Britain to use rather than iron mined either in the same province or nearby.

But it's not the same. You need convoys to ship it and that's an additional cost. Also the level of the trade route increases and decreases over time so its not like having it next door. There's only so much you can ship at any given time.

 

[Troyen]

More rollign releases (CD) would be helpful for the development cycle too.
So many things where the Softwaredevelopment of Paradox seems to lack behind industry standards.

I think they're getting there. Stellaris has hit a fairly consistent 3 month release cadence the past year and a half. CK3 adopted one for 2022 and indicated plans to keep it going forward. It still means bugfixes can take a while and I wish they'd budget more post-release bugfix patches but overall a massive improvement over the random 6-9 month silence some Paradox games get (or used to get). Hoping Vicky 3 also does a quarterly release once the 1.1/1.2 post-release rush dies down.

 

[HelloFolks]

I'm happy with the patch but for next one you should prioritize geography rework. There are a lot of mistakes Iin the map. Lots of region to be split or other not so wide. Some region traits are wrong.
Too many regions without a trait.
There is the map thread in bug forum. check it for the next patch.

More patches you wait to update it more difficult will be

 

[MachopPower69]

Will the 1.1 fix the crashing when clicking on the government tab? I made a thread about it in the tech support which a kind moderator is helping me with that is getting nowhere

 

[TempestM]

But it's not the same. You need convoys to ship it and that's an additional cost. Also the level of the trade route increases and decreases over time so its not like having it next door. There's only so much you can ship at any given time.

If they are in the same market you don't need trade routes. Britain and Australia are.