Research - Formulas and Changes Explained

[crispy131313]

ORIGINAL: Hubert Cater

There is one research bug as well that Crispy discovered that will be corrected for the next build. The good news is that this is a very old bug, likely going back quite a few years, and it is a rarer bug dependent on a few rarer combinations and doesn't affect the majority of research calculations.

If the bug is as I understand, then wouldn't it have had a larger effect on research though? If higher S&I cancelled not only the bonus, but also reduced the ceiling on regular research progression of the nations that got behind it would have a rather snowball effect, where the advantaged side was getting a bonus, while the other side a penalty as well.

[crispy131313]

The bug would also explain how Germany sometimes starts pulling away very quickly in the tech race. I think it was only really noticeable because I had set some research progression to 3% in my mod, so when i allowed a country to stay at S&I 0 and rely on a co-op ally which was S&I 3 I was capped at a 0% increase per turn.

[Hubert Cater]

If the bug is as I understand, then wouldn't it have had a larger effect on research though? If higher S&I cancelled not only the bonus, but also reduced the ceiling on regular research progression of the nations that got behind it would have a rather snowball effect, where the advantaged side was getting a bonus, while the other side a penalty as well.

The honest answer here is I'm not even 100% sure what the bug was. Reason being was when I had come around to testing your particular example, some of the code had already changed on my end as I was reorganizing things to have a better understanding myself, as well to clean it up and clarify the steps as there are a lot of moving pieces when it comes to the final calculations.

Even still I had run a number of tests (based on a variety of combinations) and the current v1.16.02 EXE was matching the results in all of my development build tests which confirmed consistency as I was going forward with the code cleaning.

When I came to your one example, at first I couldn't even repeat it, and was only able to repeat it once I made the US fully mobilized. But then when I ran it against my development build, it was working fine. So what changed in between is as I mentioned above here not quite clear, but it is now working as expected on my end.

What I can say is that I don't think it is related to what you are thinking it was as S&I and all the catch up bonuses do not cancel out the regular advancement rate (as these are added in and all final values for these should be above 0 as I'm using NATURAL_8 to store the data for these), there might have just been an oddity where I fudged something in another way and I think the 3 in S&I and the 3 in the progression are just coincidental in the sense you would believe one cancels the other, i.e. 3-3=0 etc.

As mentioned all other tests were fine and all other combinations seemed to be fine as well.

Edit: I just tried running another example with the progression set to 4% instead and it did advance so maybe it was related. Either way it is corrected going forward now.

[crispy131313]

Ok well whatever it was if it's fixed then it doesn't matter thank you!

[Hubert Cater]

My pleasure and it was an odd one for sure, as a 3% progression with all the details as you provided would provide an advancement range of [2,5], which didn't work, but when I bumped it to a progression of 4%, the advancement range moves to [2,6] and it worked.

Either way, as mentioned it works fine now so all is good for the next build [:)]

[Hubert Cater]

We are also considering displaying the advancement formula for each Major and for each research category where a research chit is applied.

It would be a mouse tool tip that displays when hovering over the chit investment area (highlighted in red) and would display something like the following:

[4,11] >> [3,8] +1 +1 +0

This would indicate the current advancement calculated [4,11], with the base value prior to catch up and intelligence bonuses applied, i.e. the [3,8]. The first + would be for friendly catch up, the second + for enemy catch up, and the third + for Spying and Intelligence bonus if applicable.

The benefit of this is it would be clear what the expected advancement results should be (also to confirm proper calculations), however the drawback is that it will reveal some extra information under FoW even though it is not clear which nation has the higher values for research level and for Spying and Intelligence.

But, overall both sides tended to know who had the better materiel, so it's not totally unrealistic and could outweigh the cons of knowing a bit more under FoW if everyone thought this would be useful?

[Taxman66]

I have mild mixed feelings about the suggestion, but am fine either way.

[crispy131313]

I think this information is best left under FOW, for numerous reasons such as spoiling a planned Sealion.

[Etzel]

I completely agree with crispy. But why not linking the tooltip with the FOW setting in options? This way you can do tests with FOW disabled, but in a PBEM game you still have the fun of not knowing exactly what your opponent is doing/planning. The FOW feature is a big one. A a computer game can offer it, as opposed to a board game, where you have limited options only. I wouldn't spoil it.

[Taxman66]

Seems to me if you are getting a bonus, you should know why you're getting that bonus.
If you're ahead on S&I those services should be able to tell you that they are outperforming their hostile counterparts.
For Enemy Catch Up bonus, those that aren't obvious as seen on the field... well the scientists/researchers should be able to explain why their work is progressing faster.

But I see the argument against as well.

[PvtBenjamin]

ORIGINAL: crispy131313

I think this information is best left under FOW, for numerous reasons such as spoiling a planned Sealion.

Agreed (marginally)

Although it would clear up the research puzzle for many

[crispy131313]

Research information shouldn't be perfect. Wasn't it not until the invasion of the Soviet Union that the Germans were surprised by the Armour of the Soviet's newer Tanks? My point is there should be some element of surprise, in a game which is otherwise numbers.

[Hubert Cater]

Thanks guys, this is why we asked and the feedback is helpful. For now we'll put in for games where Fog of War is disabled, and at least this way it can be a guide and ensure that research is unfolding as intended.

[PvtBenjamin]

I see your point crispy and agree but think its somewhat on the marginal side.

Most opponent upgrades are apparent, in the case of Sealion I always go on the basis that its going to happen.


Could help tip the scale on whether Sealion is a bluff or real which is a determinant on Egyptian forces. That is a big deal.

Maybe it could be a variable like soft builds etc.

[Taxman66]

Hubert,

Preuming no bonus and the base range of 3 - 8, is the mean really 5.5 (with equal chances of each value appearing? Or is the formula/random number generation/rounding weighted to 5?

If weighted to 5, how does this work when bonuses increase the high end of the range more than they increase the low end of the range?

[PvtBenjamin]

ORIGINAL: Taxman66

Hubert,

Preuming no bonus and the base range of 3 - 8, is the mean really 5.5 (with equal chances of each value appearing? Or is the formula/random number generation/rounding weighted to 5?

If weighted to 5, how does this work when bonuses increase the high end of the range more than they increase the low end of the range?

Glad you re visited this


If its equal weighted its 5 1/2.

If the mean is 5 its a log normal distribution with actually 3,4 occurring more than 6,7,8.

Its also possible 3,4,5,6,7 have the same probability and 8 an extremely small probability. This would be if there were fractions used in the RNG (random number generator) and the range is actually {2 1/2 to 7 1/2} then rounded. For example RNG returns 2.5000 - 3.4999 rounds to 3 but only 7.5000 - 7.5000 returns 8. Thousands of iterations would return a mean very close to 5.

Love to get in the weeds!

[Taxman66]

I suspect that last option (fractions in the rng) is the most likely. It fits with the formula

[Taxman66]

With a +1 bonus then:
(5+1)/2.rounded
(5+1) + (5+1)/2.rounded

Is a 3-9 range where 3 and 9 occur less often (gen number 3.00 - 3.49; 8.50 - 9.00 and other numbers more often e.g. 4: 3.50 - 4.49) and a mean of 6.


With a +2 bonus:
(5+2)/2
(5+2) + (5+2)/2

Is a listed range of 4 - 11, but is really 3.5 - 10.5
With a mean of 7 and tiny chance of getting 11.

This keeps the mean increasing by 1 for each +1 bonus.

[PvtBenjamin]

Yup that does it


So probably the mean (or very close) determines the distribution starting at [2 1/2 - 7 1/2] then rounded.

Odd numbers have very low chance of highest number, even numbers lower/equal chances of low & high.

Not equal distribution resulting in a higher mean (5 1/2)

[Taxman66]

Hubert, I just recently observed a 2% (single chit {base 5, due to low tech level}) gain. This was with a mobilized Russia in July 1942 (and should have a +1 enemy catch up bonus to boot) in a current latest version pbem game.

This doesn't jive with your formula.