Today's Japanese R&D lesson

[Lowpe]

7 factories size 30 yielded a 1 year and 4 month advance for a final generation fighter, but I cannot find the start date for the factories. Highly doubt it was from the beginning of the game.

[GetAssista]

ORIGINAL: Lowpe
7 factories size 30 yielded a 1 year and 4 month advance for a final generation fighter, but I cannot find the start date for the factories. Highly doubt it was from the beginning of the game.

Can easily be done for Ha-45 planes, engine bonus is much easier for those

[Lowpe]

The example I cited above was with the Ha 43 engine...[:)]

My point is simply to illustrate that researching planes more than 2 years in advance doesn't yield almost a 0% chance of repairing as earlier mentioned. Yes, the repair chance isn't fast, but getting a final generation fighter 1.5 years early isn't to be discounted especially in a pdu on game.

However, it does come at a steep opportunity cost.[:)]

In a pdu off game, getting that final generation fighter (say the Ki94) doesn't help much if you haven't gotten all the other upgrades in the army fighter line: oscar line, tojo line, tony line, frank line. You will have the plane in the pools, but no squadrons that can fly them![:D]

[GetAssista]

ORIGINAL: Lowpe
The example I cited above was with the Ha 43 engine...

Checked mu calcs, they are ok. Must be earlier factory repair then for you, I dunno if Obvert's date for Ki-95 I used is representative. Or engine bonus. Or Shinden which is later than Sam for couple months

ORIGINAL: Lowpe
My point is simply to illustrate that researching planes more than 2 years in advance doesn't yield almost a 0% chance of repairing as earlier mentioned. Yes, the repair chance isn't fast, but getting a final generation fighter 1.5 years early isn't to be discounted especially in a pdu on game.

That is correct, you would be better off overall not neglecting later planes from start. And early planes too. Useful planes that is )

[Numdydar]

Gents thanks for the discussion. But I will definitely have to disagree with your viewpoints.

I never said that the repair rate for 2+ years in advance R&D factories was zero, I said almost non-existence. As shown by my 0(31) for a 6/45 historical date plane. So even after getting within the 2 year window (it is 5/44) I still do not have a single one repaired. I also have quite a few still at 0(2) as well.

If you are playing something other than Scenario 1/2 then it may work differently. But for these two it is not worth it for me to do this. Like I said your RND may be better than mine [:)]

One thing that was not mentioned is you ALSO need to expand the engine factory. No good to get the planes early but no engines [:)]. So they need to repair (although they do repair faster) and start producing before the planes are ready. So another cost to Japan.

I hope it works out for you. It will be interesting to see what if any impact it will have to the game depending on when you get them. Of course it will be a few years before we find out [:(]

[PaxMondo]

ORIGINAL: GetAssista


i made some illustrative calculations:
E.g. assume you started the game, decided to allocate say 10 factories to your navy fighter research, and ponder A6M5 vs A7M2 allocation. Assume Rufe->A6M5 path, bypassing M3s for simplicity. Key thing is that when A6M5 arrives you would switch all research to A7M2.
1. A6M5 factories repair (as Rufes) in 3/42 and start researching A6M5 with upgrade chain switch.
2. When do Sam factories repair? Luckily, Ki-95 has same arrival date as Sam at 10/45, so we can assume all A7M2 factories fully repair at Obvert's date, which is around 5/44 (no engine bonus and 1 factory, 100 days per advancement, 4 advancements made until arrival at 6/45, 5/44 + 100*4 ~ 6/45 sounds about right).
3. What about engine bonuses? Ha-35 bonus is almost always there, highly sought for, and we can safely assume it works for the duration of A6M5 research. Ha-43 is tricky with arrival date at 9/45, so it would be a strain to get bonus for Sam research in 1944. Assume no bonus for all duration for simplicity.
4. Now what are the dates of arrival for both planes given all research done? Some guesses here about the timing of A6M5 switched to A7M2 repair but nothing wild:

Code: Select all

 A6M5  A6M5    A7M2  switched fac   A7M2 w/o switch A7M2 with switch
 fac   arrival fac   repair at?     arrival         arrival
 --------------------------------------------------------------------
  0     8/43   10    -              10/44        =  10/44
  1     2/43    9    late 44 nogood 10/44        =  10/44         
  5     8/42    5    ~10/44         12/44           11/44
  9     6/42    1    ~9/44           6/45           12/44          
 10     6/42    0    ~9/44          10/45            1/45

You trade earlier A6M5 vs earlier A7M2 as expected, but giving some R&D to each model in the beginning (like 5/5) gives you much bigger combined time advancement. In the end it all depends on how you value different planes early availability.
Supply cost is also different but I assumed 100k supplies is manageable
Engine bonus presence/absence tilts things somewhat but does not change fundamentals

Edit: there is luck involved here and I might've unerestimated its importance. When R&D repair is lucky for you, you can get couple fully repaired factories earlier and this is not a small matter as with engine bonus 1 advancement a month not only gets research done but also makes arrival date closer and allows other factories to repair faster. So quantity of R&D factories is a quality of it's own )

You should definitely search for the earlier threads on this. This has all been discussed to the fullest extent possible with all of the knowledge that the devs are willing (or able) to share.
You have some erroneous assumptions above that you will want to correct. Again, all has been discussed and then tested. You need to look back ~4 years to find the threads.

YMMV.

[Lokasenna]

FWIW, I have had some A7M factories since game start. I have been watching one every day for the past few days, now in November 1943, as it is nearing fully repaired status. I suspect it didn't begin to repair at a decent rate until mid-43, but it was definitely part of the way along before then. According to Tracker, my combined factories were over 50% repaired by turn 556 (the earliest turn I have in this Tracker database). That's sometime around June 1943, which would be the 2-years-out date (standard arrival for A7M2 is 45/6).

It's extremely easy to burn hundreds of thousands of supply on aircraft R&D when you don't need to. Just set your factories at the beginning, even with PDU On - you don't need wonder planes in 1943. You just need better planes than the Allies have. Or even just comparable planes, if your pilots are OK.

[GetAssista]

ORIGINAL: Numdydar
I never said that the repair rate for 2+ years in advance R&D factories was zero, I said almost non-existence. As shown by my 0(31) for a 6/45 historical date plane. So even after getting within the 2 year window (it is 5/44) I still do not have a single one repaired.

That looks extremely unlucky. I think you meant 5/43 as your running date though, as 5/44 and 0/31 definitely look like a bug or no supply on base (which I'm sure is not your case).
I set some size 30 factories from start in my last vs-AI stock Scenario 1, and by Feb 5, 1942 that is only 2 months Sam (9/45 arrival) gets 0,2,3 repaired, while Grace (12/44) gets 0,0,1,2,3. No other late planes for me.

One thing that was not mentioned is you ALSO need to expand the engine factory. No good to get the planes early but no engines . So they need to repair (although they do repair faster) and start producing before the planes are ready. So another cost to Japan.

Those are really very predictable as engines always repair 1 factory per day, so it's only about your willingness to expend supply on expansions. Some of the engines are no brainer to advance and even get full bonuses to all R&D because of early arrival date (Ha-45). Others are trickier but still easily manageable to get into production on time if you not sweat about engine bonus, because engine research starts much earlier (practically from Jan 42) than for airframes. Ofcourse you still need to make choices early because you can't accelerate everything at once

ORIGINAL: PaxMondo
You have some erroneous assumptions above that you will want to correct.

Care to point those out in a few words?

Edit:
While all is known about how R&D points are generated (and I assume 30 size factories always with 1 point per day w/o engine bonus and 2 points with), repair algorithm seems more obscure.
tm.asp?m=3022096
This is the only thread with R&D repair estimations where guys refer to actual simulations done to gauge the formula for R&D repair. Results in a nutshell mean that between the date of starting research and scenario date of arrival of the plane R&D Factories repair for 63% of available time on average, hence leaving 37% to actual research. Size of the factory does not matter timewise
With my example above staring R&D repair on A7M2 in 8/12/41 with arrival in 1/10/45 (it's wrong, is 9/45 in stock 1, but this does not change the conclusions) adding 63% means 4/5/44. Which is exactly the month I inferred for my calculations using Obverts Ki-95 experience. Woohoo!

Edit2:
I let the scientist in me out for a while, built myself a quick R&D simulator using repair probabilities mentioned in the thread I linked, so here is some interesting resulting eyecandy for you all. Results from several simulations of R&D from gamestart using different number of factories involved on an airframe available in 10/45. Illustrates nicely the law of diminising returns.
The actual ingame probability of R&D repair might be somewhat different, need to test it myself sometime in the future. But the mechanics is the same, with hyperbolic decrease in repair ability for far-off planes.
Surprisingly for me, luck does not play as much a factor as I expected. While first fully repaired factory can arrive with wildly varying timing, actual plane arrival is pretty tight

[PaxMondo]

[tm.asp?m=3022096
This is the only thread with R&D repair estimations where guys refer to actual simulations done to gauge the formula for R&D repair.
[/quote]
No, there is a much earlier thread must be 2010? where someone (Yaab maybe? I forget exactly who) sandboxed the R&D several times to examine how the '46 plane R&D factories would repair. It was quite an interesting study. Now, it is possible that the algorithm was changed in the ensuing 100 or so beta's since that work was done. Maybe someone will repeat the testing ....
Alimentary's formula, while great math, never had empirical testing to support it or at least I never read it.

[GetAssista]

ORIGINAL: PaxMondo
Alimentary's formula, while great math, never had empirical testing to support it or at least I never read it.

I was referring to kmitahj's remark of his own sandboxing results coinciding with alimentary's formula. Seen some other referrals to 63%/37% guideline here and there too. Maybe I'll do sandboxing when I have some time, it's pretty straightforward, only time consuming.
I'm also wary to look way too back in the past. There were some bugs introduced into R&D and later removed, that made their way into AARs and experiences, with people achieving ridiculous advancements that are no longer possible now.

[Crazypantoufle]

I usually burn a lot of supply in scenario 1 as i use all the R&D factories.
But i don't think it's a waste. It all dépends on the chain for each model.

I don't hesitate to use like 10 R&D factories for models like Zero, as it's not only about getting the A6m5 much earlier, but also the A6m8 (armored A6m5!) much much earlier. Same for the Tojo, as the IIc version is a real improvement compared to the IIa (armor+ slightly better armament) and it will be my main fighter type for most of my IJA squadrons, even when Frank (and his horrible service rating) will appear. So for this models i really burn supply to get them earlier because i feel it really worth it, and most of the factories will be kept on R&D when the first models of the chain appear.

And for the other important but later models with short chains (Frank, Georges, Judy, Sam, ect), i use 3 or 4 R&D factories. If i get them a few months earlier, it's fine, but the real interest is to have fully functionnal production factories from day 1 when the plane model becomes available .

I even use 1 R&D factory for future secondary plane models i plan to use (like transports, recon or night fighters) to get the production ready.

It's my main point for most of my R&D supply investment: the supply i have to use in expending the R&D factories won't have to be used in changing and expending again production factories from scratch + i get full production chains from day 1 for the new models.

[Lowpe]

GetAssista: Run your program with a different date for me, pretty please.[:)]

Say you want to get a 10/45 plane, but you start your research effort 3/43(say at this point you just concluded a massive Oscar r&d plan and have up to 10 factories to re-allocate). What advancement can you expect with 2, 5, and 10 factories?

PS: I guess my point being is can I run my r&d in two massive waves...the 43 planes I want early, and then the 45 planes I want early. Is this a better strategy than setting up say 10 factories for the A7 at game start...other than in cost of supplies?

RE: Cost of supplies argument. If you fall too far behind the air war curve, and allow the Allies an upper hand early, then the 300K supplies spent on getting the Frank A and R early or A7 will pale in comparison to the beating you will take from the Allies.

[GetAssista]

Lowpe, here you are. No engine bonus, mind it

I included both set of simulations of individual factories + a separate calculation based on 63/37 rule. The latter estimates expected arrival date pretty nicely, although luck with individual repairs can still shift arrival date +-1 month. Divergence between simulation and 63/37 grows for larger number of factories because early research outliers matter a lot I guess.
Re your question it looks like the best strategy would be to allocate 2-3 factories to important late planes from the very start, and then help them with additional factories later when those become free from researching earlier planes.