MoO 1 101: "seeding a tech"

[Zalson]

I've seen a bunch of people mention that they'll "seed" a tech.

What's the user mechanism? Is it: max out spend in one field and reduce all other fields to one click to preserve investment? Then, next turn, adjust spend to get the return on investment  and redirect the rest of your research into another category/categories?

Or is it something else?

Thanks!

[RefSteel]

"Seeding" really just means committing a large amount of research to a tech as soon as you begin researching it in earnest, no matter how. The idea is to maximize your initial investment, so that your future spending on that tech will get the maximum benefit out of MoO's tech investment bonus: Each turn, all spending on a tech up to 7.5% of the total amount you had previously spent on that tech ... is tripled! So if you "seed" a tech with 200 RP on the first possible turn, and then spend 15 RP on it the next turn, you will actually get 45 BC added to the tech that turn - and then the following turn, up to 18 RP will be tripled: 7.5%, rounded down, of 245. In general, it is not a good idea to take away spending from other (~equally important) techs to do so if this would drop spending on them noticeably below the level where the LED by their bulb is lit (unless they're well into the percentages) because you'd be "robbing Peter to pay Paul" - giving up the opportunity to triple your research spending right now so that you'll be able to get triple points for a smaller amount of spending later on. As with everything in MoO, there can be exceptions: Strategic considerations always take precedence.

In theory, the most efficient tech spending path (ignoring rounding errors, which actually make this very inefficient) would be to barely seed each tech at all, and then spend just barely enough each turn to keep the LED lit. In reality though, this is a terrible move, because even though nearly all your research points will be tripled, you won't be spending enough on research total, likely pouring the rest into ships or defenses you don't really need, and you'll fall way behind the other races in the galaxy. The idea behind "seeding" is that you're anticipating spending some RP "inefficiently" and you can minimize this amount by getting as much of it as possible out of the way as early as possible. (The way I think of this is that "normal" research spending means building labs and schools and other experimental facilities dedicated to that tech, and the later spending that gets tripled - often referred to as "trickle spending" because the amount you spend per turn is much smaller than the "seed" at first, and slowly increases each turn, "trickling" upward) - means staffing and taking full advantage of your existing dedicated research facilities. It makes way more sense to build a bunch of labs when you start a new research project than to do so at the end!

For an important tech, I will often temporary halt construction of factories, terraforming, and the like to pour more BCs into the new tech for a turn.  At other times, if I have rich or ultra-rich planets in the empire, I may empty (a significant portion of) my reserves to splurge-spend on research planets to start up a new technology without reducing net spending on the other ones I'm teching.  If all my planets are maxed out and focused on research, and I don't have any rich worlds to feed the reserves ... ... I probably haven't been expanding enough (whether in peace or by war).  But if that is the case anyway, unless a tech I previously thought important has lost a lot of value to me for some reason, I'll generally wait until another tech is far enough into the percentages that I want to reduce spending on it, and use the excess to at least start to seed the new tech.

[rho21]

Excellent summary.

A few notes:
Kyrub's patch provides a display for this: a single pixel for each tech. It is dark if you are spending less than 7.5% of your current investment on it, and lights up otherwise.
Spending nothing at all on a tech causes it to lose invested research. Avoid this.
Trickle spending continues to work into the percentages, but you will often want to cut spending to one click at some point anyway.

As for my personal way of playing:
When I want to open up tech early on, I will tend to give all fields 1RP each if possible, just to see what's there. Next turn, I turn up research spending and seed one of the techs I want to research in earnest. After this, I will set the spending for any seeded techs to just enough to maximise the bonus, with the rest of spending going into seeding another field. Once everything I want to research right now is seeded, research spending is slashed in favour of infrastructure, leaving just enough to get the maximum trickle effect.

In the mid-game, I will tend to follow a similar strategy: techs will trickle along except for one field (whatever looks most important) that gets any spare spending. Whenever a tech comes in, I will drop spending on that field to one click (if it isn't already there from waiting in the percentages). Once enough techs have finished, I will have the spare spending to start seeding the new ones. Obviously, this varies when a new tech needs to be finished ASAP.

At some point (roughly when I'm sure I'm winning), my interest in micromanaging this disappears and spending gets approximately equalised.

[Zalson]

Thanks for the thorough explanation, rho21 and RefSteel.

To summarize (accurately, I hope): "seeding" is dumping RP into a tech to get a 3x bonus on 7.5% of the total spend in the following turn.

Incidentally, how many ticks are possible on the search slider? n = 100% so n-5 is about: 95% of RP? 90%? (yes, I'm trying to figure out about how many ticks are equal to 7.5% of RP spend).

[RefSteel]

Each click represents 2% of your research budget. Using the "=" key, you can subdivide that further with four "half-clicks" representing 1% apiece. But ... I think you're misunderstanding the way the research system works: There's nothing significant about 7.5% of the current turn's research budget (which is what the clicks represent) unless you happen to have an identical research budget this turn to last turn and you put 100% of your research into a single tech last turn and it was the first turn you put any research into that single tech. So to get exactly the best bonus you can, just add whatever number of clicks it takes (and/or change the sliders on some of your planets so you're doing more or less research in place of other stuff) to make the little LED light by the lightbulb "light up."

You always get triple value (so a 2x "bonus") on the total amount you have previously invested in every tech you're researching. So the amount on which you can get the bonus increases every turn as long as you keep spending research on the tech, because the total investment so far keeps going up. So if you seeded a tech for 200 on turn 0, then you get the bonus on 15 of the research points you spend on that tech on turn 1 - whether that's 100% of a total imperial budget of 15 RP or 5% of a total budget of 300 RP (and whether you spent 15 on the tech to get 45 actual RP invested or spend 200 more on the tech to get 230 actual RP invested, thereby "seeding" the tech further). The number of clicks in the field determines research spending on a given tech relative to the present research budget.

So you don't seed a tech to get the bonus: You're going to get the bonus anyway. You seed a tech so it will complete sooner! Again, the "most efficient" way of teching is to just make sure the LED for each tech is just barely lit (that is, make sure one less click in the field would mean the LED "turns off") every turn, but this would mean spending almost nothing on research and therefore waiting forever before you actually get any tech. "Seeding" the tech means you can still follow this "most-efficient" practice after seeding it, but actually complete the tech in a timely fashion because you've skipped many turns of trickle-spending up from nothing with a single turn of "inefficient" spending.

[RFS-81]

(The way I think of this is that "normal" research spending means building labs and schools and other experimental facilities dedicated to that tech, and the later spending that gets tripled - often referred to as "trickle spending" because the amount you spend per turn is much smaller than the "seed" at first, and slowly increases each turn, "trickling" upward) - means staffing and taking full advantage of your existing dedicated research facilities.  It makes way more sense to build a bunch of labs when you start a new research project than to do so at the end!

I like this idea, it adds some flavour to a very abstract mechanic!

Whenever a tech comes in, I will drop spending on that field to one click (if it isn't already there from waiting in the percentages).

This seems like a good place to warn about an UI bug that has bitten me recently: In the early game, if you have less than 50 RP in research, a single click doesn't contribute any RP to the chosen field (2% of 49 is rounded down to 0). If you invest no RP, there not only is no discovery chance, but your accumulated research in that field even "decays". But if the slider isn't set to 0%, the interface will still display a discovery chance.

[rho21]

Thought I might throw together some concrete examples. Suppose you want an investment of 3000RP towards your tech.

These are the investment strategies:
Slowest trickle: invest 14RP as a seed (the minimum to qualify for a bonus), then trickle spend.
Seed and trickle: invest 200RP as a seed, then trickle spend.
Constant spending: invest 40RP every turn.
Instant spending: invest 3000RP on one turn.

And the results:

Plan	Turns	RP spent
Slowest trickle	30	1009
Seed and trickle	15	1133
Constant spending	35	1400
Instant spending	1	3000

Really shows the value of trickle spending, and how a seed jumps over the slow part of the trickle without actually costing all that much more.

Hopefully I've got my formulae right.

[RFS-81]

Slowest trickle, seed-and-trickle, and obviously instant spending look roughly correct to me, ignoring rounding errors in the game. But how did you calculate the constant-spending case? I've written a Python script to calculate the time for seed-spending and constant-spending taking rounding errors into account (see attachment), and my result for the constant-spending case is 29 turns and 1160 rp.

Also, can someone double-check if I understand the bonus correctly? As far as I understand, it works as follows: The maximal bonus is 15% of the invested rp rounded down. The actual bonus is 2 times your spending this turn, or the maximal bonus, whichever is lower. That also means that if the maximal bonus is odd, you have to choose between rounding your trickle-spending up or down. If you seed a very small amount, this can actually make a huge difference: With my script I see the following outcomes (the first number is the number of turns, the second the amount of rp paid):

Code:

>>> seed_spend(14, 3000, round_up = True)
(29, 1222)
>>> seed_spend(14, 3000, round_up = False)
(30, 1023)

In an actual game, you can only adjust spending in 2% clicks anyway, so I don't think this matters, and also the effect is much smaller if you seed 200 rp.

EDIT: The forum won't let me add the script as attachment, so here it is inline:

Code:

from math import (floor, ceil)

def max_bonus(rp):
   """The maximal bonus one can receive: 15% of the invested rp."""
   return floor(rp * 0.15)

def new_rp(current_rp, rp_this_turn):
   """Calculate the research points at the end of the turn when
   current_rp points are already invested in a tech, and rp_this_turn
   is the number of points added this turn."""
   # The number of bonus research points is max_bonus, or twice the
   # number of points we spend this turn, whichever is lower.
   bonus = min(max_bonus(current_rp), 2*rp_this_turn)
   return current_rp + rp_this_turn + bonus

def seed_spend(seed, cost, round_up = True):
   """Calculate the time and money required when you invest seed rp in
   a tech, and then spend just enough every turn to receive the full
   bonus. When max_bonus is odd, this amount is rounded up if round_up
   is True and rounded down otherwise."""
   paid = seed
   rp = seed
   t = 1
   while rp < cost:
       if round_up:
           spending = max(ceil(max_bonus(rp)/2), 1)
       else:
           spending = max(floor(max_bonus(rp)/2), 1)
       paid += spending
       rp = new_rp(rp, spending)
       t += 1
   return (t, paid)

def const_spend(rp_per_turn, cost):
   """Calculate the time and money required when you invest a fixed
   amount of rp every turn into a tech."""
   t = 0
   paid = 0
   rp = 0
   while rp < cost:
       paid += rp_per_turn
       rp = new_rp(rp, rp_per_turn)
       t += 1
   return (t, paid)

[haphazard1]

I agree with RFS-81 that the constant spending numbers look off. It does not seem like it should be slower than the slowest trickle, which is trading time for maximum bonus/efficiency of spending.

Thanks for putting this info together -- very interesting to see the comparison. I knew that seed-then-trickle should be faster, but that it would be that much faster for very nearly the same total cost is unexpected. For me, at least -- maybe this was already known to stronger players.

[RefSteel]

It is indeed possible that constant spending of exactly 40 RP would be slower than pure trickle research (or at least a very small seed) for a 3,000 RP tech. This is because "trickle research" doesn't actually mean spending a tiny amount every turn; it means increasing spending on a tech just enough to maximize the investment bonus every turn. So, once the total investment gets into the mid-500s, constand spending of 40 RP will all be tripled (so effectively putting 120 RP into the tech each turn) - but trickle spending will also all be tripled as usual, while spending increases above 40 RP per turn. By the time you approach 2,500 RP invested over the lifetime of the project, "trickle" research would mean spending nearly 200 RP on that turn, all of it tripled, getting you to the finish line immediately, while a steady rate of just 40 RP would continue to slog away slowly, tripling a much smaller amount (because you're spending a much smaller amount) each turn, and take four or five turns to finish as a result. The same is true to a lesser extent for the turns leading up to the neighborhood of 2,500 as well.

Of course, spending 40 RP per turn gets you up to the mid-500s in total investment much faster, but it doesn't suprise me that a pure trickle's slow start and 40-forever's slow finish come close to balancing out.

[EDIT: I haven't run the numbers in depth, but believe it or not, I think the difference between rho-21's results and RFS-81's is just an assumption about rounding. rho-21 assumes the maximum research bonus is:

Code:

Round Down ((Total Investment So Far) x .075) x 2

Meaning you'd get up to 2 bonus RP after 14 RP have been invested, 4 after 27 RP have been invested, or 6 after 40 RP have been invested. Whereas RFS-81 assumes the maximum bonus is:

Code:

Round Down (Total Investment So Far) x .15

This gives roughly the same bonus as under rho-21's assumption, but would provide an additional 1 RP maximum bonus whenever the investment totals 40n + a, where 'n' is any whole number and 'a' is anything in the range of 7-13, 20-26, or 34-39. You'd be surprised by the speed with which these extra bonus RPs add up, as they add to the future "total investment" as well.

I ... don't remember which assumption is correct though.]