Tuesday, October 9, 2012

Everything You Never Wanted To Know: Invention

Since it will be nearly a month until I have any decent results to show off, I might as well fill the blog with useful articles.  I hope to make this a weekly (or twice per week) segment on topics the general EVE public may not know about.  The sort of "mysterious black box" that scares most of the PVP elite off because of :numbers:.  These won't be in any particular order, so as to not take 3 weeks to get to the interesting parts.

Introduction

I could start with T2 BPOs and their weird niche, but it's been 4 years since invention replaced the BPO lottery.  If you've been around EVE that long, you've heard all there is to know about them.  Instead, we will focus on the mechanics around invention and the sort of industry that builds up around it.

At its most basic, invention is a chance-based mechanic that turns T1 BPCs into T2 BPCs.  These T2 BPCs are negative efficiency, so they take more materials and more time than BPOs, but make up for this shortfall with volume.  On the surface, the system looks like: throwing away money on datacores THEN throwing away money on inefficient building.  On the contrary, BECAUSE of that inefficiency, there are enormous profits to be had.

I cannot stress enough how much invention is a VOLUME game.  The temptation for small-time industrialists and T1 is to mine what you need and build it yourself... making it all free (wrong, by the way).  Invention has the dual problem of not being feasible to DIY materials (moon materials, enough datacores, etc), and produces very large quantities of product, that no one person is going to churn through in any short while.  Seriously, what are you going to do with 100 Ballistic Control Unit II's?

:Math:

Inventing

There are some excellent guides on the exact :math: on sites like Grismar and Chruker.  Go play with their tools to figure out the exact numbers.  I will review the top-level :math: and boil down the concepts.

One of the basic tenants of probability is: do something enough times, the probability falls out to a percentage.  Flip 1M fair coins, you'll be pretty damn close to 50/50.  Don't focus on the streaks, failures, successes; only focus on the long-term picture.  The secret to normalizing datacore costs is to invent enough to be able to use the P[success] numbers without tracking real successes.

This means settling on a build quantity BEFORE you start inventing.  With all-4 skills (except ships), this is essentially 50% success rate (48.5% if you want to be a stickler).  Factor in the datacores as a build expense by using:
  • (cost per attempt) * (1/P[success]) = 1 Blueprint
  • 1 T2 BPC (usually) = 10 runs
  • Invention cost = [(datacores per attempt) * (1/P[success])]/(BPC runs)
Also, some answeres to basic FAQ's for invention mechanics:
  • Copying: always max-runs (except ships).  T2 BPC output is modified by (runs/max runs)*(T2 BPC max runs)
    • (30 runs/ 300 max) * (10 runs T2) = 1 run T2 BPC
  • ME/PE research?  
    • NO: T2 BPC output is only governed by decryptor or lack thereof.
    • 0/0 BPO's are highly encouraged for invention BPCs
  • Data Interfaces, Datacores, Decryptors, Meta Items
    • Data interfaces: not consumed.  Essentially a "key" to allow invention
    • Datacore: consumed every invention attempt
    • Decryptors: consumed (1 per attempt), OPTIONAL
      • Changes ME/PE and resulting runs
    • Meta Items, OPTIONAL

Manufacturing

The real bottleneck in T2 invention is manufacturing.  Due to the -4 PE penalty, and the requirement to consume the entire BPC to make up the datacore cost, build times can range from 8hrs/10-run to 3d/10-run or even more.  Therefore, a focus must be paid on how much throughput one builder can actually make.  Consider the following:
  • 1 BPO can yield 20 copies at a time
  • 20 copies * ~50% success rate = 10 T2 BPCs
  • Build time of ~1d = 5-7x per week that "kit" can be run (with adv mass production 4)
The obnoxious thing that crops up is that no T2 BPCs match a normal human schedule.  2, 4, 8, 16, 24 hours fit well into a normal schedule.  Many products fit in 18, 22, 28, 36 hour schedules (even with POS bonuses).  This can make throughput tricky to calculate, since most people won't log in at 4a, or from work, to make sure their production chains are running 100% efficiently.  Take careful consideration of build time to understand what is actually possible to fit into a schedule.

Also, do not forget that T2 requires intermediate products.  Components like microprocessors are MUCH cheaper to build yourself, but take manufacturing time away from T2.  This is why T2 is such a great team-effort since work can be combined and more T2 can be produced as intermediates are handled in more efficient batches.

Kits

The easiest way to account for all the costs is to treat the process like a more-complicated T1 build.  Instead of only requiring minerals, your T2 kit will need to look like:
  • BPCs (POS makes these essentially free)
  • Datacores
  • Components (or the moon materials to build them)
  • R.A.M.
  • T1 counterpart
  • Minerals
This way, at the end you have purchased the above kit for X and sold the product for Y.  Profit = Y-X.  

Advanced Topics

Ships

Ships are a special case.  The P[success] is much lower, and the invention time is much longer.  This makes the previous assumption of "high volume simplifies probability" less valid.  Also, the problem of manufacture being the bottleneck is switched with invention, since the time required to generate BPCs is much longer than the product build time.  Also, with volumes reduced to 1 ship per blueprint, the existing BPO stock has a higher impact on profitability (though still a small one).  Lastly, ships are sexy, everyone wants to build ships, many people are not accounting their math right.  Very fine attention must be paid to profitability.  There is a lot of ISK to make on T2 ships, but it requires a lot more effort.

Tricks to keep in mind:
  • Only make single-run BPCs.  Ships default to 1-run BPCs on success, max-run concept does not apply
  • Get a POS: invent time is halved, copy time bonus helps too
  • Component build is very large: Almost half the build time for ships is components
  • Expensive kit costs (normalized by 20-copies per BPO):
    • Frigs: 200M/BPO
    • Cruisers: 1B/BPO
    • BS/BC: 2.5B/BPO
  • SOME ships benefit from decryptors, but MOST are barely worth the extra trouble

Decryptors

Decryptors modify the efficiency, P[success], and runs per T2 BPC.  Accounting for them is the same as datacores ((1 decryptor + datacores)* (1/P[success_modified]))/(runs_modified).  But since the resultant BPC is changed in ME/PE, it's not as simple to just add that cost to the existing calculation for the default -4/-4.  The math for handling the calculations is a little obnoxious, but EVE HQ's Prism tool can handle negative ME/PE correctly, so that's a good base point to start with.

Some tricks:
  • The 3rd tier decryptors are most often the most useful
    • Increases efficiencies to -1/-1
    • Increases P[success] to ~60%
    • Does not modify runs
  • Basically ONLY for ship invention (reduces datacore costs)
  • Will increase kit cost, but will also increase profit/unit

Conclusions (TL;DR)

The main point to take away from this guide is T2 invention is a large volume game.  Because the invention mechanic forces large batches to account for datacores, you will be producing items in the hundreds.  This is one of the reasons people shy away from invention; they cannot DIY all their equipment because of the various restrictions.

As such, I have always run invention programs as "cash generators".  Even if we went and bought the items we sold back, getting the cash to get the small volumes required for ship fitting was so much easier than accounting for surplus and stockpiling.  Also, to get effective stockpiles you would need to run a very broad program building nearly everything, which takes a ton of man-power and cash.  I will cover more topics on accounting in future segments