Researching, the Future

By CCP Greyscale
| Comments

Greetings space-labrats,

Let's talk about research!

Older blogs in this series can be found in the archive.

There's two major changes we're making to research: Copying is going to become way shorter, and we're changing how Manufacturing Efficiency (ME) and Production Efficiency (PE) work.


The headline here is that, during EVE’s summer 2014 release, we're going to change base copy-time on all blueprints to be the same as base build-time. With Industry giving a maximum 20% reduction in build times, and Science giving a maximum 25% reduction in copy times, this ends up meaning that (with maxed skills) copying takes 6.25% less time than manufacturing from a given blueprint.

The primary reason for this is to try and unify T1, T2 and T3 manufacturing around being dominated by build-from-copy rather than build-from-original. This gives a more unified experience and allows us to make more liberal assumptions about how much risk people are prepared to make with their build jobs. To give a concrete example, this makes us much more comfortable about removing the ability to build from a starbase using a blueprint in a station - when you're only risking the copy, this is a much smaller concern than when you're risking a BPO.

For comparison, currently the trend is that T1 blueprints take 20x longer to copy than to build, and T2 blueprints (for those lucky few BPO owners) take 100x longer to copy than to build.

Speaking of T2, this change is obviously very relevant to invention. We've had a good look at the numbers here and are currently of the opinion that this ends up being much less significant than it at first looks. The majority of invention work currently seems to be of the invent-to-build variety rather than invent-to-sell-blueprint, and on a character with the same number of possible research and manufacturing jobs, the bottleneck is pretty much always in the build time rather than the copy or invention time -- and frequently there's a further bottleneck in that invention anyway takes longer than copying. Unless this leads to a major expansion of invent-to-sell, the actual throughput of invention should not significantly change as a result of shorter copy times.


ME and PE

First of all, we're changing the name of PE (Production Efficiency) to TE (Time Efficiency), on the grounds that it's then much clearer what it actually does. ME is now called ME (Material Efficiency) everywhere, rather than sometimes being "Material Level" instead.

The major change underpinning most of the work done here is to move blueprint research to a more skill-training-like system, with a fixed number of researchable levels with identical bonuses but increasing research time.

The current system is interesting and flexible but also pretty unintuitive and with some core accessibility issues. Right now every level of ME or TE for a given blueprint takes the same amount of time to research, but gives fairly sharply diminishing returns: The first level is worth 50% of the total available bonus, the second level takes you to 66%, the third 75% and so on, so that by level 9 you already have 90% of the total available bonus. You can then go off and research into infinity to close off as much of that last 10% as you can stomach; in reality you reach "perfect" at the point where the potential savings round to zero, but this is a seven-digit number for battleships.

The major reason for changing this system is that the current relationship between ME/TE level and actual savings is difficult to understand without wrapping your head around some reasonably non-trivial math, which adds fairly substantially to the mental barrier of entry for industry and generally makes the system more opaque than it really needs to be.


No more waste

We are removing the concept of "wastage" from the system. With the reprocessing changes we no longer need to be quite so careful about build requirements dropping below reprocessing returns (which are now 50% rather than 100% of build materials), so we're removing the current system of taking "perfect build" requirements, adding waste on top and then removing it again with ME.

In the summer release, all base build costs will be increased by 11%. ME will then reduce build costs by 10% of that new base value, which brings us back to where we started. The rest of this blog is working from that new, higher base.

Production time already uses this dynamic, so no changes are needed there.

The Material Efficiency skill will be repurposed, stay tuned for more information on that in a future blog.

(The reason we’re going up by 11% is that we’re multiplying the new base by 0.9, so we want to divide the old base by 0.9 so the two cancel out. 1/0.9 = 1.1 recurring, so the actual multiplier we use will be 11.111111…%.)


The new system

Blueprint research will then be moved to a ten-step system. Each step of ME research will reduce material requirements on that blueprint by 1%, and each step of TE research will reduce manufacturing time on that blueprint by 2%. These values will be displayed as their actual percentages, rather than their step numbers, so a blueprint that has been researched six times in each will show as ME 6% and TE 12%.

Research times will follow essentially the same curve as skills, with the additional five steps interpolated between the existing skill steps, and blueprint "ranks" multiplying the research times for more advanced blueprints.

For a "rank 1" blueprint, such as T1 ammo, the unmodified research times will look like this:

Level Seconds Approx Total
1 105 1m 45s
2 250 4m
3 595 10m
4 1414 24m
5 3360 56m
6 8000 2h
7 19000 5h
8 45255 12h
9 107700 1d
10 256000 3d

Skills and other bonuses will then reduce times as they do currently.

Negative ME and TE levels work pretty much as you'd expect in this sort of system, being converted into direct percentage values. TE -4 will thus now be shown as TE -100%, and all the various decryptors (and related code) will be updated to match.

This gives us a pretty clear system that is easy to wrap your head around, works for pretty much everything, and provides a point at which blueprint research is "done" for a given blueprint and needs no further work.



What we're doing with non-standard blueprints

We would like to unify everything around a clear, consistent set of numbers, which means rooting out and eliminating as many of the anomalies as possible. This will for example mean moving Warfare Links to the same ME 10%/TE 20% system as everything else, removing the weird outsized TE bonus on fuel blocks (I'm pretty sure that's my fault and I'm pretty sure I did it by accident), and generally eliminating most of the other quirks, except where the blueprint or product is sufficiently unique anyway that it's justifiable to keep its oddball status.

How we're selecting ranks

T1 ammo will be the Rank 1 baseline. These blueprints all currently have a base research time of 6,000 seconds (i.e., 1h 40m). To determine ranks of other blueprints, the current plan is to simply divide their research times by 6,000, maintaining the research ratios currently present. This means T1 modules are generally rank 2, T1 frigates/cruisers/battleships are 20/40/60, and the highest current rank is Titans at 3414 (good luck maxing that out!). This is pretty easy to further adjust, and we may for example want to pull ships up or down a little.

How we're going to turn old blueprints into new blueprints

Obviously we need some way to remap old blueprint research values into new values, and equally obviously we're not just going to turn an ME 120 blueprint into an ME 120% blueprint.

Our current line of thinking is to do the math such that no currently researched blueprint gets any worse in terms of the bonus it provides (weird anomalies like fuel blocks getting a general TE nerf notwithstanding).

This would mean that ME/TE 1 become ME 5%/TE 10%, ME/TE 5-9 become ME 9%/TE 18%, and anything over ME/TE 10 currently move to ME 10%/TE 20%.

This has the advantage of everyone's blueprints staying the same or getting better, but the drawback that people with blueprints researched to really high levels will find that, while their blueprints are still generally getting better (i.e., will be actually perfect rather than just a reasonable approximation thereof), other blueprints which were less-well researched will have caught up with them.

We're very aware that some of you will feel that you've lost your previous advantages gained by researching blueprints for a really long time, and this is one of the areas we're preparing to focus the most on in terms of receiving feedback and making adjustments or additions to smooth the transition. Everything is on the table in terms of finding a reasonable solution that meets everyone's legitimate concerns, so please approach the feedback in terms of telling us what you'd like to see rather than simply expressing frustration with the changes as described here. We're not done with this yet!

For transitioning negative ME blueprints, we're just multiplying by 10, while for negative TE we're subtracting 1 and multiplying by 20, which keeps both values roughly the same before and after for the reasons outlined above.



That covers pretty much all the changes to research. The functioning of the system is pretty solidified design-wise, but all the numbers are very easy to adjust. We're anticipating making further balance changes in response to feedback, so, please let us know in the comments if there's anything here you see as especially problematic -- and why -- and we'll try and address as many concerns as possible!


-CCP Greyscale, on behalf of Team Super Friends