Ratjoy.com » Forums » EoS Club Rats Den » Working team: R&D
You need to register or login to post a reply.
Working team: R&D
|
Mister Death RJ: McFlono McFloninoo Post Rating: 4 + / - Total Posts: 266 Karma: 300 Joined: Feb 6, 2012 |
Posted on Apr 25, 2012 Summary: Come up with a workable R&D system that has the following properties:(a) Still based on research levels; (b) No matter how mature the game gets, there is always incentive to produce Q0 goods; (c) No matter how mature the game gets, there is always incentive to do research. There are lots of new complaints about (a) the old R&D system being unfair for a variety of reasons, and (b) the new R&D system being unfair for a variety of reasons. Rather than another ad hoc fix, I'm proposing a round-table discussion by all interested parties, starting here, starting now. Here's the starting point, which I think we should take as a given: The quality scale shall remain the set of natural numbers, with higher numbers offering benefits to players. Pretty much everything else is on the table - prices, times, benefits, and drawbacks. I'm just brainstorming in this post, so by all means tear down my ideas and/or build up some of your own. BENEFITS I think the strictly-linear benefit scale is the thing that's probably the "most broken" in the current system (and the one thing the reforms didn't touch). Established players are going to have higher research levels - not only is that to be expected, that's as it should be. What we need is a benefit scale such that Q0 goods are still worth making. One possibility - not the only one - is to make the benefits decrease with increasing quality, so that going from Q50 to Q51 confers less of an advantage than going from Q0 to Q1. Anything slower than linear would work for this - squar root, logarithmic, or something with an upper bound such as arctan. But I'm not going to dwell on this because I have another idea... DRAWBACKS Congratulations! You now have the technology to grow a Q50 apple! But... growing a Q50 apple, even if you know how, isn't as easy or cheap as growing a Q0 apple. You need to grow that apple in the right soil, with just the right water, or maybe in a greenhouse, with extra workers hovering over it to make sure that nothing bad happens to it. Instead of producing 100 bushels per tree, you've pruned all but five apples from it so that each one comes out just perfect. Some options: - Higher quality comes at a production hit. Give players the option to produce a higher number of lower-quality goods instead. - Higher quality comes at a price penalty. Give players the option to produce lower-quality goods for less. - Allow factory improvements. No matter how hard you try, you can't grow a Q50 crystal in a Q0 factory. I yield the floor. |
|
Moe Jack RJ: Moejack CO: Moe Jack Post Rating: 30 + / - Total Posts: 128 Karma: 74 Joined: Apr 7, 2012 |
Posted on Apr 25, 2012 I need to make this quick, so it's not super detailed.Say higher quality goods cost, relatively, more to produce than lower quality goods, but they fetched correspondingly higher prices at market. Then the demand could be segmented to provide for competition at different market segments. Cheap jewelry stores don't compete with Tiffany. Ferrari and Bently aren't in the same market as Kia. If you leave production the way it is, that you cant intentionally produce lower quality goods (other than buying cheaper inputs on the b2b or import) then whole markets are created for q0 and q10 and q20 products. To maximize profit, a retailer would need to buy goods of different qualities on the b2b, or else focus on one market niche. A integrated producer would be able to move up in the segments, making more money as they progress their tech, but able to compete at lower levels of quality. |
|
Random Tester RJ: Tester CO: Random Tester Post Rating: 0 + / - Total Posts: 26 Karma: 46 Joined: Jan 21, 2012 |
Posted on Apr 25, 2012 (Last edited on Apr 25, 2012) boost or penalty not for newbies, but for anyone in the industry.As a reply to a statement in the other thread, the above point is intended - why should it cost the same when you intend to move into a less saturated market? Of course this may mean scaling the Import Quality as some have suggested. EDIT: Another thing to keep in mind is the store sales is not a simple +2% per quality level. The quality part of the store equation looks like this: P_eq_part = ( max(0.3, (1 + 0.02 * (Q - Q_avg))) / P )^2 Meaning it's possible to get a 0.09 (or -91%) sales from when your product is more than 35 quality levels lower than the average sold product quality. |
|
Reiter Hexenmeister RJ: Ritter Hexenmeister Post Rating: 0 + / - Total Posts: 22 Karma: 10 Joined: Mar 26, 2012 |
Posted on Apr 25, 2012 (Last edited on Apr 25, 2012) The solution is standard deviation. Consider for example, glass Q20 as average quality and using that as origin or zero bonus with a SD of 11, I think that using that gives a better trust. It have the same or lower computational cost than powers, so will not relentize server calculus at midnight.If you are a newbie that produces Q0 glass your price bonus with old system be -40% with SD it be arround (hypotetically) -18% (Q - Qavg) / (10 * SD) which makes easier to start. On the other hand if you are a glass tycoon and you produce Q50 glass with old system your bonus been around +60%, with SD it be around 27% or so depending of multipliers. I supposed that in the game, like in reality, majority of bussines people tend to stay in the mean of a gauss distribution slanted to lower values, but if a perfect bell or slanted to higher values, give a a long term players (richest) major advantage than temporary or newers in Q level and, for the others makes the way not so harder, due to, that if you played EoS for several months or years, yor distance, in quality, from cuadratic mean is longer than new or lazy have. Proposal have the effect that, in heavy produced goods (e.g. fruits) SD is higher, flattening a little the differences in Q, which i think is the aim of the actual change, but in low produced goods, SD is low, thus Q research give aditional reward to pioneers, forwarding more people produce that. Also to incentivate research, time be proportional to 2-2.5 power instead of 3, and cost proportional to 1.3-1.7 base for exponentiation multiplier insted of 1.2 that is actually (elder players can spend more money). Above satisfy (a), (b) and (c) points posted by Mister Death EDIT: The new (actual) system have two nasty side effects, with a 10m^2 R&D you can reach average quality without effort, which unbalance maintenance and salary costs and give a big advantage to companies with high income, for that reason research must be harder than now. Second, if quality of your product are low away from average, sales drop down dramatically, unachieving pont (b), for example, with metals I have a big profit due to exploit quality improvement but I start to sell glass bottle two days ago, and have a huge low quality inventory, then my sales drop to 25% from yesterday, for fiberglass, rubber and silicon which I puchase on B2B or import, where are cheaper (for hourly cashflow reazons), with a 400m^2 store, sales level drop to same level when store have 90m^2, make low Q items unprofitable, if I'm a newbie or incursioned in other sector without previous research... there is nothing to do. |
|
Emile Lasto RJ: Emile Lasto Post Rating: 0 + / - Total Posts: 8 Karma: 10 Joined: Apr 21, 2012 |
Posted on Apr 25, 2012 I don't think the new system is as broken as people feel it is. It's just that the transition is messy.I think the follow-the-leader dynamic will end up being interesting. Relative differences in average quality levels will reflect peoples' preferences, and become much more interesting and "thematic". People will argue about whether it is worthwhile to be market leaders or not. But whether or not it is exactly optimal for profit to be a market leader, there always will be market leaders. |
|
Nwabudike Morgan RJ: CEO Nwabudike Morgan CO: CEO Nwabudike Morgan Post Rating: 15 + / - Total Posts: 108 Karma: 344 Joined: Apr 4, 2012 |
Posted on Apr 25, 2012 Goals:
Previous research system:
Current research system:
Current impact of quality on items:
I believe that the current impact of quality is appropriate, but that the manner in which research times scale is not. In my opinion, trailblazers should feel rewarded, not punished, for progressing in R&D. Being the forerunner already incurs a greater cost by paying full price for research, but the temporal costs become extreme, especially in poorly populated industries. Using the average research level means that any closed industry with several major players can internally advance as far as it likes, while someone in a small industry gets hit with extreme penalties. As the maximum research level in an item's production grows, the available buffer between the leader and the rest of the pack should increase. If the current leader has Q30 research and the pack has Q15, then the leader has no incentive to improve to Q60 research and drag the pack up to Q45. If the leader could improve to Q60 research and only allow the pack to easily improve to Q25 or Q30, then the R&D leader would gain some persistent value from further research. I think that such a buffer could be given by forcing all research to cost "full price" above a threshold that increased more slowly than the quality of the leader. For example, say that no discounted research exists before the best quality anyone has researched is Q20 (comparable to low-quality foreign goods). If the R&D leader is above that threshold, then begin to discount research that is well below the leader's research level. Continuing the example, say that research becomes cheaper if the research level is no greater than 0.6 * LeaderQ - 4. This would introduce a reasonably sized buffer (12 levels at first) that would grow another level every time the leader advanced 2-3 levels of research. |
|
Scott (Admin) RJ: Ratan Joyce CO: Ratan Joyce Post Rating: 0 + / - Total Posts: 1175 Karma: 5083 Joined: Jan 13, 2012 |
Posted on Apr 25, 2012 Thanks for all your inputs, I believe in general everyone understood the goals, but the implementation is a bit tricky.So far it seems the best thing to do is to first isolate the leader, I think I'll go with this: the average of the top 3 leading companies in an industry. And then instead of the old way (insane costs at high levels): f(level) or the newer way (too much relief in some industries, not enough in others): f(level+20-avg) Perhaps something like the following: f(level-0.3*max) Once again, another reason for the change is it's NOT a simple (1 + Q/50)^2 for the stores, but rather max(0.3, (1 + (Q-Q_avg)/50))^2 |
|
Innocent Bystander RJ: Matthew Matician Post Rating: 0 + / - Total Posts: 76 Karma: 54 Joined: Apr 2, 2012 |
Posted on Apr 25, 2012 That would be better than how it is right now, but would you consider something more akin to factor*f(level) instead, where factor was a function, say, of the single-person highest tech? (so factor is near 0 or .5 when your quality is near 0, and could scale linearly until factor hit 1 when your quality was 50-75% of max level or higher.)Doing f(level-.3*max) is still going to help the high-level players just as much as those entering the market, which I thought was what we were trying to avoid. Also, my rationale for single-max is that, for fields with only one or two active researchers, they will be penalized harshly compared to industries with 3 active researchers. Above all, thanks for (and congrats on) creating a game that a lot of people love enough to flood your forums with critiques and suggestions. |
|
Scott (Admin) RJ: Ratan Joyce CO: Ratan Joyce Post Rating: 0 + / - Total Posts: 1175 Karma: 5083 Joined: Jan 13, 2012 |
Posted on Apr 25, 2012 The thing is, the purpose for this change is NOT to make everyone suddenly catch up with the leader, it's more to mimic a system where research cost decreases as time passes. Yes, for the higher level players as well.On the surface it may appear that they are just being helped as much, but to stay x levels ahead of the curve, they're always paying (for example) $100 mil for that particular average + x level, whereas the amateur who follows is constantly paying $10k for the average - y level. Standing on the shoulders of the giants, here are my revised goals:
|
|
Nwabudike Morgan RJ: CEO Nwabudike Morgan CO: CEO Nwabudike Morgan Post Rating: 15 + / - Total Posts: 108 Karma: 344 Joined: Apr 4, 2012 |
Posted on Apr 25, 2012 Like the above poster (e: sniped, make that two posts above!), I feel that an important part of this balance is that only players at a lower research level should feel the benefits of higher technology leaders. There should be some distance below the maximum tech level in the field before any cost reductions appear at all, and in my opinion, that distance should increase at a modest rate as the maximum tech level increases. The companies below the leader, even #2 and #3, aren't particularly important--they might not even be trying to compete with the presiding research leader.If you think that the costs at high levels are "insane", then you should adjust the growth rates of the cost functions themselves. If you have an adjustment that is always applied to the function, then it isn't much of a special case for one player versus another. |
|
Scott (Admin) RJ: Ratan Joyce CO: Ratan Joyce Post Rating: 0 + / - Total Posts: 1175 Karma: 5083 Joined: Jan 13, 2012 |
Posted on Apr 25, 2012 If that's the case, then it all comes down to an ideological difference for point 5.I'm not sure how many would support me on this one. Although I'm in favor of balancing all industries, I do believe if more money is spent on research for a particular industry - then future research should cost less for that industry simply because of the data at everyone's disposal. Just made the 3rd change: f(level-0.3*max) |
|
James Grey RJ: Dairy Aire CO: Dairy Aire Post Rating: 0 + / - Total Posts: 26 Karma: 21 Joined: Apr 17, 2012 |
Posted on Apr 26, 2012 Setting up a reward such as influence everytime you sell so many of a product to the public of quality that hasn't been achieved before might be a way to increase incentive for further research.
|
|
Nick Wright RJ: Marx Post Rating: 0 + / - Total Posts: 8 Karma: 15 Joined: Apr 1, 2012 |
Posted on Apr 26, 2012 I've already plugged this idea in the other two threads, but maybe you could try the following.Instead of Cost = x * (level+20-world_avg)^1.2 Time = y * (level+15-world_avg)^1.2 Try Cost = x * e^((a - b)/a) * (a^1.2) Time = y * e^((a - b)/a) * (a^1.2) or maybe Cost = x * e^((a - b)/a) * a Time = y * e^((a - b)/a) * a Where a is the tech level being researched, and b is the median tech level (or a fudge, like median + 10 or whatever you prefer - just not an average, for the reasons posted by Innocent Bystander in the other thread). For newbies at tech level zero, the cost/time of research will be about a third of what it was previously. At the median (or whatever you've set 'b' to), the cost is the same. Going beyond that, it increases. So if the median tech level is 50, and you're trying to research from 99->100 to acheive double the median level, you're paying (relatively to the tech level) more than if you were going from, say, 74->75 to achieve 50% on top of the median. That achieves all your goals and it also stops super rich players from racing way, way ahead of the median all on their own because of the diminishing returns per $. Just like the real world. |
|
Scott (Admin) RJ: Ratan Joyce CO: Ratan Joyce Post Rating: 0 + / - Total Posts: 1175 Karma: 5083 Joined: Jan 13, 2012 |
Posted on Apr 26, 2012 Too late, I've already made the change yesterday to:Cost = x * (level-0.3*lead_tech)^1.2 Time = y * 3^(level-0.3*lead_tech) There was a "typo" in the earlier released equations, time was and is always 3^(factor) and not (factor)^1.2 Let me work on the queue first and see how that'll go. But thanks for the suggestions Nick, if the new eq doesn't work out then maybe I'll go with using exp. |
|
Nwabudike Morgan RJ: CEO Nwabudike Morgan CO: CEO Nwabudike Morgan Post Rating: 15 + / - Total Posts: 108 Karma: 344 Joined: Apr 4, 2012 |
Posted on Apr 26, 2012 You're switching things around in the explanation above. In your research formulae, time is cubic (x^3) and cost is exponential with base 1.2 (1.2^x). Exponential growth quickly outstrips polynomial growth.
|
|
Scott (Admin) RJ: Ratan Joyce CO: Ratan Joyce Post Rating: 0 + / - Total Posts: 1175 Karma: 5083 Joined: Jan 13, 2012 |
Posted on Apr 26, 2012 Yes it does. Again research time is always cubic and never exponential, it's one of those "early morning" mistakes that just got spread around.
|
You need to register or login to post a reply.