Looking for expert fitters/mathematicians


(Monkala Demeir) #1

Hi All,

I’m presently working on a project to develop a rough “value multiplier” guide for rudimentary fitting. Effectively fitting in eve is a nightmare to start with and to initially get your head around and there is presently (to the best of my knowledge) no real benchmarking system to gauge whether or not you’re on the right track when building your first fits and thus, myself and few others have begun work on a rough output-value tool.

Effectively this is a baseline calculation to work out the tradeoff of any given ship, between output stats and the overall cost.

Presently we are in need of a bit of consulting. What we are looking to do is assign a valuation mechanism that can create a degree of parity between the major ship-output stats. These being:
DPS, EHP, Average Active Tank, NAV-m/s, Range and number of targets. What we want is to create a rough matrix in which a value-multiplier is assigned to each stat, depending on the primary use of the ship.

As an example
Ship X is going to be used as a Fleet PVP ship

Fleet PVP ship contains the following multipliers: DPS20=score, EHP0.8=Score, Average Tank * 15=score etc etc.

If this project is something that would be of interest please feel free to mail me in-game or reply to this thread. This description may be a little convoluted and so I am more than happy to discuss this project further with anyone interested.

Lastly, there will be compensation for anyone able to help with areas of the project and so I am more than happy to discuss payment if someone is able to help with this current aspect of the project.

Thanks so much all,

Monk


(QuakeGod) #2

An “effective” fit is entirely subjective. What I think is effective for a given purpose may not necessarily be what you think is an effective fit for that given purpose, but it works nonetheless. You also have to take into consideration a players skills and experience. An “effective” fit will not compensate for lack of skill and experience.


(Monkala Demeir) #3

Hey mate,

Thanks for the reply.

Totally understand that this is not a catch-all when it comes to actual fitting and people will need to fit to their own skill level. Nor is this an exercise to determine that “x ship will make you a flying god!” but more just to benchmark various ships against each other.

Thanks for the input mate, will definitely use your thoughts as a caveat when applying the valuation mechanism!


(Shailagh Rose) #4

Im, like, really smart at math


(Jenne Wain) #5

This is an interesting thought experiment, if nothing else. I guess the way I see this is one of your fundamental challenges is going to be defining what “value” is, quantitatively, in a way that is useful.

First, there is the situation where each hull has a theoretical “ideal fit” with a given selection of modules that work with the ships attributes to maximize benefits within constraints. Odds are that would be really expensive, but take that as a starting point.

But not so fast, because this is where quantitative analysis is going to begin to fall into qualitative subjectivity. Say, I’ve got X cap and Y PG to be split between tank and DPS and Z ISK to spend on it all. Which is more important- tank, DPS, or cost/benefit ratio? For example, what would be ideal for an Orca vs. what would be idea for a Oracle, and how can a noob Alpha and vet Omega alike figure what works for them. Thus, you can get a sliding scale of splitting those resources between attributes of the final fit which can be considered subjectively. That’s going to be a challenge to deal with and present in a useful way. Given that, perhaps a one number rating might not be as helpful as giving a rating across different dimensions or attributes. That may have a lot of value because it helps people understand how to make trade-offs when working on fits in a more general way.

Lastly, there are the extreme number of variables that will radically impact this in the real world, again, depending on how you define value in the algorithm. Market price fluctuation, skills, implants, fleet role, etc. can all have a major impact on what “value” a given fit has.

I’m not mentioning any of this to be discouraging- this is an intriguing idea and concept. Just trying to float out some additional considerations to try to be helpful.


(Jenne Wain) #6

One other question I have is are you looking to create something that is a tool for benchmarking different fits or to provide a tool to help optimize different fits to achieve a given value (probably against multiple evaluation criteria)?

Put another way, are you just looking to grade fits or to provide a tool for evaluating and optimizing trade-offs?


(Eternus8lux8lucis) #7

EFT used to have this feature in one of its upgrades before it died. It was for buffer tanks for null fits only though but same idea.

But yes what you will run into are functionality and homogeneous issues over time. Whatever your formula is and however successful this final programs distribution ends up being itll just enhance the ideology of “fit like this or nothing” mentality.

Good luck either way though as there are raw statistical data dump calculations you can create that would spit out such a result quite easily. Though we have a lot of different ships and modules the actual number of raw variables is still quite low.


(Estuary Algaert) #8

In my earlier days of play, I looked at doing the same…

Part of the issue you war going to run into is exactly what you stated in the first sentence. There are two skill levels, your personal skill, and your character skills. Both of these make and break any degree of comparison between ships… even comparisons within the same ship.

Sure, there are relative comparisons within hulls that can be made, particularly around the performance/isk ratios. But frankly, once you start comparing one hull to another that math gets rather crazy.

Think of ships more like tools for construction, something will always do X job better than something else. I can use pliers to remove nails… but there are better tools. EvE ships work the same, pick the right tool for the job you need it to do, inside that scope, you can do the comparisons you want… but that is still a large scope.

Also, anything beyond solo and small gang, those ship differences start to matter less and less. When you can have all your friends basically alpha something off the field… well… it no longer matters that ship X does 100 more dps and has 1000 more EHP than ship Y. They will both work in the setting you need them in, now it is just cost.


(Monkala Demeir) #9

This is an awesome reply thank you!

I’ll try to go through each point with my general thoughts.

  1. As far as costing goes, the final part of the formula we’re looking at will include allowing for the cost tradeoff. This is most likely going to be through a division of the final derived score by a function of the cost ( ie this may be X score / (cost/1000000). That, however, is going to undoubtedly return mixed results as ship pricing doesn’t quite scale that easily. My current thinking is allowing for the cost by implementing an average cost for each hull across the database we will eventually build. In this sense, the final score could end up being - X score/ 1+(current ship price/ average price of that hull across the database). Take the following example. Monks PVP Punisher arrives at a score of 45.9 and is priced at $4,700,000. In this scenario we could look at the average price across all t1 frigates in the database coming to lets say $6,000,000. We would then go 45.9/(1+(4,700,000/6,000,000)). From there we could apply a multiplier to get a slightly “nicer” looking index but that is one way I’ve thought about looking at the price trade-off.

  2. If i understand your point correctly this is where we start to look at the actual tradeoff between output and value. To your point, I believe looking at the exact trade-off between a monetary value and a specific metric may start to get way too intense, doable, but probably a little deeper than I’m looking to go in this infancy stage. Currently, as far as the final derived score and application goes I will probably look at developing some kind of (subjective) matrix.

Look at the following:
If the purpose of the ship is to be at small-gang PVP fit. @Estuary_Algaert, this may also apply to your statement as well. Continuing with this line of thinking.
Under our team’s parameter we may find that, when associating a given ship with small-gang PVP, the following multipliers apply: DPS50 , EHP0.9, Average Active Tank *40, NAV-m/s *300, Range *0.7 and number of targets *1000. If that was what we ended up with, the ships stats would be multiplied by each value and a raw score will be derived from whatever that final value is.

Below is the initial stages of the rough table I think we need to create. As you can see there are the usage fields for the ship on the left. Please note these are just initial possibilities, we will hopefully broaden these uses out later to quite an extensive list.

Multipliers
Usage DPS EHP Average Tank NAV Range Targeting
PVP/Solo
PVP/Fleet
PVP/Sniper
PVP/Tank
PVE/Solo
PVE/Fleet
PVE/Sniper
PVE/Tank

  1. The third point i’d love to go in to is your identification of the number of different variables.
  • Market price fluctuations will theoretically be built into the price formula at any one time. For example, say you’re tossing up between two fits that you have built, a vindicator and a malestrom. IN this example, say the vindicator produces a better overall score, when cost is accounted for between the two ships, in December the vindicator comes out with a better score-cost ratio. January rolls around and all of a sudden vindicators spike in price. Whilst the vindicator still maintains the same output score, it now arrives at a lower efficiency score as the price has increased dramatically, now when looking at pricing the maelstrom now comes out with a better efficiency ratio and so the user may now decided to move to that ship instead. This is one example of the price fluctuation formula and the efficiency trade-off

-Skills will have to be set to a standard minimum for a viable benchmark to be arrived at. Right now we use osium as our primary ship building tool and so for ease sake at the moment we may continue to use the default “all skills at V” to make life easier BUT down the line we may be able to create a program that takes into account current skill levels when building your ship and arriving at a score. This will most likely come in a later phase of the project.

-Implants I see as a cost-side factor. If the ship you build (assuming all skills @ level V) still requires 3 different implants to use, these 3 implants will be added to the cost of the ship as they are a necessity, we may allow for the full value of these implants or just a determined fraction but regardless i presently see this as another cost-influence

  • Fleet role is the final element to this which I really like and will really give our multiplier matrix a whole other dimension. If you look above you’ll see that right now we have it segmented quite broadly (pvp solo, pvp fleet etc etc) What we hope to eventually do is broaden out the matrix to potentially take into consideration all possible ship uses eg PVP Fleet damage. In that scenario, DPS would get the highest weighting or a higher weighting than a ship designed for PVP Fleet Logi.

I’m not sure how comprehensible that whole response was, i’m still fleshing out my thinking and as a continuing caveat this project is about 5 days into its infancy stage so we’re still right at the beginning, i’m going to undoubtedly have missed a few key elements and this will take a while to flesh out (as all good things do) but is a project i’m willing to stick to ! :slight_smile:

Thanks so much again for your response, would love your ongoing thoughts and will continue to post ongoing updates and what we’ve arrived at


(Monkala Demeir) #10

Right now all, I know there has been a lot of conceptual talk but what I love to start trying to do (and would love some help) is begin to get a crowd-sourced idea for what multipliers can be applied to each stat using the rudimentary matrix I created.

Usage DPS EHP Average Tank NAV Range Targeting
PVP/Solo
PVP/Fleet
PVP/Sniper
PVP/Tank
PVE/Solo
PVE/Fleet
PVE/Sniper
PVE/Tank

Given this matrix, would anyone be able to think of a way to begin to identify what multiplier should be assigned to each “cell”. Right now we have 6 primary stats and 8 possible ship usage-classifications. Would love to get in touch which an expert fitter who could begin to say " ok lets look at pvp/solo/sniper. In that scenario, the weighting will favor DPS and range" That way we can start to apply multipliers and start a few trial runs applying it to ships.

Any continued thoughts or help on this is always appreciated and thanks so much all for joining the discussion!


(Jenne Wain) #11

Wow, very cool response, and thanks for all the detail. I definitely have some thoughts and ideas about the points that you mention, but before I go there, I’m going to be a bit of a jerk and pull way back into some more conceptual talk :rofl:

There is a lot of really good theorizing here for what the technical solutions can or could be, but if you’ll indulge me, I’d like to have a quick go-round from the opposite end of the project- the user. The reason that I’d like to push on this a little bit is because your goals will have an impact on some of my suggestions to the items that you raise. So, I’ll put on my “product champion” hat and ask some of those annoying questions.

Who is the user?
Who are you targeting this tool toward? Is this something that is a general purpose tool for anybody, or is this more biased towards new users with limited fitting knowledge and fitting flexibility, or veteran users who have more ability to make tradeoffs due to experience, skills, and wealth? If you had to pick one (the noob or the vet), who is more important?

What is the pain that you are trying to solve?
This is probably the most challenging question and the one that will have the biggest impact on the direction of the project. This is likely where choices will have to be made as well. For example, are you:

  • Trying to give users a starting point with a fit that is optimized to their role and skills (based on your skills, heres a template for you to tweak for this hull and role-think “Wizard”)?

  • Trying to let users take an existing fitting and see how they can improve or optimize is based on their role and sills (based on your skills and the hull attributes, you should change those passive resists to active-think “Optimizer”)?

  • Trying to teach how module tradeoffs work in a more easy-to-understand way while fitting, so users can more transparently see how adding a given module for one attribute affects others (beyond raw stats like DPS or EHP) and show what kind of role a given fit is most appropriate to (i.e. those blasters won’t work on a sniper, so you’re now a brawler and you better add some tanks as well- think “Comparator”).

  • Something else?

What is the context of use for the user?
I see that you mention Osmium above (I’m a PYFA girl and that team might be an interesting one to reach out to, but nonetheless…). But what is the environment and workflow that you imagine the users operating in? For example, when I do a fitting, here is my general process:

  1. Think about what I am trying to do and what role I am going after
  2. Identify the skills that I have (or will need) to fulfill that role
  3. Search for ships that generally align with that role and my skills
  4. Search for fitting ideas online, wade through the silly ones, look for patterns, and find something that looks like a good starting point,
  5. Drop that into PYFA (and get depressed that I’m not as cool as the people online and the fit isn’t as good for me as it is for them)
  6. Assess what I see as trade-offs that I can make (i.e. I can drop that mag stab to add an AR)
  7. “poke-and-hope” with different modules working with the results in PYFA since I can see all of the attributes at work and changing (*BUT-I have experience and know how to do that fairly well, noobs certainly will not)
  8. Go to sleep
  9. Wake up and think, “Huh- maybe I can live without that AR if I add a better AB for more speed for kiting and escape…”
  10. Tweak the fit again. Get depressed again.
  11. Identify some skills and better modules that I can train for/buy to make things better.
  12. Tweak the fit again.
  13. Build/buy everything
  14. Take it out and try it. Learn what works and doesn’t (paper vs. reality)
  15. Tweak the fit again, swap modules, test revised fit.
  16. Be (relatively) happy with the result

Now, I’m sure people will read this and think I’m a complete moron, but I suspect that this is not an uncommon sequence for a lot of new players. There are a lot of pain points in that sequence, so what do you see as opportunities to create “pain killers”?

Will this be a stand-alone tool, or something that can integrate with another tool (Osmium, PYFA, etc.)

How will you present information to the user?
Another hard one. One-number ratings are attractive because they are simple, but they are usually TOO simple. For example, if I look at a 1000W loudspeaker that doesn’t actually tell me anything useful beyond how much power I can put into it. What I really want to know is how loud does it get? To know that I also need to know impedance, sensitivity, frequency response, etc. Same with things like “horsepower” -cool, but what’s its 0-60 time? Towing capacity? One-number metrics are easy but frequently not helpful.

At the same time, being too atomic in your metrics gets messy and confusing- see the current situation with other fitting tools. I can see all the specs and stats, but it’s difficult for the new user to see correlation and tradeoff relationships. It can become “data” rather than “information”.

I think again, this gets to a fundamental question of what information you want to convey to the user- do you want to tell a user “A is better than B” or to help them understand why and how to improve it?

What is your MVP?
What is the least you can do that will provide a significant enough benefit to the user? How will that be built upon in later versions? What will you consciously not add? I like that you are looking at starting with a set of basic role archetypes as constraints for the project, but what other constraints do you want to add so you aren’t boiling the ocean on the first stab at this?

OK, that’s probably enough- and I’m sorry to answer your questions with a bunch of other questions, but I do hope that this is helpful! I’m not trying to troll you or be pessimistic or anything like that, just trying to add some additional considerations and framing for the discussion that can inform the technical solution.


(Ramona McCandless) #12

If you can solve the quandry, the tabletop gaming world will be interested in how it works for the purposes of Points based games.


(Estuary Algaert) #13

Something to keep in mind with ship costing.

For PvP, ship cost is a larger factor than PvE. Generally constantly losing and replacing your PvE fits is not desirable whereas you generally expect to lose the PvP ship. PvE fits on the other hand have operating costs over time, ammo, charges… forgetting your drones… this can weigh heavily on your pocketbook over time. Sure, ship X’s rating is really high, but that ammo selected is pricy.

This brings me to something else to think about, PvE. It is far more predictable than PvP, you basically know what is going to happen, what the opposition is, ships, damage types, etc… Great, ship X is awesome vs Y kind of rats but your are doing content against Z now.

One final note for PvP and PvE, where you PvX also changes desirability of ship. A mega awesome ship of ultimate death is rather SOL if you drop into WH space with no probes (or without probe support) to find the exists, or no ability to refit, specially after the one you came thru closes :slight_smile:.


(EOH Lummox) #14

When considering fleet effectiveness you would also need to account for tracking, signature and resists to allow your calculations to be meaningful.

As a real game PVP example AB armor HACs were a direct counter to BS fleets due to their combo of speed and signature combined with resists for remote rep. A BC fleet would have had better ‘value’ without those being considered but would get stomped by the BS fleet.

Equally a real game PVE example would be using T3Cs for DED plexes instead of BS as their sig and speed allows them to sig tank so much dps that they can tank more incoming dps than a BS with 4 times the active tank numbers can do.


(mkint) #15

What I assume you are trying to do is make an app where you can put in your loadout and your use scenario, and it’ll automatically rate your loadout. I’m gonna say that’s not a thing that’s possible, at least not in any meaningful way. What might be more interesting is gather the primary characteristics of each ship, namely tank, DPS, speed, agility, ewar, etc, figuring out the absolute theoretical max on each. Make your app output a radar chart showing the performance of their loadout relative to a theoretical max.

edit: On the same chart, maybe have a second layer showing a statistical median score for each axis extracted from killboards, so that they know what’s “normal.”


(Old Pervert) #16

I haven’t read the rest of the thread so I expect this has already been said, but you cannot arbitrarily assign an importance to tank/dps/speed/etc

Ships are fitted with the intent to exploit a weakness in other fits. Be they fleet, solo, or gang.

It may be a nano doctrine, which means it doesn’t need (or can’t have) a lot of tank. Or it could be a triple plated armor doctrine (dual plate baddons tank very nicely for their price).

Or it could be a cruiser doctrine, T3Cs vs arty fleets and all you’re waiting for is GFs in local.

Then there’s your gila fleets, where some dude in a cepter gets to feel like a god whose warp disruptor also makes ships disappear into a cloud of drones.

Or combat cepter fleets, whose #1 most important stat is a sub 2 second align time. DPS, tank, etc, are all secondary to that mobility.

The list goes on and on. Many doctrines are judged by the same attributes, but those attributes are determined by what you fight not what you fight with.

Edit:
Reading the post by mkint above me, that’s actually a fantastic idea. You could try and get a dump of fits from O.Smium for a list of theoreticals, once you validate it’s a fit that actually fits. Many of those fits are insanely stupid, but you can always count on some dickhead to officer fit his t1 destroyer to say “hey look I can get XYZ out of it” there.


(mkint) #17

I’m thinking about my idea a little more, so some clarifications.

-First, a reasonable limit on what the “max” performance might be… say, no officer mods, or above a certain price point, or no mods that don’t hit a certain limit of units traded. Some methodology to avoid obscure mods, but still leave room for lolfits.
-Still have tactical roles to select from, but what will matter here are which attributes will show as axes on the radar map. For example, in most battleships, there aren’t really any roles where sig radius matters, so don’t include it by default on the radar graph. An ewar support ship might have the axis count numbers of ewar mods where any other role might not. Look to previous posts in this thread for more examples.
-The median layer might require some curating of loadouts and roles. You probably can’t just scrape that data from some other database. Maybe use information from published fleet doctrines, or killmails from known engagements.
-the radar graph should have minimums to hit, and when they are not hit, the axis lights up red. For example, a battleship meant to be sniping but can only hit out to 70km would light up red on the range axis. Incursion fleets typically want a certain amount of ehp. Logi needs to put out a certain amount of reps/s. lvl 4 missioner needs a certain ehp/s. Published fleet doctrines should cover most of that.
-have default roles built in, but also have an interface where an FC can put together his own requirements and distribute it to his fleet. I always thought the way EFT dealt with copy/paste was a great way to distribute information.

I might be pushing a little feature creep on you, but a single score isn’t particularly useful unless it’s clear where that score comes from. What is far more useful is knowing which attributes matter for which roles, and where those attributes need to be in order to be functional. There might be a better graph out there than the radar chart (I’m not exactly an expert on varieties of charts) but I can’t think of any that would convey importance as intuitively.


(system) #18

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.