what exactly do you mean by complex fit?
i am not sure, what makes a player understand ship fitting.
no need to talk about blaster rifters or other abominations by newbros.
but i have seen many fits by experienced players, that could have been made better, with less ISK.
when i started the game, i would ask for fits in helpchannels, like most newbros. but i came to the conclusion, that a lot of players donât have a good idea about what makes a good fitting, even after years of playing. i mean, if you throw enough SP and ISK against a task, it will somehow work anyways.
but i do think, for almost every task there is a Tech2 Fitting that can do it. ALMOST every task.
there are some things, that need a bit of bling. or even a bit more.
what i am trying to say is, newbros need good reference, to learn what makes a good fit. if you get handed shitfits by your corpmates from day one, as example, and you deem them good fits. well, you will most likely build shitfits yourself and think you got it
1 Like
The sort of fits people arrive at only when they understand all the options available and how they inter-relate.
1 Like
A âgood fitâ has very little to do with the modules being T2 or not, or worse or better. An actual good fit is about understanding the use, situation, player capabilities and preferences and how different modules affect actual gameplay, not just numbers. Itâs therefore no use to create some library of âgood fitsâ unless you very specifically specify the use and situation for each and every one of them.
Case in point being solo fits for for FW players, thereâs no point giving a newbie a super amazing fit for a ship that requires a very active play style with lots of buttons to press at the exact correct time and relies on (pre)kiting. It doesnât matter how good that fit works in theory or in the hands of an experienced pvper, a newbie or someone new to PVP isnât going to make that work.
Then thereâs issues with just basic understanding of whatâs going on. Like people talking about a cloak/MWD fit how the mwd blooms their sig and how thatâs a problem. Not realising that theyâre going to be cloaked while running the MWD cycle, so it doesnât fcking matter.
The REAL problem isnât new players not knowing stuff, thatâs just fair and understandable. The problem is older lazy players still being clueless while pretending to know things and feeding newbies nonsense, and not just on fits. The inevitable guy in the corp whoâs been playing for 15 years (mining) who takes control of every conversation on voice (probably doing the âIâve been a WOW dev for 7 years dontyouknowâ) and creates this aura of knowledge while actually being dumb AF, and he canât stop giving âadviceâ⌠on anything.
3 Likes
That you wonât like the result?
I am just guessing of course, I should have added âI donât think you will likeâ. The reason is because you are coming in with the âAIâ idea, and the results are very different - at least the 2 I can think of. Nope not telling you need to try make the AI to understand the one.
1 Like
yeah, that is absolutely true and it is sad.
those are 90% of players hanging out in help channels for example.
so, would you say it isnât about the fits. it is more about education, right?
how can we adress this matter then?
thatâs pretty much not possible for the entirety of new players i am afraid.
there are tons of guides on youtube that are absolute trash for example.
how should a newbro decide what is worth his time and what isnât?
also, how could we make it clear to veteran miners, that they are not good at the game, just cause they play it for years?
(except ganking them like the Princess and her Court)
it seems to me, that this is something worth thinking about.
i mean, every hunter loves the clueless, overblinged newbro, that is going to lowsec in a 1b astero without using his cloak for example.
but wouldnât it be much better for the game, if there was something like quality control for newbro education?
isnât it much more fun, to fight good players, that give a gf in local after a confrontation?
rather than blapping a someone, that doesnât even know what happened, not to speak of what a conversation is good for?
oh well, i am gonna open a new thread for this topic
Please dont. Thereâs enough of them
1 Like
too late muahahaha
1 Like
I donât think EVE does work that way. A ânewbroâ or âcasualâ who doesnât really understand how his fit works, how the modules and the ship stats and abilities achieved by those fitted modules are playing together, will never be a âgood fightâ but always a victim to a player who does. It just might need 10 seconds more to kill him.
I think an AI can do basic quality checks, yes:
- are rig/module slots unused?
- is there a cheaper/moderately priced faction/deadspace item that can replace a T2 one and free up some PG/CPU to upgrade other items?
- are the weapons supported by the ships boni?
- are too much items of the same type (stacking penalty) used?
- is the targeting range sufficient to even use fitted long-range weapons?
- are weapons with a low application supported by modules that actually allow them to hit properly?
- is a fit generally viable for the specified purpose: âbrawlingâ âkitingâ âsnipingâ âactive repsâ âbufferâ etcâŚ
However, that still doesnât say anything about the ability of the pilot to actually use the fit in the way it is intended.
After all, I encourage you to use the project to learn and get better yourself - both in programming and fitting ships. But donât forget, itâs always better to teach newbros why they fit or combine certain modules and how to build a ship around a certain playstile or purpose rather than âgivingâ them âgood fitsâ. And last but not least: Very successful fits are often âone-trick-poniesâ, and the wins are scored by being very selective in the engagements. It will be very hard for any A.I. to âunderstandâ this.
Good luck!
5 Likes
Per usual, the why is much more important than the what. Copying someoneâs fit (or âbuildâ in other games) but not understanding WHY theyâre doing what theyâre doing even when showing gameplay, doesnât really work. They all go âah see, itâs easyâ but when they then try it themselves it horribly fails and then the excuses come.
You can give a man a fish, etc etc.
1 Like
Sorry but no you wonât be training an AI on fits.
Unless the AI is training on the actual game (i.e. playing it) then it canât learn anything other than the fits, which on their own are pretty useless.
The AI would need to be training on actual fights for starters, and this already blows your scope into the sea.
Instead, as a learning exercise, Iâd make something that does what PYFA does, but maybe in JavaScript so you can run it in locally your browser easily. Store all the item names/ids in JSON and pull with Ajax calls as required, and this eliminates the need for a database (or indeed any back end at all).
Then, once you have that, start pulling in player skill data, so all your values match with the playerâs bonuses (i.e. Missile Specialisation buff) etc.
AI isnât the silver bullet everyone things it is.
the idea was, for example, to feed it with letâs say 20 brawling fits for a ship. then it would know, what a brawling fit for this ship looks like.
then give it some parameters like ISK and ehp for example, that the user can input.
then it would generate a fit, based on the database of 20 fittings and the parameters.
i am aware that it will not magically create perfect fittings
actually, the more i think about it, the less useful it looks to me xD
1 Like
The problem is, that there are âideas behind fitsâ that an AI canât comprehend.
For example, for the Legion alone I have saved like 50 fits, half of them are brawling fits. But most of them are for different scenarios. Smallscale, Covert, Fleet, PvE, Neutralizing, HAMs, Lasers⌠Doesnât mean any of them are bad, they are all pretty good if I use them in the right way. How should your AI find out which one to recommend?
Now, to be a bit constructive: What you can do is to grab very successful fits from zKillboard for example. Not âcreate own ones by the AI based on âstatsââ.
Lets say, a user asks for a good fit for the âHarbingerâ.
Step 1: Your Program checks the alltime-top page for the Harbinger, which would be this one: Harbinger | Topalltime | Ship | zKillboard
It will find a list of pilots with MANY Kills in a Harbinger. That doesnât nessessarily mean that those people are âExpertsâ, but they are very active and they probably know the ship by now.
Step 2: Your Program checks the LOSSES pages of lets say the Top10 Harbinger Pilots.
For the Top1 Harbinger Pilot âVirus Knightâ, this would be this one: Virus Knight | Losses | Character | zKillboard
Step 3: From all losses of a pilot, your program ignores all losses that are not a Harbinger.
Repeat until you have X Harbinger losses from a pilot. X can be 10, can be 20⌠donât go too high here.
Step 4: Your Program now checks all those Harbinger losses for the modules and rigs used. For example, if 90% of the destroyed Harbingers you find use Heavy Pulse or Heavy Beam Lasers, you can be pretty sure that those Weapons are probably part of a very good fit. Because if Focused Medium Pulse or Quad Light Beam Lasers would be similar good in performance, you would see them more often be used by the most active Harbinger Users.
That means, every module used gets a âscoreâ.
Step 5: Your Progam now âcreatesâ a new fit, beginning with the highest scored modules. And fills all slots with modules ranked by score, until you run out of CPU/PG. If so, check for moderately priced faction/deadspace/storyline items which would free CPU/PG. If not possible, âdowngradeâ the lowest scored T2 item to a more fitting-efficient Meta4 item. If still not enough, check for alternatives that push the same attribute, but require a lot less CPU/PG. For example a Coating instead of a Membrane. A Small Cap Booster instead of a Medium one⌠At this point, your fitting should definitely work.
In the end, for the Harbinger you will come out with something like that:
- 50MN Quad Lif MWD
- some 1600er Plate
- T1 Trimark Rigs
- 1-2 Energized Multispectrum Membranes
- Damage Control
- Heavy Pulse or Heavy Beam Lasers
- Warp Disruptor or Scrambler
- Stasis Webifier
- Capacitor Booster
âŚ
Build that into a Fit, and it will look like this:
And voila, while this fit might not be the ultimate best fit for every imaginable situation, its absolutely solid for many situations and probably a lot better than what 90% of the newbros scrap together.
2 Likes
The person need to stop believing the â â â â people fill them with everywhere. It is a mental development needed that most humans are incapable of.
For some reason humans in most cases, choose to believe the most stupid of the answers they are given - and then expect people who are not stupid enough to fall for it, to take the time to explain to them why they are wrong. If the smart person refuses to waste their time, the stupid person goes on with the belief it is now a proven fact that the stupid answer was correct.
Humans also tend to think the first thing they try is good or great. Without having anything to compare with at all. In the case of fits this can be handled by giving them an actual good fit early on, CCP has failed hard there, letting really bad fits into the community fits. Mikeâs bus helps some.
You also need some base experience to be able to understand a subject. So you need people to simply open the fitting window, and put some modules on. And not because Aura locks up the GUI and donât let them play, but because âThe enemy here is known to fire Thermal missiles at you, so you want to to fit this Shield Hardener you just got, when you activate it, it consumes cap and increases your resistance to thermal damage, so use it during the fightâ.
Back to the fit evaluator - simply telling someone good or bad, takes a billion tries to train them, that works for an âAIâ a human dies of old age. If you want them to learn to actually fit, you need to give them basic knowledge, something a computer is incapable of using, but a human can understand context and actual connect thermal resist with thermal damage on other than the words.
1 Like
At least current ones and once they can we will have much greater problems to worry about. 
1 Like
Go for it! Iâve not read all the comments here but there are arguments on both sides. I am Head of Software Engineering and the biggest lesson Iâve had when wanting to build âcool stuffâ is first interview your target market, then build what they need.
It sounds like youâre going straight to thinking you know what they need. An idea or path you are taking might or might not be the solution. Or it might be a nice to have. If you go down that route without knowing your clients then you might build something they think is cool but never use.
Talk to newbros and ask them what they are struggling with when it comes to fittings. The aggregate that research and come up with a product roadmap.. This part of the work is typically done by Product Owners and Business Analysts. As a developer itâs a good skill to have and really expands on the critical thinking when you are actually building an app and prioritising what gets done first.
Second lesson I had was donât build perfect. Build the first thing that is needed, then release, and repeat. Youâll end up in and endless cycle of improving features that get you less value than the effort put in. Once your base is done and your users ask for improvements, then you do them as you know exactly what they want improved.
Looking forward to see what you come up with.
1 Like
thank you for this elaborate post
thank you very much