Glorious Alpha Two Testers!
Phase I of Alpha Two testing will occur on weekends. Each weekend is scheduled to start on Fridays at 10 AM PT and end on Sundays at 10 PM PT. Find out more here.
Check out Alpha Two Announcements here to see the latest Alpha Two news and update notes.
Our quickest Alpha Two updates are in Discord. Testers with Alpha Two access can chat in Alpha Two channels by connecting your Discord and Intrepid accounts here.
Phase I of Alpha Two testing will occur on weekends. Each weekend is scheduled to start on Fridays at 10 AM PT and end on Sundays at 10 PM PT. Find out more here.
Check out Alpha Two Announcements here to see the latest Alpha Two news and update notes.
Our quickest Alpha Two updates are in Discord. Testers with Alpha Two access can chat in Alpha Two channels by connecting your Discord and Intrepid accounts here.
Comments
He didn't go to the Wikipedia link
Look who's trimming context out of quotes now.
Neither.
It is entirely possible for developers to adjust more than one tuning knob when balancing their games. It happens all the time. I'll offer again. Would you like me to find you some patch notes that show this is possible?
I don't need to bother going to the last two wikipedia links - I know enough about both of them to know there is no point in me bothering to read any more.
To the first one, if you honestly think a startup MMO company is going to put machine learning to use on class balance this decade, I'm not sure what to say to you.
To the second, this would be about as slow as measuring variance individually per ability - or the hiring of someone specifically to perform this one task. When you consider that the entire combat team of an MMO the size Ashes is expected to be (including balance and new content, and often the coding as well) is exactly one person, this is a tall order getting someone in just to run figures on balance.
The actual problem with Bayesian probability though, is that it requires a pre-existing assumption about a given event, and if that pre-existing assumption is wrong, things go badly.
At least, that is my understanding of it.
You stated it was impossible. Not expensive or challenging logistically.
The math involved in solving a problem can remind someone of the math involved in solving machine learning problems without requiring that Intrepid has to do any ML.
Also, ML isn't particularly difficult, plenty of open source projects like tensorflow would be super helpful for stuff like class balance, and wouldn't take a lot of dev effort (as someone who is in the field).
Also, and again, none of this is relevant. This variance model this was an off-the-cuff answer that was supposed to be an example, not an actual proposal.
You are getting the points of this discussion mixed up. It's as if you are just skim reading my posts looking for something you think may be a "gotcha", without actually paying attention to it at all.
I said it is impossible to balance a stun with no RNG in an open world MMO. That is what I said is impossible, and is all I have said is impossible.
What is being discussed in the post you quoted is not that - it is a different point.
This goes back to what I said earlier - dense or deceitful?
Edit to add, it may be a few hours before I can reply.
Fair point, more clarification is always good.
First, I never said it is impossible to balance using variance - I said it is too time consuming.
Adding in Bayesian stats to it isn't going to speed that up much.
Interesting citation of the Bayesian inference as a possible method of game systems balancing, but i believe you are aware of the limitations prior of something abstract like "balance" and its possible implications when translated into game design would be right?
Bayesian inference would usually be used for something less abstract in terms of prior and more palpable like player engagement and retention, like in this article i can give as an example. https://home.cs.colorado.edu/~mozer/Research/Selected Publications/reprints/KhajahRoadsLindseyLiuMozer2016.pdf
Aren't we all sinners?
You can split that out into it's component pieces and little baby priors like 'i am x% confident that a player with y skill will win' etc. This is all half-baked atm because it's 1:55am, but I'm confident that you could get there
Again, neither.
Beau's post about Bayesian Inferencing was an example of how the balancing of a stun with no RNG, is infact possible. But admittedly, I've taken some cheap jabs after you condescendingly asked me if I'd heard of blocking or avoidance. That wasn't cool of me and I appoligze.
Sorry for being a jerk. I'll stop with the jabs.
The limitations i was refering to in my comment for an abstract prior like "balance" in game systems, would be not only the amount of non-abstract variables in the builds, like gear, stat allocation, classes(64), skills, augments and etc. (using your pre-existing assumption as a base: I am x% confident that y character build will defeat z character build)(using ashes as example).
But also abstract variables like playstyle, encounter situation or game knowledge.
btw i never before thought of the possibility of using something like Bayesian inference as a tool for game system balancing because of the abstract prior limitations of its setting, your citation gave me something to brainstorm about, like making Priors less abstract and solid enough to be used in this concept.
I could dissert further but the horary prevents me. I wish you a good night.
Aren't we all sinners?
His post was a way in which he believes using an observed variance to assist in general balancing may work.
That was never the core of the reason I said you cant balance a hard CC with no RNG in a combat system based on RNG.
I know enough about it to know it isnt something I want to get in to a specific discussion on with someone that does know about it, but also enough to know this isnt an immediately applicable situation to apply it to.
The variables that would need to be accounted for are near limitless.
False dichotomy. Your assuming hard cc that isn't technical, have universal break out mechanics, or any set up, should exist in the first place.
How does rng balance stun any better than if it was 100%
I am unsure what you mean by technical, I am not assuming anything about break out mechanics or set up existing or not as neither of these would alter my argument either way,
...
RNG as an opposed check, not just RNG. It is important to not just isolate the part you don't like from a system.
The point I am making here is that Intrepid need to somewhat even out players of slightly different skill levels some. RNG is the only real way to do that in an open world PvP setting - and that RNG is most important on hard CC's, as they are often the key differing ability in the arsenal between an ok player and a good player.
Adding RNG means that good player is still going to be able to win most of the time using their standard approach - but every so often that not-quite-so-good player may get lucky and get a resist on a key CC against them, turning the tide of the battle.
The fact that both players have some control over that CC;s chance to land means neither player can really blame the system or the RNG- they can only really blame their build.
The thing I find amusing here is that if for some reason that soo-called good player started losing more than they were winning due to RNG on hard CC, then they weren't good players at all, they were just one trick ponies that had their trick taken from them.
There are a number of people in this thread (and the previous threads) that seem to me to fit in to this category.
the ONLY way to do that? JustVine listed a bunch of other (imo better) ways to manage it than RNG stat checks.
This is literally the definition of a comeback mechanic, which I do appreciate myself, but we need to remember that comeback mechanics are antithetical to your point in the directly preceding paragraph "...differing ability in the arsenal between an ok player and a good player." This makes your argument a paradox: RNG separates the good from the bad, but also blurs the difference between the good and the bad. Personally, I think RNG works great as a comeback mechanic, but I get the feeling what you're really trying to say is "RNG management is a skill too", yes?
I do want builds to be relevant, but you might be the only person I know who thinks CC can ONLY be controlled by RNG management. People are telling you alternatives exist and they work well, does that make sense?
Oh boi, I wish I wasn't a one trick pony who got rekt by RNG checks. Those players are too skilled for me.
In context, yes.
That context is that Intrepid need to slightly compress the player skill paradigm.
In many games, the idea is that a small amount of additional player skill over an opponent will almost always result in a win. That paradigm in Ashes will be a death knell for the game.
By compressing the effect of player skill, more players are more evenly matched to more players, making PvP in the game better.
The suggestions offered are all fine if the idea is to separate players by skill even further. If you are indeed not a one trick pony, then RNG on CC won't bother you at all - especially if there is player input to the outcome.
If you are indeed a good player, you should be able to overcome missing a CC. If you do indeed want good PvP, then you want as many people being a good match for you, rather than an easy win.
It always becomes meta at the top because it's free damage. RNG is just wishfully hoping your design goal will happen or worse is an illusion designed to trick the player into thinking 'it could be' rather than just building better combat.
The build 'skill' it 'adds' to the game is even more of an illusion. If it meaningfully decreases stun chance it becomes copy paste mandatory. Always. If it doesn't no one uses it. It restricts builds for some wishful prayers to the RNG gods. Ugly lazy design.
This is true if players have no say in the RNG.
Hand players a say in it, and players build characters around having a high chance of CC - at the cost of other aspects of their character.
Far from lazy design. It is literally a lot more work to do it - which is the opposite of lazy.
I believe JustVine is arguing that the concept is lazy. Beau does a really good job explaining this in the blog post you keep refusing to read, which is a shame.
If he wants to make an argument on that point, we have a discussion right here. I've asked him several times to do so - it is on him if he refuses, not me.
This is on my hall-of-fame for "getting people to accept absurd positions". Say that millions of fights between players of different builds are happening all of the time. Some of those fights will be between a player who builds to have a cc with no-nrg and a player who builds to not have one. When you adjust any of the properties of the high-impact cc, then on a statistical level when you look at millions of fights, the win rate shifts. If you increase the cast time, then maybe now the opponents were able to get one more ability off or one more heal in before the CC landed.
If you decrease the duration, then now the opponents are able to act sooner and will take less damage while under the effects of CC. If this crosses particular thresholds, then now it disables some discrete combos like how a 3 second CC might guarantee a 2.5 second cast nuke, but a 2-second CC would give the defender time to act.
All of this stuff changes the winrate. It changes the probability that the worse player will be able to win.
If you want to have an argument on that point, you can read the article. I've shown you the link several times. If you refuse, that on you, not me. You're capable of clicking links.
This doesn't work in an open PvP MMO game.
I'm more than willing to bet my house, farm, animals, and blacksmithing career in a game of League of Legends drunk than in Super Mario Party sober...