Poll
Question: Should we adjust the RNG to represent THC probability more accurately?
Yes - 16 (43.2%)
No - 21 (56.8%)
Total Voters: 37

Pages: [1] 2 3 4   Go Down
Print
Author Topic: Design Topic #2: RNG  (Read 2286 times)
Vince
Developer

Posts: 7949



View Profile
« on: May 29, 2018, 08:06:04 am »

"My THC was 70% but I missed 3 times in a row, which can only mean one thing – the game is horribly broken" is one of the most popular complaints, so let’s talk about it. Let’s start with what 70% THC really means. If you attack long enough, you will reliably hit 70 out of 100 times. It does NOT mean that you’ll reliably hit 2 out 3 times, but that’s what many players expect.

Overall, the phenomenon of player’s expectations vs actual probabilities is well documented:

https://www.pcgamer.com/the-gollop-chamber-the-good-the-bad-and-the-ugly-of-rng/

Quote
I now understand that human beings are not very good at evaluating probabilities. In particular when an RNG generates repeated sequences a human will cry foul. For a human, randomness usually means ‘evenly distributed without any detectable pattern or repetition’. This is basically how random numbers are manipulated in many games to meet player’s expectations. One poor result immediately results in a bias towards a better result.

Players without any board game or pen-and-paper roleplaying game experience tend to be a lot more hostile to the explicit use of RNG in video games. They do respond well, however, to the more subtle psychological manipulations of randomness which developers and publishers employ these days.

https://www.nexusmods.com/xcom2/mods/482

Quote
Ever do all the work getting your soldiers into position, flanking the enemy, making all the right moves to set up the perfect shot, one your soldiers just can't miss then BAM, that 95% chance to hit misses wide? But no big deal, you planned for this and still have another shot, except, holy crap the next 90% shot also misses! You played perfectly, made all the right decisions but now you are out of actions, your turn ends and the enemy proceeds to wipe your entire squad on their turn. That's it, mission failed, all that work for nothing. How UNFAIR does that feel?

Well then this mod is for you! To put it simply, it aims to make the Random Number Generator and the resulting combat rolls "FEEL" more fair. You're going to miss those 95% shots less often, about as much as you'd naturally expect. Double-extremely-low-chance misses should come up so rarely that you hopefully won't be as upset by them. Basically this mod aims to change the RNG to make it feel and play much more FUN!

REASONING

OK, first of all I'll address the obvious, the RNG in XCOM is already fair, it's been proven to be mathematically accurate many times.  So why this mod and why this name?

For many people who have played this game and previous in the serious, you quickly realize that despite the mathematical assurances, it sure doesn't FEEL fair. Missing a 90% shot twice in a row, is like a punch to the gut, especially after all the work maneuvering to set it up in the first place. And depending on the difficulty level (especially in this sequel XCOM 2, which seems more punishing), what those key misses might mean to the missions success/failure. Yeah sure, it's technically accurate, it's "Good" mathematics and statistics, but it's also BAD game design.

The solution seems to be simple: rig the RNG to deliver what the player expects (or at least avoid what everyone hates – missing 3 times in a row despite seemingly high THC), so I have two questions for you:

1)   Should we rig the RNG to meet players’ expectations?
2)   If yes, how? Meaning what should we aim it? What outcomes should never ever happen when your THC is 70-80%?

Keep in mind that both your party and the enemies will use the same system. Remember that awesome turn when your enemy missed you 3 times in a row? Well, if you won’t be able to miss 3 times in a row, neither would your enemy.  Choose wisely.

While you’re thinking, here is how our RNG works. It draws numbers like cards from a deck, meaning you can’t draw the same card twice until the deck is out of cards and reshuffled. We round up, so if you draw 17, for example, you cannot get numbers 11 to 20 until the deck is reshuffled. This approach ensures that if your THC is 70%, you’ll miss 3 times and hit 7 times. If it's 63% though, you're not guaranteed to get 63 out of 100. Each 10 rolls you'll get 6 guaranteed hits, 3 guaranteed misses and 1 can go either way.

Ideally, your misses would be spread out evenly but as bad luck would have it, sometimes you’d line up your 3 misses in a row and then hit 7 times in a row. Nobody ever complains about hitting 7 times in a row, but missing 3 times does tend to agitate some folks.

To be fair, nobody wants to miss 3 times in a row IF your skill is high enough –  this simply isn't fun, especially if the enemy hits you every time. So we can count consecutive misses and once you hit 3, the next roll is on us and it's a hit! The hit card will still be removed from the deck, so you won’t be able to draw it twice, i.e. we will simply spread out your misses evenly but won’t give you free hits or help you win.
Logged
Fed
*
Posts: 135


View Profile
« Reply #1 on: May 29, 2018, 09:50:19 am »

Well... You don't have to exactly rig it. You can redefine what random is.

Say, instead of rolling a d20 each time - make a sequence of numbers from 1 to 20. In random order. When you need a random number - take the first one in the sequence. Then - the next one. When the sequence is done - generate a new sequence.
It's fair - more or less - in both ways. You don't artificially change the actual results to make them more even. At the same time you don't get thee ones in a row.
Logged

English isn't my native language... So if I say anything stupid, outrageous, wrong, offencive, disgusting, it's misscomunication.  8)
Vince
Developer

Posts: 7949



View Profile
« Reply #2 on: May 29, 2018, 10:48:34 am »

Well... You don't have to exactly rig it. You can redefine what random is.

Say, instead of rolling a d20 each time - make a sequence of numbers from 1 to 20. In random order. When you need a random number - take the first one in the sequence. Then - the next one. When the sequence is done - generate a new sequence. It's fair - more or less - in both ways. You don't artificially change the actual results to make them more even. At the same time you don't get thee ones in a row.
Unless I missed your point entirely, you can easily generate 3-4 misses in a row with this system too. There's no real difference between generating a sequence of numbers once and generating a sequence one number at a time, provided you don't generate the same number twice, which we don't.
Logged
Wrath of Dagon
Archmaster

Posts: 2337



View Profile
« Reply #3 on: May 29, 2018, 11:03:34 am »

True RNG is better if you reload, worse for Iron Man, so just depends what's the higher priority. Personally I reload.
Logged

Secondly--MURDER? Merely because I had planned the duel and provoked the quarrel! Never had I heard anything so preposterous.
Kirov89
Apprentice

Posts: 51


View Profile
« Reply #4 on: May 29, 2018, 11:13:51 am »

Considering the fact that no RNG could ever account for all the variables of RL combat, if the system gives you 7 hits out of 10 attempts for a THC of 70%, then it does it's job right. As for several consecutive misses, heck, even a sharpshooter can have a bad day or a yokel get lucky. I say leave it the way it is.
Logged
Etelvino
Neophyte

Posts: 2


View Profile
« Reply #5 on: May 29, 2018, 01:31:39 pm »

I don't think it would be a good idea to rig the RNG, mostly because it leads to unresonable expectations and messes with the game balance.

Also, in my experience, the issue with XCOM and other RNG-dependent games is not the sequence of failures or successes, but rather that a single sequence of 3 or 4 failures could doom you altogether. What I mean is that the battles are very prone to swings, with the possibility of eliminating a character in a single turn if he's been focused or an almost certain victory turning into a crushing defeat because no shots hit their mark on the previous turn.

I think that countering these swings is actually more important to make the player feel like tactics and skill are more important than the RNG. This could be accomplished by reducing how much damage an attack does, relative to the target's total health, or making damage mitigation matter more than damage avoidance (which is actually something AoD did fairly well).

Edit: formatting and typos
« Last Edit: May 29, 2018, 05:38:23 pm by Etelvino » Logged
MF
Journeyman

Posts: 107


View Profile WWW
« Reply #6 on: May 29, 2018, 02:36:06 pm »

That's just the way RNG works. Fiddling with the knobs won't change the core problem: uncertainty. I think the only solution that caters to every player is to use a deterministic system. I used to be on the fence about deterministic systems until I played and enjoyed Hard West.

The downside of deterministic systems is that, when optimal paths emerge, those paths will always be optimal and so encounters can be 'solved'. This trivialises non-multiplayer gameplay in the long run.

The downside of RNG is that, no matter how well you play, sometimes you just lose. Risk the boardgame is probably most notorious for this out of all gaming systems because attacking players overestimate their advantage, where, in reality, they should expect six losses for every seven wins with a 3:2 die advantage. In RPGs, people usually overestimate their odds in the same way, except it's a bit more obfuscated there.

Anyway, I don't think a deterministic system is an option for TNW, so I'll refer you to a common option for reducing ragequit-inducing losses on good odds in Risk: If the attacking player has an overwhelming advantage, he or she gains an extra die for every standard deviation away from an even match. For your example, that would translate to  anything above a certain threshold (could very well be 70%) getting a free reroll on a consecutive fail. It's an easy and generally applicable enough rule that you could make it a menu option or something.

Edit: I forgot about the usual way this is dealt with in RPGs: The Dungeon Master decides to fumble some of his own rolls to compensate. You could emulate this by causing the AI some grief after the player suffers some RNG hardship.
« Last Edit: May 29, 2018, 03:03:12 pm by MF » Logged
Wrath of Dagon
Archmaster

Posts: 2337



View Profile
« Reply #7 on: May 29, 2018, 03:42:09 pm »

Hard West has chance to hit, so how can it be deterministic? (Except for a some special abilities where THC is 100%. But that only works because you're usually heavily outnumbered, and tends to be a win button anyway).
Logged

Secondly--MURDER? Merely because I had planned the duel and provoked the quarrel! Never had I heard anything so preposterous.
Sparacul
Apprentice

Posts: 69



View Profile
« Reply #8 on: May 29, 2018, 07:02:16 pm »

I prefer non-rigged rng. It's more fun when its random. Maybe tweak some edge cases here and there.

To be fair, nobody wants to miss 3 times in a row IF your skill is high enough
Ignore first miss (on good chance to hit) per combat if weapon skill maxed?
Logged
MF
Journeyman

Posts: 107


View Profile WWW
« Reply #9 on: May 30, 2018, 02:10:26 am »

Hard West has chance to hit, so how can it be deterministic? (Except for a some special abilities where THC is 100%. But that only works because you're usually heavily outnumbered, and tends to be a win button anyway).

Hard West's chance to hit depends on the enemy's luck stat. If your 'chance to hit' is 50%, your enemy needs 50 luck points to not get hit. If he doesn't have 50 luck points in the bank, he gets hit for the weapon's determined damage and his luck points are replenished. If he does have enough luck, he dodges the bullet and 50 luck points are substracted from his total. Exception: 100% always hits and sub-20% shots are not allowed.

It's fully deterministic, there is no RNG involved whatsoever.

But like I said, it's probably not very relevant to TNW.
« Last Edit: May 30, 2018, 05:49:42 am by MF » Logged
Judge Mental One
Neophyte

Posts: 2


View Profile
« Reply #10 on: May 30, 2018, 12:37:35 pm »

Can leave the rng alone IMHO. Having played rpgs (p+p as well as computer) I know that unlikely things happen, like 3 or 4 1s in a row on a d20. Allowing the 3rd miss to be re-rolled (still not a guarantee) as suggested above when over 80% might be nice.

Still as long as we can save frequently, it should be fine.

 
Logged
Pladio
Craftsman

Posts: 346


View Profile
« Reply #11 on: May 30, 2018, 03:31:18 pm »

Actually, some games use an RNG but smooth it out differently.

Battle for Wesnoth for example has a chance to hit, but most units have multiple attacks per hit.

For example:

THC 50%, but instead of having only one attack, you have 3 attacks of 3 damage each at 50% THC each.
This differs from XCOM whereby you have 1 attack with 9 damage for THC 50%. The average is the same, but you end up having a higher chance to hit at least something.

The reasoning behind this is to avoid players complaining about just that, but actually also adds an additional dimension to the combat.

In a futuristic games with guns, this can easily be translated into burst mode with 5 attacks at lower accuracy or damage for example.
This could be done in any game though as we are talking about TB combat anyway...

The view would be that stronger attacks can either have lower accuracy and/or number of attacks for higher damage whilst you could have a weaker bunch of attacks.

I suggest playing the tutorial in Battle for Wesnoth to understand what I mean if you haven't yet.
I also believe Phoenix Point is using a similar system in their beta version from what I saw in their demos.

Hope this makes sense.

Logged
Dewey_Master
Craftsman

Posts: 348


View Profile
« Reply #12 on: May 31, 2018, 12:08:34 am »

Here is a sample case from the real world:

Kansas City QB Alex Smith completed 67.5% of his passes in 2017, averaging 33 attempts/game. That corresponds with about three shuffles of the AoD deck.

Game-by-game percentages fluctuated within a fairly small range. 15 single-game completion percents from highest to lowest:

80, 78.4, 76.7, 76.2, 75, 73.5, 73, 69.4, 67.5, 64.1, 63.9, 58.8, 57.6, 55.9, 45.2

These values mostly vary within +/- 10% of the season average:

12.5, 10.9, 9.2, 8.7, 7.5, 6, 5.5, 1.9, 0, -3.4, -3.6, -8.7, -9.9, -11.6, -22.3

*************

My takeaway from this is that actual performance should remain pretty tight to the predicted average while still admitting better and worse games. You accomplish the first part by using a 10 card deck. However, I think your system could be improved by shuffling after 8 draws rather than all 10.

With 10 draws, not only does Alex Smith complete exactly 70% of his every 30 throws, but you will regularly run into scenarios where the next outcome is predetermined.

With 8 draws, Alex Smith completes:
5/8 (62.5% completion or 7.5% lower than expected) on 47% of his shuffles
6/8 (75% completion or 5% higher than expected) on 47% of his shuffles
7/8 (87.5% completion or 17.5% higher than expected) on 7% of his shuffles.
Moreover, virtually every draw will still be undetermined before selection.
Logged
Black
Neophyte

Posts: 8


View Profile
« Reply #13 on: May 31, 2018, 12:29:41 am »

Quote
1)   Should we rig the RNG to meet players’ expectations?
No, never. You should never negotiate with people who don't understand how probability works.
Logged
puppyonastik
Craftsman

Posts: 282



View Profile
« Reply #14 on: May 31, 2018, 03:19:47 am »

Are you guys planning on a combat demo? I think that was a genius move for AOD. I think it would be wise to follow suit for TNW. If not, I trust in the teams' judgement.
Logged

Quote from: Mayday
Can a single dick even enter a group setting?
Quote from: Fez
You've never lived until you've tried it.
Pages: [1] 2 3 4   Go Up
Print
Jump to: