Has anybody else noticed that glitches have some odd probablities associated with them? Used house rules to fix this?

Below is the output from a program I wrote to show (roughly) the chances of glitching for various pools and varieties of glitch; note that in all cases these are normal glitches, not critical. This data is based on 100000 rolls; the pools used for each column are the same, and smaller pools are sub-sets of larger pools, so you can at least compare "apples to apples", even if the specific values are off a bit due to sample error.

The "glitch" column is for the basic glitch rule. Worth noting is the chances for glitching with 4 dice is HIGHER than that for glitching with 3, 6 is higher than for 5, etc. Its in fact almost twice as likely.

The "grem1" column shows the chances of glitching if its a test using the Gremlins Quality, rating 1. Again, the chances are higher for even number dice pool sizes than for the next pool size down- again, almost double.

The "rush" column shows the chances of glitching if its a test using the rushing a job rule (1's and 2's count towards glitches). Again, there's the same issue, and its pretty noticable, though the jumps upwards are not quite as severe.

The house rule I came up with to address this is something I call the "hold one" rule. Basically, one of the dice in your pool (physically it would be an odd colored or otherwise distinctive die unlike any of the other dice) is "held back" for glitch purposes, and only if the roll is a glitch both with and without the "hold back" die is it really a glitch.

The "HO" columns show the same role types as the previous 3, using the "hold one" rule. The basic effect is obvious from these charts- the chance of a glitch with an even sized dice pool is the same as that of the next smaller odd sized pool. This removes some of the oddities from glitching, but also decreases chances of glitches over-all. Are glitches a big factor in play? There's ways to bump up the chances a bit so the curve matches the basic glitch curve but is (much) smother (adding an extra die that contributes if it comes up a 1 works well), but is the added complexity worth it?

I'm also considereing replacing the gremlins rule with one that has characters using something like the rushed job rule, because I like the fact that large pools can still glitch, and small ones don't do so automatically.

pool glitch grem1 rush HOglitch HOgrem1 HOrush
2 0.306 1.000 0.555 0.167 1.000 0.332
3 0.074 0.423 0.259 0.074 0.423 0.259
4 0.132 0.518 0.408 0.074 0.423 0.259
5 0.035 0.196 0.209 0.035 0.196 0.209
6 0.061 0.261 0.318 0.035 0.196 0.209
7 0.018 0.095 0.172 0.018 0.095 0.172
8 0.030 0.134 0.258 0.018 0.095 0.172
9 0.009 0.048 0.144 0.009 0.048 0.144
10 0.015 0.070 0.213 0.009 0.048 0.144
11 0.005 0.025 0.121 0.005 0.025 0.121
12 0.008 0.037 0.177 0.005 0.025 0.121

I'd not really thought about it before, but you're right, there is an increased chance to glitch on even DPs over their previous counterpart.

However, as an easier solution to the 'hold one' rule, you could just have odd numbers around down instead of up when determining how many 1s are needed for a glitch.

Another thing to think about though, is that a critical glitch is more likely on 3 dice than 4, so I'd still rather take 4 dice over 3 even if I am taking an increased risk in getting a regular glitch.

I think your solution very elegantly fixes the problem as you identify it, but it does add a little bit more complexity. I always ask myself if that complexity is worth it- i.e. is the problem more bothersome than the fix. In this case, that little bit of added complexity will impact every roll you make. I guess my question would be: do you find glitches so bothersome that its worth this?

What Karolina said. What you're not taking into consideration is the critical glitch rate, which are more common on odd dice compared to the even dice partners.

Mhmm, each even numbered pool downgrades the Critical glitch chance, converting it to a regular kind. Since hits are derived on 5 or 6 and a glitch is only on a 1 under normal circumstances, each dice is twice as likely to hit then it is to glitch as I see it.

Imaginary number time cause i'm lazy and not doing the real math:
lets say out of 20 rolls of 5 dice you have a glitch rate of 3 and a crit glitch of 2. Add another die to that and it converts the rate to more like glitch of 4 and crit glitch of 1. Both add up to a total of 5 glitches of any kind.

I think my logic is right, but I'm not math wiz.

Considering that crit glitches are catastrophic failures compared to regular glitches I rather like the trade off.

Actually, with book rules, the chances of a critical glitch on 3 dice is 4.63%, while the chances of a critical glitch on 4 dice is 5.17%. So not only are there more glitches with 4 dice than with 3, there are more critical glitches.

When applied to critical glitches the inversion does persist at higher dice counts, but it is much smaller.

Probably the uglist cases are at the 1 to 2 dice range. 1 die has a 1/6 chance of glitching, and such a glitch is always critical. One would assume that would dominate the 2 die glitches, where there are both critical and non-critical glitches. But it turns out there are slightly more gritical glitches with 2 dice than there are with 1.

Yours,
Joel

PS: Mongoose, your numbers are actually for critical plus non-critical glitches.

Well, that'll teach me to not run the math on odds before posting. I still say the much easier fix is to round the number of dice needed to glitch on odd numbers down instead of up. Technically the the rules do indicate that you should be rounding up, but like I said, this fixes that oddity.

Now, correct my math here, but on a 1 die roll, you have a 1/6 chance of glitching, which is a 16.66% chance of critical glitch.

On two dice though, you have to roll a 1 and then a not 5 or 6. That is a 1/6 chance times a 4/6 chance which is 16.66% * 66.66% = 11.106% chance of a critical glitch.

of 36 possibilities, (1 1 to 6 6) there are 7 critical glitches (1 1, 1 2, 1 3, 1 4, 4 1, 3 1, 2 1) - this is a 19.44% glitch rate.

on 6 possibilities (1 to 6) there is 1 critical glitch at, yes, 16.67%.

your probability would be right if we rolled one die, and then the other [edit] only glitching on the first die, but we don't; so there's a few extra combinations (notably the probability of 'not a hit not a glitch' on die one * the probability of a glitch on die two (1/2)*(1/6) = 8.33% -> which is conveniently the difference between your result and mine.)

However, rolling that 1 means that you still glitch. So, you still have a 1/6 chance to glitch on a 2 dice roll. You just have a lower chance of actually hitting a critical glitch.

I was afraid it might do something like that. I figured it counted that. Odds was never something we went over properly in statistics class, which depressed me as a gamer.

There are a number of resources online; really, approaching it from combinatorics is probably more helpful than approaching it from statistics, as simple counting problems don't really need powerful statistical tools.

Plus you get to not make me sad by performing random trials to generate approximations for easily-calculated probabilities.

~J

Since they were asked about in general, I use a house rule where you glitch if a number of 1s show up equal to the lowest of the associated Attribute or Skill. Critical glitches follow the same rules as normal glitches. Say you're shooting a pistol and have Agility 6 and Pistols 2. It's only going to take two 1's for a glitch to occur regardless of how many dice you're throwing, as your Pistols skill is only a 2. If half the dice pool comes up as 1s, it's a critical glitch.

The philosophy is 1) that I like glitches as both a player and a GM and 2) that unskilled people and untalented people both do a shoddy job of things. I also like to insure that critical glitches actually come up from time to time. Which they don't. I don't think I've ever seen one outside of my own dumb luck; but that's par of the course.

Note that I treat glitches as simple little mistakes, not the crippling acts of stupidity a lot of other GMs do. A glitch on a Pistols Test, for instance, might mean it jammed after firing, but a Simple Action to whack it on the side will get it working again. A critical glitch means it's jammed until you can sit down and work on it. Which is pretty much how it's supposed to be for the mos tpart, but I've seen way too many GMs treat glitches like critical glitches. Note that I do allow actions to succeed if the test itself was still a success,, it's just the consequences of the aftereffects that vary.

Just to complete the story properly, due to the way dice interact, the total of critical and non-critical glitchs on 2 dice is actually 11/36. With 7 of those being critical (as compared with 6/36 for 1 die), and 4 of those being non-critical.

Yours,
Joel

PS: Yes, I enjoy doing this sort of computation.

PPS:Rounding the number of dice needed for glitches down instead of up does not produce quite the effect you intend. For 3 dice, this would produce 48/216 (8/36) non-critical glitches and 37/216 (just over 1/6) critical glitches. I.e. 3 dice would critical glitch slightly more often than 1 die, and have regular glitches to content with as well.

uh, no? (5 1, 6 1, 1 5, 1 6) is the actual 'plain' glitches, that 4/36 or 1 in 9. (11.11%)

if you mean glitches and critical glitches that's (11/36) or 30.5%.

you actually have a higher chance of crit glitching with 2 dice than glitching with 2 dice. (if you meant 1 die versus two then you're still wrong, as 19.44%(7/36) is greater than 1/6.)

This is slightly off topic, but is this actually stated anywhere in the BBB? I thought it was, but I did a cursory search the other day (SR4A) and couldn't find any hard fast rule.

And I agree with Dr. Funk. I think glitches are a little too rare and I tend to house rule toward more glitches. The ol' stove pipe jam is my favorite non-critical glitch because it's basically an annoyance that slows the player down (and I happen to think that SR combat is ridiculously too fast).

[As a side note, is it weird that there is a baby in the background of that shooting video?]

edit: Critical Glitch

Well, I say technically because it says "If half or more of the dice pool rolled come up as 1s, then a glitch results." and half of 3 is 1.5 Since rolling 1 1 is not at least 1.5, and you can't roll half a die (Well, you know d3s not counted), it seems that you would have to get at least 2 1s for a glitch on 3 dice.

Ah I see. I thought maybe there was a blanket "always round up" rule somewhere (i.e.- not just glitches). Maybe I'm on crack.

No, fairly sure Shadowrun follows the standard rules for rounding (.5 or greater means round up, .499999 or less means round down) when it occurs and whole numbers are required, which isn't all that often outside of initiation.

There is one thing to think about as an advantage to 4 dice over 3. You can just buy the hit and not worry about glitching under non-stress circumstances. I think that is good reason to want to go from 3 dice to 4 (As well as the increased chance of some success, even if it is a glitch)

They seem to occur quite often in my experience. Off the top of my head you have glitches, things like electrical damage that (obnoxiously) halve armor and the limitations on the Take Aim action.

And I seem to point this out a lot on the boards, but you cannot buy successes with a DP of 4. The rules state that the GM should only allow purchasing successes for large dice pools when there is little-to-no risk of failure.

Does it actually state that as a rule, or is that another of the helpful "suggestions" that were so liberally scattered throughout the SR3 rules?

Edit: ok, I just dusted the cobwebs off of my copy of SR4, and unless they've changed it in SR4A it is indeed a suggestion. You can buy successes with 4 dice under the same conditions as with any other number of dice, that is to say "if the gamemaster allows it".

~J, who recently had to go through every sourcebook and make a ruling on every last suggestion and optional rule in the game. This is why optional rules and suggestions for fudging are design failure, people.

(Bitter, me? Never!)

Pretty sure SR4 actually states you always round fractions in favour of the players in one of the introduction sections. It would seem that it would influence all rounding.

Where does rounding come into a problem with Take Aim? I didn't know there was any division at all involved in Take Aim.

You are right, there are a few other instances of rounding where whole numbers are needed, but that still isn't many. Glitches technically doesn't round because it technically doesn't require a whole number. You are right about half armor being one of the few other places where you have to round.

almost without except SR4 indicates which way to round on things that result in halves, generally this is in the PCs favor* ex. condition monitors specifically state '8 + Attribute/2 (round up)' but for instance, the have your gear catch you while falling test (pg133 SR4A) is half the PCs body (round down).

considering the rule is 'half or more' I would rule that 1 of 3 is less than 1.5 and therefore not subject to the glitch rule (regardless of rounding).



As per page 62, SR4A, the rules do not specifically prohibit buying hits with a DP of 4 (e.g. it's in the table.) the rule statement is specifically: 'Gamemasters should only allow this when the character has an exceptionally large dice pool (and is unlikely to fail) or when the situation is non-threatening and non-stressful.

it is important to note that, per RAW it is an "or" statement. (That is, exceptionally large** pools are permitted by RAW to purchase hits even in threatening or stressful situations and you are permitted to purchase hits in non-threatening** or unstressful** situations even when your pool is exactly 4.)


* A change from previous editions.

**exact value subject to GM discretion

I believe I'm right in saying that this applies, accepting where it is specified in the rules to round one way or the other.

Thats what I thought, but I couldn't find the actual statement.


So in this case it is stated that you round down, but in other cases its round up. Thats why I was curious.

At any rate, we should probably stop derailing Mongoose's thread.

I can't find a citation for it, but every case a brief search turned up of the whole word 'round' in my pdf copy indicates that this is the case explicitly.*


*I stopped after eight or nine, there may be an example against the players later in the book.

[edit] and method has located one. so there's at least one exception.

Yeah I'm not 100% but this has definately come up on dumpshock atleast half a dozen times before that I can remember.

Edit: In any case, it's what I did last time I ran a game.

While we're on the topic and you checked, is SR3 explicitly round against the players? or did I just incorporate an old house rule of my own into my mental ruleset as canon?

So, searching through my SR4 PDF the only mention I could find of how to round was in jumping tests where it specifically states to divide by two, rounding up.

Oh, just found one on condition monitors, once again round up.

Melee damage is once again round up.

Impact armor against falls (Halved) rounds up, so presumably so does v electricity.

Actually quite a few other instances of rounding are popping up.

It seems like the rule is "Round in favor of whoever the rounding is affecting directly"

Drain is down, damage to vehicle from ramming is down, a penalty to a particular drug is down.

The only oddities I see are that you round down when halving concealability for a pat down, and you round down for weapon mounts on drones... oh, and round down on how many throwing weapons you can ready in an action.

So yeah, some oddities it seems that doesn't really allow you to put down a general rule all that well after all. It does seem that rounding is accounted for in everything that rounding is needed for though.

My guess is that its due to a concerted effort on the part of the SR4A writers.

In SR3, pool calculation is always round down, as are chargen knowledge and language skill points; in several places during these calculations there are references to "always round down" which seems to apply specifically to the current calculations but could be misread. Both halving armor for armor-piercing munitions and halving Power for normal weapons against vehicles round down; it's not clear to me if that's round-for, round-against, or just round-somehow. Dice rolled for the grenades/explosives optional rule (half Power vs. TN 4 for staging) are rounded up; again, favor is unclear.

Wait! I found one! Visibility Modifiers other than Full Darkness are halved, round down when in melee combat; that's probably something that you could consider rounding in their favour. Also, Flux Rating for Sensors and ECM is Rating*1.5, rounded up, which sounds pro-player.

Heh, there's even a bit of flux I forgot where you round to the nearest half. Oh, Shadowrun.

Anyway, there's a lot I can't firmly call one way or another in terms of being pro-player or anti-player, but there are some cases where it seems to be pro.

~J

Except I'm just using the SR4 regular book

Well more props to them then.

Most of the time in SR3, you round down. There are a few things that specifically mention rounding up (Kage mentioned soem of them, and I can't think of any more off the top of my head), but nothing is player specific. You round it how you round it, and it screws whomever it screws. (In fact, my group houseruled it as Always Round Down, because simple, universal rules are easier to remember. Like, Tie Goes to the Defender.)

Technically almost everything that rounds is going to not really be for or against the PCs, because the NPCs run under the same rules. So half impact armor round up when defending against e damage helps the PCs when they are being hit, and hurts them when the 'bad' guys are being hit. Only ones that basically only affect PCs would be karma rounding up/down in the case of initiation.

In principle you could have a round-for or round-against PCs (so a Heavy Pistol round against a hostile NPC vehicle would have an adjusted Power of 4, while the same against a PC or PC-allied vehicle would have adjusted Power 5), it would just be sorta dumb.

~J

always rounding impact armor down would generally be more dangerous to players since they take more shots on average than the typical npc.

kage summarizes the absurdity of 'round against the players' rather succinctly however. (although a rule like 'round against the defender might work ok.)

Occasionally you get situations where it's rather hard to decide who is the defender. I think I prefer to round .5 up and .49999 down, except if explicitly noted otherwise.

And all ties not explicitly noted are resolved with rock paper scissors. It's faster than looking stuff up.

The thing is .49999 doesn't come up much in SR because you are generally working with whole numbers and generally halving them (cyberware essence costs and grades being the chief exception I can think of). So if you start with an even number you get an obvious whole number result, but anything odd will result in a .5 and thus round up per standard mathematical rules. So really "rounding up" should be the default, but there are some situations where that might be silly.

And I want to apologize to Mongoose for derailing his thread.

True, I didn't differentiate. Didn't seem important to the discussion.



So if you default, you always glitch? Or if you have equally low skill and low attributes, you are less likely to glitch than if you have a higher rating in just one of the two? Seems... glitchy. But hey, if it works for your goroop, cool. I agree that glitches are a bit rare, and that was one problem I had with the house rule I considered- it makes them rarer.

As for rounding, I'm an old SR2 player (where things like halved armro most certainly did round down) so assumed you always round down unless otherwise stated. It seems to apply in SR4- most notably when determining maximum augmented skill ratings, and (iirc) for drain. But it does seem to generally be covered in the new text on a case by case basis.

Back in SR1 and 2, almost everything was "round bad". Now, the rule of thumb is that if rounding a 1 to a 0 doesn't makes sense, you round up.