Now, I know that I may be frowned upon by everyone for creating such a heretical thread, but after yesterday's test I realised how slow, unwieldy and confusing the basic mechanic of Donjon is. I've already seen some solutions to the dice problem in the game, but none of them is exactly what I'd want if I actually ran the game (and sure I will someday). After contemplating alternative methods of rolling dice in Donjon I came with the following variant rules:

Test formula:

1d20 + Attribute + Ability + equipment bonus + modifiers
(or 1d20 + base difficulty + Donjon level + modifiers)

Whoever gets higher total gains one success, plus one additional success for every full 5 points he has over the opponents result.

Natural 20 means that the player rolls open-ended 1d6 and adds the result to the total. Natural 1 means that the player rolls open-ended 1d6 and subtracts the result from the total. These rolls are open-ended - as long as the player rolls 6 he gets to reroll and adds/subtracts from the total (only in such "critical" rolls d6s are open-ended).

The final total can never become lower than 0.

If there is a tie in final totals, both players roll 1d6 and add to their results. Repeat until the tie gets resolved.

Law of Success:

One success still means one die - 1d6 is added to the total.

In Spending Tests, spending every Wealth or Provision point adds 1d6 to the total.

Every other bonus or penalty directly modifies the total. E.g. -3 range penalty means that the player simply subtracts 3 points from his result. Only law of success and Wealth/Provision points generate additional dice.

Actually, I'm not sure what to do with bonus and penalty dice originating from spells.

I noticed that default Donjon mechanic is very random, and the randomness should be preserved with rolling 1d20. Still, I added Alternity-like bonus/penalty dice, because pool mechanics are usually more fun for me. There should be no problem with buckets of dice with this method, though. The problem is that the variant changes probabilities and messes with the value of certain modifiers (dice generated with Law of Success will influence the results more than the rest of the modifiers). Also, it would be very difficult to gain more than 3-4 successes in one Test, I think.

Now, as for the initiative, I like the method presented in Donjon - it reminds me of 7th Sea, and generates action patterns that support tactical thinking well. On the other hand, the mechanic is impractical - it requires big numbers of twenty-sided dice to be rolled and they have to remain on the table (unless you note things down or use some kind of initiative tracker, but such methods are bothersome and slow things down). To preserve the way initiative works and solve the "buckets of dice issue", I'd adapt initiative rules from Threads and use cards. Instaed of rolling dice, every player draws one card per dice he would normally roll (from his own deck, or from "public" deck, or maybe two mixed decks). Then, actions proceed from the best card to the worst, and cards are discarded after they are used for action. Decks are reshuffled after every flurry. This method has two advantages - even with one deck there is enough cards to represent actions of quite a number of combatants, and it's easy to track initiative and strategize when all action cards are visible on the table.

Thoughts? Comments? Suggestions?

Hey, Filip.

I disagree that the Donjon mechanic is slow or confusing - possibly unwieldy for small hands - but I recently played with an 8- and an 11-year old, who had no difficulty understanding the dice mechanics; I just don't see a problem there.

That said, I think your solution to the non-problem is heinous: it converts a modest dice pool into a single d20 roll, and completely screws with the probability curves. I find that my intuitive expectation of success is more satisfyingly reflected with pool mechanics like Donjon's than with linear d20 probabilities (like D&D), which drive me towards cataclysmic temper-loss and bad behaviour. They make characters who are supposed to be good at stuff fail, all the time. Furthermore, you are suggesting the replacement of one simple die-comparison mechanic with math, multiple die sizes and a deck of cards; I would not call that a simplification.

Go with the bucket of dice. You won't regret it.

Like any other mechanic it takes some getting used to.  But I expect your initial struggles with it come from less than optimal interpretation rather than actual high handling time mechanics.

The actual steps are these:

Player #1 (scans dice where they fell):  What's your highest roll, mine is a 17.
Player #2 (scans dice where they fell):  Letsee...I got a 19...and 1...2...3 other dice that beat a 17.
Player #1: Drat, you got 4 successes.

That's it.  There's an extra step in there if the high dice are tied, but its just adding the number of tied dice to the winner.

Often times people try to sort the dice, or arrange them, or match them up, or all kinds of other fiddling with the dice.  That's totally unnecessary.  If you touch the dice at all you're just creating extra work for your self.  Pan and scan and spot the high die...pretty easy once you get in the habit.

That said, I do prefer using d10s to d20s solely because d20s tend to roll more freely increasing the amount of space to scan.  But if you roll in a box lid to keep them all contained, its not a big deal.

Er, I understand how the procedure should go, I've already seen those suggestions many times. And somehow I still don't buy it. The default mechanic is simply out of question for me, at least for real life games.

We've been testing how the game works on Irc. Scanning a string of 5-10 numbers made my eyes hurt, and I often failed to notice the highest die. Ok, after getting used to the thing, it might work for me on Irc. Although somehow I'm not too enthusiastic if I imagine going through, say, 20 or more dice.

Now, when it comes to real life, I'm a proud owner of 11 twenty-siders. That's a lot in Poland. Probably I have the biggest collection of d20s in town, save game store owner (but he doesn't count, since he's selling them ;)). Most of the people I know have no more than 3 twenty sided dice, provided they have their own dice at all. In Poland it's fairly common for people to come to the game with no dice, expecting to borrow them from the GM.

So, that's it for the buckets. I can easily imagine low-level characters throwing more d20s than the whole group has at disposal. And I don't plan buying a new bucket of d20s just for Donjon.

(Jeez, how many dice do you people use?)

Well, I could use d10s or d6s, they are more common and I have bigger piles of them, but they mess with probabilities anyway.

But the source of my issue is not even the number of dice - we've been dealing with dice pools of 30 or more in Exalted with no problems, and pretty fast. Mostly because grabbing the dice with successes back and rolling the pool in 10-die increments is easy, fun and quick.

In Donjon the dice have to stay at the table for the whole Test, or there would be some complicated re-rolling of low-number dice involved. When it comes to initiative, it looks even worse - for three low-level characters and a number of monsters possibly around 20 dice would be needed only to indicate the action numbers, for advanced party much more.

That's so impractical that my head hurts ;)

And no matter how weary I am with 1d20 based mechanics, it feels just in place for Donjon. And despite the mathematics involved, it seems impossible to me that scanning the dice for highest numbers and comparing could be faster, even with practice. Adding two or three numbers doesn't take much effort.

At this moment I'm not very worried with probability curves, as long as no side gets privileged in the Tests (and it doesn't seem so to me), and the method produces roughly similar amounts of successes.

Now, I only need to know how much this d20 method of mine differs from the default when it comes the last point. And if it differs much, how could I make something similar produce more equal results?

What's interesting - usually when I approach new game, I see lots of minor mechanical details that irritate me. In Donjon, absolutely nothing looks out of place for me - save the basic rolling method ;)

Hey, Filip.

Well, in practice, at the table with real dice, I found the dice pools easy to read, as did the 8- and 11-year-old I played with. IRC is a whole different bucket'o'dice, though, so YMMV.

I didn't have enough d20s, so I used d10s - which does less bad things to the probabilities than using a single d20. The distinction is between a normal curve, vs. a flat line. The flat line of a single d20 is horrible for matching player expectation with dice results. Surprisingly bad (or surprisingly good) results happen much more often with that flat line, than with a nice curve. You can't change the linear nature of the probability of a single die-roll with another linear transform, which is pretty much all you have available to you: adding and multiplying. Once you get into exponentiating or other fun math stuff at the table, you might as well roll more dice. Busting out a calculator is not usually much fun.

Initiative is not a problem if you have a pencil and paper.

The short answers to the "only need to know" part are: a lot, and you can't. Roll more dice, and get those nice normal distributions butting heads. I would always use pools of d10 before switching to a single d20. (Heck, I'd use pools of d6 before a single d20, every time.)

(P.S. my stats are a bit rusty. It's been twenty years or so.)

I didn't have enough d20s, so I used d10s - which does less bad things to the probabilities than using a single d20. The distinction is between a normal curve, vs. a flat line. The flat line of a single d20 is horrible for matching player expectation with dice results.

Initiative is not a problem if you have a pencil and paper.

The short answers to the "only need to know" part are: a lot, and you can't. Roll more dice, and get those nice normal distributions butting heads. I would always use pools of d10 before switching to a single d20. (Heck, I'd use pools of d6 before a single d20, every time.)

Filip,

You are not going to get an answer to your question, I don't think. You should be able to figure this out with math.

Everyone on this thread has been exceptionally nice to you, and you're driving them away. You say, "Donjon is too complex," they say, "we've played it and it's not," and you say "it is!" I understand you've tested it, and if it's not for you, ok! They've all played at least one session and enjoyed it, and ok for them, too!

I'm not sure what to tell you. E-mail me if you have any questions.

Ok, I've spent almost two hours rolling dice and counting averages today, and despite being driven away by the reactions on this forum, I'll share the results of my experiment. (And sorry, but I tend to miss on niceness when I feel like being attacked by someone; and attempts to convince me that my problem is not a problem do not help me at all)

So, I tried with the following dice pools, using IRC dice roller: 5 vs 5, 10 vs 5, 10 vs 10, 20 vs 10 and 20 vs 20. I didn't go past 20 dice per side, since that would involve rolling over 40 dice in total and that's way too much IRL (since, how many people actually have so insane loads of dice? ;)), and with the pools of 20 scanning the numbers become bothersome enough. That covers the first 10 levels in Donjon anyway.

Generally, I got a range of 1-4 successes. I've seen only a few rolls, less than 10% total, that gave me more than 4 (5-7).

Bigger dice pools gradually produced slightly higher average - I suppose it's either one additional success on average for every doubling of dice, or for every 10 or so additional dice, as long as the pools are roughly equal. With a difference in pool size, an average difference in successes became roughly doubled as the pool difference doubled (but it didn't go past 1-4 range anyway).

I tried the d10 method - it gives a lot more ties than d20 method, and I can't agree that it doesn't hurt the probabilities much.

I also tried a method in which I rolled pools of d10, counting successes on 6+, and then counting the difference in successes (so, an equivalent of tossing coin pools). It produced lower average range of successes, and favored the side with bigger pool much more than the standard method.

Anyway, these methods involve rolling insane numbers of dice only to get 1-4 average success range. Also, the number of dice grows with experience unproportionally to the increase of average results. So, is there actually any other benefit of bothering with buckets of dice other than the tactile appeal of dice pools? (Well, the tactile appeal certainly has its merits when it comes to dice pools, but it seems like taking things to the extreme here, without real justification)

Now, looking at my method, it produces almost the same range of successes as the standard method and keeps a decent probability curve (with 1d20 vs 1d20 extreme differences in result are not as probable as with d20 system standard of 1d20 vs. static number). Its handling time is comparable on lower levels (i.e. seconds vs. seconds) and most probably faster at higher levels (when dice management obviously starts to slow things down). It gets rid of those insane buckets of dice, but still keeps tactile benefits of a pool with those bonus d6. Finally, it gives more mechanical impact to the law of success and makes spending Wealth/Provision points more worthwile (these two things may just as well be a benefit or a disadvantage).

To sum up, it is roughly comparable with the standard method but solves the problem of ever growing buckets of dice. Calling it "heinous" without giving it any consideration is, well, heinous.