# Dungeon Feedback **URL:** \ **Category:** Official News\ **Created:** [September 30, 2022, 5:57am UTC](https://community.gemsofwar.com/t/dungeon-feedback/77122 "2022-09-30T05:57:13Z")\ **Posts on this page:** 1\ **Showing post:** 257
**Author:** ![Stratelier](https://avatars.discourse-cdn.com/v4/letter/s/977dab/32.png) [@Stratelier](https://community.gemsofwar.com/u/Stratelier)\ **Post date:** [October 6, 2022, 5:34pm UTC](https://community.gemsofwar.com/t/dungeon-feedback/77122/257 "2022-10-06T17:34:16Z")
Oh wow, long post time! > [@Kafka](#): > > It was necessary that we fix the broken RNG in the Dungeon though, because we told everyone the doors had equal chances for everything, Technically speaking, where was this? As far as I know this was just a default assumption on our part (because “every outcome is equally likely” is the _literal definition_ of “fair RNG”), and the original announcement only said that the Dungeon _(as a whole)_ would contain 3 bosses, 2 traps and 1 bonus. EDIT - as [originally stated here](https://community.gemsofwar.com/t/6-5-update-dragons-path/76669): > [@6.5 Update: Dragon's Path](https://community.gemsofwar.com/t/6-5-update-dragons-path/76669/1): > > Players will have 6 Dungeon doors to pick from, behind each door is either a Battle, Trap or Boon Room. Each Dungeon will have 3 Battle Rooms, 2 Traps, and 1 Boon Room. So technically speaking, this never specified what kind of RNG was being performed (other than an assumed fair RNG). But here’s a simple idea that would be easy to implement – the “Minesweeper rule”: The first door opened should be _required to NOT be a trap._ Because for every player hunting Dragonite specifically, if they open a trap on door 1 they will quit playing for the day because of the immediate confirmation that the desired outcome has been excluded. And the LAST thing you want the player to do is quit playing, right? > **Analysis of how it works** > > 1 - The first door opened is limited to 4 outcomes (3 bosses and 1 bonus). > 2 - _After that,_ the remaining 5 doors get shuffled by a fair RNG. > > Of the 60 possible combinations of runs (6 of which are Perfect Runs), 20 of them (trap on first door) are removed outright, boosting the overall rate of a “perfect run” by 1.5x (15%, from 10%). This may sound like a significant boost to the Dragonite economy over time but (1) it’s STILL a very long grind to acquire a full set of Dragons, and (2) you can always just adjust the base Dragonite payout to compensate for the higher “win rate” (e.g. 50+ instead of 80+). > [@Kafka](#): > > A very common strategy when it comes to random chance selection is to choose the same doors every time because eventually your chosen combination will be successful. So if you were one of the players who would choose a combination of Doors that was affected by this bug, your chances of getting a perfect run were significantly reduced - and that’s just not fair. This was one of my reactions once the patterns were discovered too, we had some players unknowingly picking “winning” or “losing” combinations and boasting/complaining accordingly based on what we DIDN’T know about the process. > [@Kafka](#): > > There was a bug in the Door shuffle code that caused it to skip the source at the index of the target slot, so when it was working out target slot index 0 (Door 1) it ignored source index 0 (Low Boss battle) etc. So this was legitimately a bug all along? I can see that. 0-based vs. 1-based indexing (or in this case, probably _n_-based mod _n_) is a _frighteningly_ easy thing to get wrong. > [@sls](#): > > Shuffling 6 unique numbers is so trivial, it is for all intents and purposes **impossible** to mess up. As long as you are shuffling the _entire batch at once._ Considering we have bug reports of duplicate stairs behind multiple doors, this implies that doors were being determined on a just-in-time basis based on the client telling the server which door you’re opening and which options have been previously revealed, a more complicated design which multiplies the opportunities for bugs. But now that it was discovered, I actually would kinda like to see it return on purpose, as a feature. The catch is that the “source index” list (Boss 1, Boss 2, Dragon, Trap 1, Trap 2, Bonus) needs to itself be shuffled by fair RNG on a weekly basis. Then large guilds CAN work together collecting data to figure out which doors have the best odds, BUT it only works for a few days and they have to start over every Monday with the weekly reset. And likewise, someone who isn’t collecting their data, just picking fixed doors every time may have “good” weeks and “bad” weeks, but over time it will average out to a fair RNG. > [@Kafka](#): > > This is why the economy balancing in this instance is also important, if we give too much Dragonite it will cause the Dragons to become early and common collectables and we really wanted them to be more special than that. Or to quote one of the single most HATED developer/publisher posts of all time, “The goal is to give players a sense of pride and accomplishment.”. However, it is impossible to feel any sense of accomplishment from RNG, because _by definition_ the outcome is essentially predetermined, nothing you do can influence it. > [@Aragorn1](#): > > When talking about potentially having Dragonite gnomes that gives me serious PTSD. You know, I don’t even remember the last time I saw a Cursed gnome. It has to be a couple of months now. Last week I found two in a single day … _without_ crafting a GAP. Again, that’s the fundamental nature with RNG: it likes to cluster, so there will be “runs” of a specific outcome AND (more often) “droughts” of it, and there’s no way to verify whether the next “winning pull” went to you or somebody else. For a quick example, just compare the RNG piece selection of retro (original) Tetris vs. modern Tetris: - Retro Tetris used a fair RNG (“dice roll”) for each piece independently. Depending on the quality of the RNG used, you could get “runs” of certain pieces, or go very long times without getting a specific shape. - Modern Tetris uses a “deck of cards” RNG, giving you all 7 shapes shuffled in a random order. You will only rarely see 2 of the same shape in a row, never see 3 or more in a row, and will rarely go more than 9 pieces between instances of a specific shape. Combined with other modern features (particularly the Hold Piece) this makes modern Tetris feel completely different from retro Tetris despite the same underlying game design. > [@Moriad](#): > > I would add that, if one of the devs’ goals is to encourage consistent daily play, it is much easier to rationalize missing a day when you _expect_ to get _nothing_ than it is to miss a day when you _know_ you will get _something_. This psychology from the player perspective cannot be emphasized enough. For example, if you log in 28 days out of a given month you will get a free Orb of Chaos at the end of it, something which is otherwise given out mostly as event rewards only. So for early- and mid- level players who are still collecting and filling out their Troop roster this is a very anticipated item, and this particular means of getting one is (while only 1 a month) absolutely guaranteed. > [@sls](#): > > As has been stated by Fourdottwoone and others: > > A carrot on a stick only works if the donkey can still see the carrot in front of it. > If the carrot is too far away, the donkey stops running. Let’s make the other obvious comparison: A literal slot machine at an actual casino will, on average, pay out a specific percentage of small wins over time, sometimes a larger win, and very rarely that massive jackpot. These recurring small wins are effectively REQUIRED because they _motivate the user to keep playing._ A long streak of losses encourages the player to walk away sooner with whatever they have left. Heck, even “Scratch-it” style lottery tickets (with predetermined outcomes decided at the factory) tend to have an overall 1/5 “win rate”, regardless of whether the win is more or less than the original ticket price. In contrast, our current “Dragonite lottery” is functionally jackpot or nothing, with none of the small wins that _actually motivate_ continued play. Another simple idea: Break down the Dragonite rewards into smaller, compartmentalized payouts: - Upon defeating the Gem Dragon boss: award (1 x Level) Dragonite (Note this is _guaranteed_ regardless of traps, very important because it incentivizes actually playing the mode, AND using the stairs to increase the Dungeon level) - After defeating all three bosses: Award (1 x Level) Dragonite _(again, this is guaranteed)_ Open all remaining doors automatically Reward +20 bonus Dragonite for first Trap not triggered Reward +40 bonus Dragonite for second Trap not triggered The overall economy over time is therefore: - Perfect run (10%) = +60 Dragonite (average +6 /day) - Single trap (30%) = +20 Dragonite (average +6 /day) - Total: Average +12 per day (plus 2 x Level) > [@awryan](#): > > Your player base has dropped significantly since the implementation. Hate to play “that guy” but your data for that claim is …? --- _[View the full topic](https://community.gemsofwar.com/t/dungeon-feedback/77122)._