Search • The Tavern

The comment about AlphaZero above brings to my mind another observation. AlphaZero also learnt how to play StarCraft expertly. I am not a StarCraft player, but from what I understand, AlphaZero just spammed a cheap and fast unit and then used superior micro to take down human players. This makes me ...

It's definitely possible. XL is a rather coarse measure -- using XP would be better. Unfortunately, there is no fine-grained measure of XP which I can get from Sequell data. At least, I don't know how to do it, maybe some Sequell wizards can figure it out. Many things in-game depend on XL, so it's n...

I do not understand what any of that really means. Yeah you can theoretically calculate difficulty based on vault weights, spawn weights, odds of various combat events etc. That's a bit like saying that you can model human behaviour by using quantum mechanics because everything is based on physics a...

Stuff. I haven't replied to your comments because: ⋅ You agree that id should be removed. ⋅ You agree that early game difficulty is harder than later game. ⋅ You agree that removing id would make early game easier. ⋅ You think this outcome is ok. The rest...

[This discussion happened in another thread , but I think it belongs better here.] tealizard dissents from this entire approach of using Sequell queries to determine game difficulty: Incidentally, the plots from the other thread are essentially a souped up version of the Sequell query arguments tha...

My proposal is to remove all identification, and make all items identified by default. This would mean consumables, weapons, armour etc.

I am not opposed to identification of some subset, but I don't really see any advantages to this proposal.

One should be careful of generalizing from veteran players' experience. Almost 60% of all characters die on D:1-4. That "should" not happen, assuming reasonably competent play. One should not be designing game difficulty based on hypothetical optimal play, but rather how people generally ...

One should be careful of generalizing from veteran players' experience. Almost 60% of all characters die on D:1-4. That "should" not happen, assuming reasonably competent play. One should not be designing game difficulty based on hypothetical optimal play, but rather how people generally p...

I actually made no argument about the shape of the "ideal difficulty curve" (I will give my thoughts on that topic in another thread). I merely made the observation that the early game is hard, indeed much harder than the later game. I inferred from this statement, that you were trying to...

Congrats on your win.

Blinking has a failure chance when being constricted, but if you try it 2-3 times, you will generally succeed. Teleportation chance doesn't depend on constriction, AFAIK.

I also actually question your implied premise -- that the early game *should be* close to as as difficult as the later game. In fact, I think that would make the game much less appealing. I would argue to most people it *feels better* if their character, once established and leveled up somewhat, fe...

The XP curve in the game is very weird. There's also another mechanic in the game: skill points . The curve for the skill points depends on character XL, so it is inherently filled with breakpoints. Also, the formula for dependence on XL is this extremely weird function (the horizontal axis should s...

There have been many posts before about removing identification. I won't rehash all the old arguments, but rather provide some new ones, based on my recent experiments. As you can see from my plots in this thread , the early game is very hard (relatively speaking). This proposition is not seriously ...

Keep in mind that you are playing 0.24 (trunk?); however, I doubt that food generation has been changed significantly since 0.23.

You could also have missed a food vault somewhere. There's that vault with the bees and there's another vault with the jellies.

You can use objsta t to check these suspicions. In 0.23, in a 3-rune game with Swamp and Snake, about 124 rations are generated on average. From your action table, you ate 109 rations and 8 rations existed in your inventory at the end of the game. This number presumably includes the rations you got ...

Character looks fine. Don't die due to hubris.

You can train Axes past 18 (which is min-delay for broad axe). The extra training increases damage, and more damage is always good.

High priests are probably much tougher to play with than warlords, but if you give them plate armour to wear, they will probably not be that squishy. Wiki says they have about 70-80 HP (maybe, with enough training, they'll reach 100 or so). Of course, it's not "optimal" (or even fun), but ...

I didn't calculate it, but I highly doubt that the game becomes much tougher after Lair, given the shape of the other graphs. All indications are that the game keeps becoming easier.

Some more plots, this time of the difficulty of each floor. Dungeon only goes till D:8 because otherwise the calculations for randomized Lair entrances become too hairy. I plotted difficulty of each floor from D:1 to D:8 (later floors become problematic because of randomized Lair entrances)....

You have a +4 vampiric executioner's axe, which is a really good weapon. 60% of the time you hit a living monster with it, it will heal you based on how much damage you did. An executioner's axe does a ton of damage, so you'll get healed a lot in-combat, which makes it very hard to die. If you are w...

Finish Orc mines and Dungeon before doing the Lair branches. Snake is a bit easier than Shoals. Keep in mind that Nagas are slow and snakes are fast, so split them apart to make them easier to fight. There's also plenty of cover in Snake and very little cover in Shoals. Watch out for Guardian Serpen...

Thanks for replacing Lab with Gauntlet. It is great. More importantly, it's much better than Lab, in every possible way.

Also, I've said this elsewhere: exploration-based piety is bad. Nemelex should give piety on kills. Unlike some gods like Ash where exploration-based piety could be justified for "flavour" reasons, there's no real justification for exploration-based piety for Nemelex. From what I understan...

Btw, about the discussion about degeneration above: I think it's a good card. The polymorphed monster typically has lower HD than the original. Also, it has great potential for lulz. From one of my old games : You draw a card... It is Degeneration. The death yak evaporates and reforms as a storm dra...

Also, it's not clear why card power uses both invo and piety. power = (invocations * 9) + (piety * (invocations + 25)) / 27 + piety * (3/2) It should only use invo. Piety is a consumable, invo is a training choice. Usually, when I want my god to do more powerful things, I train more invo. That behav...

Good to see goldified decks from Nemelex. Not sure if I quite like the hard cap on deck sizes, but maybe it's ok. From a little bit of code-diving, it looks like gifts from Nemelex increase the gift timeout whether you receive cards or not. If I understand the timeout mechanic correctly, this means ...

From the graph above, we can see that there are roughly two parts of the graph: from turn 1-1000, and from turn 1000 onwards. Here are zoomed-in graphs for the two periods, with significant events labeled. (This is for all players; the tenpercenters graph is similar). Keep in mind that these are all...

Oops double post.

Just go to Orc if you''ve cleared it and spawn some jellies; you'll hit max piety almost immediately.

A lot depends on what god you choose. Good gods for new players are Trog, Okawaru, Makhleb and Yred. If you want to play with Trog, you might want to start as a Berserker instead of Fighter, since you get to worship Trog from the start. Minotaurs are good with all weapons and you can use shields if ...

There are labels on the horizontal axes -- you just have to scroll a little bit since the image is a little too big.

I created plots for difficulty level based on turns. The turns are on a log scale. Data under the spoiler tag. I used queries of the form: "!lg * !boring 0.22 / !won turn>50 turn<76" The left end of the graph starts at turn 25. I used small bucket sizes at low turns and bigger buckets at h...

Another idea could be to calculate difficulty in game turns, instead of XL. Assuming a player plays at roughly the same rate throughout the game, we could then see how "smooth" the difficulty level they experience, as they play the game. The relevant Sequell query would be "!lg * !bor...

First derivative of difficulty (the data is underneath the spoiler tag): All players XL Prob of dying at XL Prob of surviving to XL+1 Probability of dying at XL given that you’ve reached XL First Derivative of Difficulty 1 14.26 85.74 14.26 2 8.49 77.25 9.90202939118265 -4.35797060881735 3 12.25 65 ...

How difficult is the game as we go along? What is the relationship between experience level and difficulty? This question has obvious implications for the XP curve in-game. We can try to answer the question using this measure: Difficulty at experience level XL = Probability of dying at experience le...

There's one idea I have for consumables which might reduce the associated power spiral. Currently, consumables are generated based on exploration. I have discussed elsewhere why exploration-based piety is bad, and similar comments apply here. What if we generate consumables only on monster kills? Th...

Did anything come of this proposal? Seems to me like it had a fair bit of interest at the time.

I am not opposed to complicated formulae which give better results than simplistic formulae. Crawl has tons of different mechanics, and sometimes complicated formulae better capture the various interactions. However, we should have in mind an objective. If we have an idea of what the XL curve is try...

The player XP need to attain various levels has the following comment as preamble: // Note: For historical reasons, all of the following numbers are for a // species (like human) with XP aptitude 1, not 0 as one might expect. // Basic plan: // Section 1: levels 1- 5, second derivative goes 10-10-20-...

Monster HD is a fundamental mechanic in the game. However, there is only a very loose relationship between monster HD and how dangerous it actually is. This fact is recognized by the game when it tries to calculate the XP a monster gives on death ("exper_value"). See this function. Now, it...

Just ignore riposte and play as you normally would. Long blades are fine as weapons. They have good damage and there's a fair variety of one- and two-handed versions. They're also fairly common. If you're playing with Dithmenos, keep in mind that long blades cross-train with short blades, and you ca...

I don't really see what's special about Abyss and Hell. The "danger level" would definitely depend on status effects on the character. Whether lignification is more "dangerous" or less "dangerous" depends on what's on the screen, the character's offence and defence. Of ...

It looks that you can cast any spell in the "library", not just the memorized spells.

Seems ok. It costs a ton of piety and Sif piety gain is pretty slow to begin with.

Once again, I'd like to suggest that there is no perfect way to create score, and while we maybe can refine the parameters of the one best scoring system, it will never satisfy everyone. On the other hand, introducing multiple systems of scoring could. Indeed, the aim of the scoring function outlin...

In a game with RNG, the influence of luck is inevitable. That does not mean that the role of time is useless. Quite the contrary: the time taken gives a lot of information on how the game went. To be (more or less) comparable, two runs would need the same game seed. Indeed, since now game seed has b...

Yes, I agree that tension is a made up measure. To be fair, the "danger level" would also be some kind of measure based on enemy HD and character XL, with a bunch of hacks in the middle. The measure needs to be easy to calculate. The aim is to approximate actual danger as well as possible....

[The following is related to this thread I created. However, this idea is completely stand-alone] Suppose we want to calculate: "how much danger is our character in right now"? It is difficult to calculate this exactly. DCSS damage formulae are insane, and there are all kinds of interacti...

* When Jelly dies, you receive a piety/200 chance of increasing your satiation by 70(Only jelly. It doesn't apply to other slimes). * When Jelly dies, you receive a piety/200 chance of recovering some MP(Only jelly. It doesn't apply to other slimes). * When Jelly dies, you receive a piety/200 chanc...

Just as an idea: one mechanic which could be used instead of the "heal on kills" is the Bloodborne "rally" mechanic. Whenever you take damage, for a short time afterwards, if you damage a monster, you regain some HP. This mechanic promotes a fast-paced style of play. Bloodborne a...

Search found 2166 matches