Data science investigations into the mechanics of the world's greatest role playing game

[Zardnaar]

You know what I most admire about Zard? His ability to pat himself on the back with sufficient vigor to fracture his arms, and then to still be able to type forum posts. Perhaps he's really Thri-Kreen?

Hey I just remember a lot of people basically saying "Zard you're wrong and off our rocker" and all theses years later the stuff I pointed out just after release has made it into the various guides people write or numerous threads about certain stuff being OP/broken.

 

[Oofta]

Hey I just remember a lot of people basically saying "Zard you're wrong and off our rocker" and all theses years later the stuff I pointed out just after release has made it into the various guides people write or numerous threads about certain stuff being OP/broken.

Or it could be that in practice it doesn't really make that much difference depending on your game. It's certainly nice (perhaps too good) but it's greatly dependent on situational factors and class. Have a lot of attacks, advantage on a regular basis and low AC monsters? It's great. Otherwise? It's okay.

YMMV of course.

 

[Harzel]

Too many simplifying assumptions to be useful.
Too enamored of technique and jargon to be credible.
Also, "melle compbat". Really.

 

[jgsugden]

Oh, wait, is the thread where we'll finally reach universal agreement and a solution to the question of the -5/+10 feats? WONDERFUL. IAMSOFRIGGINGGLADWE'RETALKINGABOUTITAGAIN!!!!!

There is so much to this game... an analysis like this is like tring to evaluate a television series... by looking at 20 pixels on one frame from one episode ... without even having an agreed upon criteria for what you're trying to determine.

 

[Stalker0]

The more interesting conclusion has that high level fighters can take on crs 150% above their level. That follows my intuition based on experience that the Cr system starts to break down at higher levels

 

[abirdcall]

The more interesting conclusion has that high level fighters can take on crs 150% above their level. That follows my intuition based on experience that the Cr system starts to break down at higher levels

These are for solo monsters though right?

CR in 5e is different than CR in 3e.

It's not 'a party of 4 at level X can take on a creature of CR X.'

It's just a guideline that the creature is powerful enough to take out a character. Think of CR as a warning.

A CR 5 creature is an easy encounter for a party of 4 level 5 characters. 2 of them though is a deadly encounter as their force is multiplied.

 

[squibbles]

So, I may have misunderstood but, as far as I can tell, they just simulated a bunch of random draws, regressed the outcome of said draws on 6 variables, and then used a machine learning tool to optimize the simulations (they are pretty vague about all of this).

Those 6 variables are AC, HP, initiative, #of attacks, to hit modifier, and average damage, and yet... they appear to have simulated wizards. I'm shocked that the wizards underperform in the simulation; wizards, after all, are known for their high AC, HP, and #of attacks. /sarcasm

underspecified statistical model is underspecified

 

[Sorcerers Apprentice]

Of course when you put numbers into a mathematical model it will tell you that big numbers are important.
I don't see how you can use data science to evaluate creature abilities that need to be used intelligently to have any impact...

 

[squibbles]

I don't see how you can use data science to evaluate creature abilities that need to be used intelligently to have any impact...

Strongly agree!

Though I don't doubt that a machine learning algorithm could identify the optimal moves of any permutation of creatures and PCs if it included complete information. Things like AlphaGo, that do this sort of thing, already exist.

The d20datascience article clearly lacks a lot of relevant information.

 

[GreyLord]

I think it is accurate to a point. Without trying to break the game...I think the conclusions are ABSOLUTELY correct.

Due to the effects of bounded accuracy and purposefully limiting how well one can hit against a creature, the amount of damage you do is probably the most important factor in KILLING (as opposed to other tactics of getting past like talking to it) monsters in 5e.

However, the difference between a Fighter that hits with a +11 - +14 against an AC of 21 will hit probably most of the time or 2/3 of the time. If they deal 20 points of damage and have 5 attacks per round after three rounds they should have done around 200 points of damage. Not bad.

A Fighter that hits with a +20- +21 (+6 Prof, +9 STR, +3 Magic, +2 Bless or Proficiency Die somehow) will hit 100% of the time. If they do 15 points of damage a round they will do 225 points of damage per round.

So, in this scenario, MORE damage is pretty important, but even if they do less damage per hit, because they hit more often, they end up doing MORE damage.

I would say the Damage and to Hit are directly proportional in 5e.

For the above scenario to balance out the Second Fighter would need to do 2/3 the damage of the First fighter (so around 13 Damage per round gets it close)...as that is proportional to the 2/3 to hit that the First fighter has. In that light, damage and to hit % might be directly proportional in how much they affect combat and taking down enemies.

For those who go out of their way to try to boost their To Hit or AC to the point that they will try to break the game if they can, I think the To Hit and AC start gaining prominence.

As such, in these instances, the To Hit or ability to hit a monster, from what I've seen is directly proportional to damage in relation to how effective it may be (at least to a point, # of attacks is ALSO highly important).