Game Data - Please help Collect

Back from page 18!
Three and a half months later I finally managed to spade this on my own for a bit (not that there's much to do while waiting for bug fixes).

One of my earlier theories was proven right yesterday. Having racked up some 800 data points, I graphed them, and the result baffled me for a second:


5, 6, 7 and 8 are damage numbers that pop up when you hit monsters; the vertical axis is how many times each number came up.

"Huh", I thought. "That's surprisingly skewed". Then I decided to graph the same pool, but this time summing up every odd and even number: 1 & 2, 3 & 4, 5 & 6, etc. So instead of visualizing [5, 5, 6, 5, 5, 6, 7, 7, ...] I tried [10, 11, 11, 14, ...].

The result had vanquished any doubts I had by that moment. An almost perfect bell curve:


This is the same pool, mind you. In fact, it now has twice as little data points - about 400, yet it's closer to the mathematical function that's supposed to be working here.

In retrospect, this was only logical. Individual numbers produced by Commando's Double Tap don't follow statistical rules for independent roll results - because apparently they're not exactly independent. The first and second hit in Double Tap affect each other, but if you sum them up you can get rid of that influence.

What conclusions can be drawn from this?
I suspect that every time you use Commando's basic attack, the game rolls for damage of the full attack (both halves, 2x60%) before showing anything to the player. Then some other algorithm divides this result into two separate attacks, which are then applied to the monster (and both of which, I assume, roll to proc any available effects).

Tomorrow I'll do some more recording, this time of Double Tap on level 2, and make a similar thing. If their mean values differ by 3, I'm right about how DT works and the universe is safe again.

This link is still up, if anyone's interested.

 

http://risk-of-rain.wikia.com there is this wiki here is it somehow possible to combine them together?

 

I'm well aware of this wiki, but the format doesn't lend itself to this kind of lenghty calculation stories. I'll put some concrete, practically useful numbers when I have them.

 

Instead of doing what I said I'd do, I decided to take a jab at character speed.

I decided to run the following routine: get a dried lake stage, get to the leftmost point on a ledge up to a wall and walk back, measuring the time from the moment the character starts moving to the moment it starts falling from the ledge.

Here's what I got:
7.93 s: Han-D
7.94 s: Mercenary
7.96 s: Sniper
8.02 s: Miner
8.08 s: Acrid
8.50 s: Enforcer
8.55 s: Bandit
8.62 s: Huntress
8.66 s: Commando
8.74 s: Engineer

Now, there is a number or measurement errors. Firstly, the recodring wasn't perfect and got a bit choppy, so the movement might've started a split second earlier than in the recorded clip every now and then. Secondly, start and end of the measurement was done by me tapping the stopwatch app, so there's a delay in eye-to-brain-to-finger reaction, which is not always consistent as I am human. Lastly, character sprites might have different width, meaning not all of them, having come up to a wall, necessarily started from the exact same spot.

With this much noise, any difference less than 50 milliseconds can be safely disregarded. Probably even 100 ms.
I'm fairly sure that Han-D, Miner and Acrid all have the same base speed. They're all melee fighters, and it would be logical to give them a boost of some sort to compensate for range. Sniper is kind of in-between, I'll have to do some more measuring. Bandit, Engineer, Huntress and Enforcer must have the same base speed too, as evidenced by some local co-op letsplays back from beta. Sadly, I don't have a controller so I can't run local co-op by myself (and good luck getting any strictly simultaneous actions in online MP with a 250 ms latency).

EDIT: slowed the videos to 1/8 speed and measured again. Yeah, melee + sniper seem to have the same speed that's rougly 6% higher the ranged classes have.
EDIT: added the jedi ninja mercenary.

 

Update: gathered 200 data points on 2nd-level DT. Average DPA values between the two graphs differ by 3.05, which I believe to be 3 + statistical error.
This means that, when hopoo said the Commando gets +3 damage per level, he was talking about the sum of both 60%-base-damage halves.

However, not all is well in the state of Denmark.

DT clvl 1 __ DT clvl 2​

Instead of even columns of 15 and 16 there seems to be an unusual spike of 15 damage per shot, almost as if the game wanted to flip the bird at me. We'll see who laughs at the end of the day, game.

Also, if you want to help, you can walk around the first stage and let basic lemurians bite you, note down how much damage their attacks do, and post your findings here. Remember that enemies get a boost every 90 seconds, which presumably means attack power, so either quit as soon as the timer hits 10, or make sure to divide your notes (0:00 - 1:29, 10 - 2:59, 3:00 - 4:29, etc). You might want to make a copy of Save.ini and Save_backup.ini so that an accidental death doesn't screw up your stats screen.

 

Did some gathering on Riot Shotgun out of curiosity. It's considerably stronger than Double Tap, I'll tell you that.

It's all fine and dandy, but we want pictures, Parker, pictures of Spiderman. Err, I mean, practical results. One such result would be the average DPS of a given ability - frankly, that was the goal from the very beginning.
But in Risk of Rain there is no such thing as seconds, or time. How does one derive ADPS in a game so dependent on frame rate? Simple - by moving to another parameter: DPF, or damage per frame.

When the game runs smoothly at 60 FPS, my Commando performs 156 Double Taps in a minute. If we take calculated damage per attack from the spreadsheet, that's 156 * 12.4661, or 1944.7116 (against a lemurian) damage in a full minute. Divide that by 60, and you get DPS - 32.41186 points of damage per second.
If your PC can give RoR all the resources it wants, that's the only number you need to know. Well, it probably is the only number you care about anyway, but for the sake of versatility I'll divide that by 60 again and get 0.5401977 - damage per frame.

Stay tuned for a damage comparison between Commando's handgun and Enforcer's trusty shotgun!

 

I would help if I wasn't on a trip, and in the middle of Internal Assesments, and college applications.