This quote from another thread got me thinking...
... and I have to say experimentation hardly ever came up in the games I played. To be honest, actually only once in one of our groups, when the Bjornaer who was supposed to design a spell for the covenant so we could do an important quest decided he would experiment on it so he could finish it in one season - he completely blew it in the first AND second season, and then ended up taking the whole year for it. So that was hardly incentive for the other players to try it, though most of those other players are no longer part of the group anyway.
I juggled the numbers around a bit before with a friend, mostly coming to the conclusion that a risk modifier +2 is best if you're hoping for a Discovery. But now I wanted to know all of the probabilities, so I did the (hopefully correct) math and came up with this...
... this obviously assumes that no Vis is used in the lab work, but all that does is up the chances of a Disaster slightly - in the end the probability of that will always stay below 10% and just take off the appropriate percentage of the "no extraordinary effects" event.
So as mentioned the risk modifier of +2 is best if you're aiming for a Discovery, but +1 turns out to probably the best overall choice - while you give up an about 12% chance of basically "nothing unexpected happens, all goes to plan", you in return get an increase of 9% (from 1% to 10%, which in this case is enormous) for the chance of a Discovery, while also slightly reducing the chance of a complete failure.
Side effects basically come down to 40% detrimental, 40% neutral, and 20% beneficial. Modified effects are highly dependent on the effect being designed and your intended use of it, so it's hard to put on an one-dimensional axis.
So I hope this will show some people what to expect from experimenting, and how to best achieve what they're hoping to get from it.
P.S.: I'm wondering if changing the experimentation bonus from "simple die" to "stress die (no botch)" would increase the attractiveness of experimentation - on one hand you might also get nothing from it, on the other hand you have that very small chance of a huge bonus. This also serves to emphasize the gambling character of experimentation.
P.P.S.: If my math is wrong, feel free to correct me. Did this after a night of being sick without getting any sleep, so there might be glaring mistakes in it without me noticing it.