Calculating Work Success

[Watermelon Cat]

I am curious if anyone else has tried to prove out work success rates. I recently decided to give this game another shot, and am enjoying the current version a lot more than some of the earlier ones.

At current though, I keep having my days get ruined (often right at the end) by silent orchestra. It gets a meltdown at the wrong time and there doesn't seem to be a reliable solution for it.

Suppression really isn't an option in my current run, as I had a rather high casualty day during one of the core suppressions, and I just want to get the gear out of a few particularly nasty abnormalities (SO included) before starting a new run. I have a few high level staff with good gear, but not enough to suppress it.

So, my current goal is twofold.

1. Just keep working with it and evacuating the department so it can do its thing early in the day to get enough PE to get the gear, and the energy reset doesn't matter much.

2. Optimize my chances for not having the thing break out on my at the end of the day in the event of a pair of unlucky meltdowns. (which has happened to me twice now.)

Toward that end, I have started doing some analysis on the work results and am finding that with the same employee and the same penalty percent, I am getting wide ranging results. For example, with the main employee I am using, applying attachment with a work success of 116 at 30% penalty, I have gotten results ranging from 18 to 9.

So, whatever the calculation is, there is a lot of RNG in it.

The wiki has something kind of unclear on it:

Every 1 point of success rate equals 0.2% chance of the bonus to Success Rate

So, if I am understanding that correctly, this means that my 116 success rate stat equates to about 23% higher chance of a positive result.

----------

What I am not clear on is exactly how that calculates, but I have a few thoughts on the subject:


1. I am assuming the abnormalities have a hidden stat of some sort which indicates their difficulty. I base this off of the fact that I can very easily and consistently get a high number of PE boxes out of Nothing There, but with Censored, I usually end up somewhere in the middle.

2. I am not clear on how multiple stats play into it. The wiki suggests that success is based off of Temperance, but if I have an employee with a strong fortitude doing instinct work, it seems to affect the success rate as well. I could be wrong here, but there seems to be a correllation.

3. I am not sure if there is a set penalty or bonus applied based off of the work type preferences. IE, is very low a set penalty percentage, and very high a set bonus?


Anyhow, I am curious if anyone else has spent time trying to figure this mechanic out. Ideally, I am hoping to be able to mathematically predict where my best chances are for a particular result.

[elipod]

First time, they did big interface overhaul. Succes chances in "work favor" table and job assignments were dispalyed with actual percentage, then they were obfuscated with "high/common/low" words. But I think, values didn't change much, if at all.

I don't know exact conversion, but I think "common" covers 40%-60%, "high" is 60%-80% and "low" is 20%-40%.

Besides values from "work favor" table, temperance adds 0.2 succes chance to any work. So if you had, say, 50% for Attachment V, you'd see 70% sending agent with 100 temperance on actual job, because end value would be: 50+100*0.2+All other bonuses and maluses.

As for end result randomness, it rolls separately for each E box, classic bell curve distribution, extremes are rare, but it's still possible to roll all good or all bad results.

Honestly, I gave up on Orchestra: farmed his gear and restarted.

[Watermelon Cat]

Thanks Elipod!

That is very helpful. It does make sense that they are bucketed out in 20% groups since there are 5 of them. I did finally get Silent Orchestra's gear and completed the day, so I am not sure how much I will need it.

I did notice that the work favor changed with agent stats, so I am guessing the game does most of the math for the player there.

From 30 tries with the thing, it bucketed out as follows for Me:

Common: 6 Bad, 9 Good, 9 Neutral
High: 1 Good, 1 Neutral
Low: 1 Good, 1 Neutral
Very Low: 2 Bad

I'll probably give it some more thought and see if I can come up with a useful formula for optimization, but without precise numbers it would be pretty rough I think.

[Watermelon Cat]

I went ahead and worked this out mathematically in detail. Specifically, I built a tool in Excel, which thankfully has a formula for just this sort of thing, and used it to calculate the odds for each percentage point from 0 to 100% success rate. Of course, I don't think a lot of these are actually possible within the game, but it shouldn't effect the results.

So, if we take 30 trials and consider the results of 13 to 18 successful results (resulting in neutral for Silent Orchestra) and then bucket out the results into 20% buckets, it comes out as follows:

Very Low: 0% to 0.31%
Low: 0.49% to 41.32%
Common: 45.48% to 54.77%
High: 50.82% to 0.95%
Very High: 0.95% to 0%

The absolute best percentage of work success for getting a neutral result is 52%, which has a 72.69% chance.

From a practical standpoint, it is hard to identify a specific percentage, but if we assume we are correct in the percentile bucketing being in 20% groups, we can theoretically identify at least roughly where we are at based on the repeat work penalty.

So, if you have an agent who transitioned from High to Common when it hit -6% penalty from the first work, we would know that they are within 6% of 60% success rate.

----------

So, from there, we could accurately say that the agent would be in the proverbial sweet spot for the next two works with the -12% and -18% penalties, since that would put them somewhere around the mid point of the common range where the bell curve peaks.

This at least gives a >50% chance of making it.

The other practical observation from this is that all other factors aside, a common work favor is the best option when selecting an agent to work on it.


Anyhow, I have the gear at this point, so I am all set for the time being, and I have my Excel tool in my back pocket for other sorts of analysis.