The Fidgeting Writer - doing the maths (SPOILERS)

Fallen London
101

Yay, another post in this wonderful thread!

So, i gather from the extensive statistics that the chances are as follows (rounded to nearest integer)

Level0 - TT=>SDV = 70%
Level1 - SDV=>GSL= 70%
Level2 - GSL=>DD = 70%
Level3 - DD=>RRB = 60%
Level4 - RRB=>LHFW = 60%
Level5 - LHFW=>LBG = 55%
Level6 - LBG=>CS = 55%

The cumulative chance to convert your Tale of Terror into a Coruscating Soul would equate to 3.74%.

Let’s do the math for 100 Tales of Terror
The total turns spent to get 3.74 CS would be 296.75
The total Item cost of this conversion will be 335.51 Echos
The total revenue would be 1167.27 Echos
The net profit would be 831.76 Echos
The profit/turn would be 2.80 Echos
Cost of brain-damaging Tedium = FW Profit per Turn - Normal profit per turn

While this seems a fine and dandy way to spend your turns, the calculation is considering you have all conversion items lying around. If you need to grind for the items you need for conversion, the numbers reduce drastically because you need to equate for those turns as well in your overall profit/turns calculation.

In my case, I have to convert all needed conversion items from Cryptic Clues (about 30 per turn), which makes my profit/turn a respectable 1.26 Echo per turn for a CS.

*EDIT: my math was broken due to recursive cost error: a measly 0.51 Echos for obtaining a Coruscating Soul because I have to spend so many turns grinding for them. In fact, in my case, it’s better to sell them right off the bat for 0.60 Echos of hassle-free profit.

I know the Forgotten Quarter Expeditions bombard you with Tales of Terror so I’ll probably put sink those in the FW once I have accumulated them anyways :)

Needed items Overview for 100 conversions starting from TT (rounded to nearest Integer)
100x Tale of Terror
70x Vision of the Surface
98x Correspondence Plaque
70x Brilliant Soul
21x An Identity Uncovered
13x Extraordinary Implication
7x Mourning Candle

Check the turns you need to gather all the above and add those turns to the Total turns spent and use those a the divisor for the Net Profit to get your own overall profit/turn.
I am very interested in the profit/turn numbers including grind-time of other players, please post them if you are willing to share them.

Source Spreadsheet
.
.
.
.
.
edited by dharthoorn on 9/12/2014

102

I get my Tales of Terror in the Flit. Probably not the best way of getting them, but it’s pretty simple and the same storylet can be used for a lot of the other necessary items.

103

It’s probably more efficient (though not as effective), to simply gather 3 x Memory of Distant Shores from the PoSI-challenges on The Seekers of the Garden and The Cities that Fell. At least that’s 1.5 Echoes/Action, and then those memories can be side-converted 50 at a time to Tales of Terror!! via Brilliant Souls. 50/3 + 2 Actions for 52 Tales gives 1.39 Echoes/Action. I am a big fan of side conversions. :)

104

thefts in flit are a lot more straight forward, but i usually get my tales from converting cryptic clues–>appalling secrets-> tales of terror. and the clues are from the university/flit investigating flip trick. might not be as efficient as side 50-51 side-conversions unless you mass convert thousands of clues at a time?

105

Could you show your math for this? I already have a huge stockpile of Clues ready to convert to Secrets then Tales. AFAIK the next best ppa grind is the Box, this just has to beat that.

106

Luck’s a Narrow quality, so there are no 55% odds. The actual probability of getting a Coruscating Soul is more probably around ~2.5-3%; 70%/70%/70%/60%/50%/50%/50%, though the last one might be a 60 (it could go either way is the descriptor for either 50 or 60% odds, the estimated odds for looking through the lens over ~200 attempts are somewhere near 55% so it’s hard to tell one way or the other.)

107

Hmmm, that’s certainly weird, because that’s not what These Numbers are saying. It seems players are hitting 70/70/70/60/55/55/55 pretty consistently. Perhaps the values the devs used are custom range instead of a preset table?

Sure, I have Uploaded Here.
I did fix an error on a recursice cost calculation and my average turn yield is significantly higher (1.26 E) than the 0.51 E for a CS I quoted earlier.

Notes on the spreadsheet;

  1. Data fields in Red are used as base data and can be modified as you like. So you can enter your &quotBatch Size&quot field and &quotCC per turn&quot and all the turns/cost data will auto-calculate.
  2. The entire turns calculation assumes you a side-converting Tales of Terror or Journals of Infamy from Appaling Secrets in turn gotten from mass-converted Crytpic Clues. There is no functionality (yet) to check if TT’s can be gotten faster. I still need to put a formula that takes the more efficient of the 2.
  3. Alternately you can just manually edit the data fields for turns next to each component and the auto-calc will work just fine.
  4. Side conversions yield 0.50 E in items per conversion, and I have not incorporated that gain into the spreadsheet (yet).

Any improvements welcome.
edited by dharthoorn on 9/11/2014

108

Dharthoorn, I notice that you have the cost of a Mourning Candle down as 12.5 in Cell D37, where it should actually be 2.5.

109

D’oh! Fixed. Good catch.

110

No, that’s not how Narrow qualities work and it’s not even accurate anyway; over nearly 4000 attempts, it’s more like 70/70/70/61/53/49/55. Narrow quality odds increase in increments of 10%, so 70/707/70/60/50/50/50 seems most likely.
edited by Spacemarine9 on 9/11/2014

111

The maths for this stuff has always been too confusing for me, so I never managed to get it right. Until today! Thanks to the obvious idea of checking to see what it would take on average to get a single Coruscating Soul, I can now present this table-turned-into-something-this-forum-can-handle:

(38.87 Actions + 38.87 x Tale of Terror!!) x 70% = 27.21 x Sense of Déjà Vu
(27.21 x Sense of Déjà Vu + 27.21 x Vision of the Surface) x 70% = 19.05 x Glimpse of Something Larger
(19.05 x Glimpse of Something Larger + 38.10 x Correspondence Plaque) x 70% = 13.33 x Deal with a Devil
(13.33 x Deal with a Devil + 26.67 x Brilliant Soul) x 60% = 8 x Room Number at the Royal Beth
(8 x Room Number at the Royal Beth + 8 x An Identity Uncovered) x 50%= 4 x Last Hope of a Fidgeting Writer
(4 x Last Hope of a Fidgeting Writer + 4 x Extraordinary Implication) x 50% = 2 x Lens of Black Glass
(2 x Lens of Black Glass + 2 x Mourning Candle) x 50% = 1 x Coruscating Soul
Total: (312.50 - 100.43) / 112.46 = 1.89 Echoes/Action

Hopefully this might help someone somehow. As for the 55% success rates, might it be possible that Failbetter Games has tweaked the luck challenges in the middle of the data collection? 60% during the first half and 50% during the second half would give those numbers, right? Not that the evil RNG wouldn’t try to trick us just because. Who are we lowly mortals to attempt to control our fates?

112

[quote=Spacemarine9][quote=dharthoorn]
Hmmm, that’s certainly weird, because that’s not what These Numbers are saying. It seems players are hitting 70/70/70/60/55/55/55 pretty consistently. Perhaps the values the devs used are custom range instead of a preset table?
[/quote]

No, that’s not how Narrow qualities work and it’s not even accurate anyway; over nearly 4000 attempts, it’s more like 70/70/70/61/53/49/55. Narrow quality odds increase in increments of 10%, so 70/707/70/60/50/50/50 seems most likely.
edited by Spacemarine9 on 9/11/2014[/quote]

Well, I took the liberty of combined my numbers with yours and this is what it looks like;
Level0 - TT=>SDV = 11296 Attempts, 8096 Success, 71.8%
Level1 - SDV=>GSL= 9876 Attempts, 6987 Success, 70.7%
Level2 - GSL=>DD = 4715 Attempts, 3347 Success, 71.0%
Level3 - DD=>RRB = 3363 Attempts, 2096 Success, 61.5%
Level4 - RRB=>LHFW = 2065 Attempts, 1107 Success, 53.6%
Level5 - LHFW=>LBG = 1108 Attempts, 585 Success, 52.8%
Level6 - LBG=>CS = 579 Attempts, 323 Success, 55.8%

From this combined data, 70/70/70/60/55/55/55 seems to be closer to the mark. No need to bicker over it, though. Perhaps the Devs can confirm the exact rates from the Sourcecode?

113

I dunno. 579 attempts really isn’t that many. It’s entirely possible that as the number of attempts increases that number will go down closer to 50%.

114

I think it should be possible to calculate an expected margin of error from the sample size, but unfortunately I don’t actually understand statistics enough to do that >.<

115

189 average ppa (assuming all the components on hand already) doesn’t sound that much better than the alternatives here. Maybe Fidgeting Writer is better as a source of Enigmas (compared to the burning of Incunables)?

116

[quote=empirimancer][quote=Gillsing]
Total: (312.50 - 100.43) / 112.46 = 1.89 Echoes/Action
[/quote]

189 average ppa (assuming all the components on hand already) doesn’t sound that much better than the alternatives here. Maybe Fidgeting Writer is better as a source of Enigmas (compared to the burning of Incunables)?[/quote]

That’s assuming the odds are 70/70/70/60/50/50/50, which they may not necessarily be accurate. At 70/70/70/60/55/55/55 the payoff is 2.73 E/A. At the exact combined statistics we have compiled to date from converting 11.296 TT the payoff is 2.71 E/A.

Then again, we’re talking about luck here and so statistics are only a very rough approximation of your mileage unless you are planning to do a 50.000 TT conversions run. This FW storylet will be a big gamble, no matter which way you swing it.

As Sara Hysaro already remarked, on the latter parts of the conversion string there is just not enough data to approach reliable accuracy so discussing percentages on the last 3 is fruitless.

And now, my head is spinning from going around in circles.

117

The statistics are just governed by the binomial distribution, so the standard deviation in expected percentages for n trials is just 100% / (2 sqrt(n)) (assuming the probability is closish to 50-50). So the 2-sigma error bars (approximately 95% confidence interval) on the measured rate for 579 trials is 4.1%. All of the last three conversions are consistent with a 55% success rate, but ever so slightly inconsistent with a 50% success rate. Therefore, if you think that the designers were just as likely to use a 55% success rate as a 50% success rate, then you should be inclined to believe that 55% is the correct number. However, if you think it more likely that the designers would choose a nice round number for the rate, then you should start to get a little worried.

–
edited by Guy Scrum on 9/12/2014

118

It isn’t a matter of a rate for the challenge itself, Luck is a predefined quality, a luck check in Fallen London is identical to a luck check in any world coded via Storynexus. Unless it isn’t actually a Luck check and they’ve been sneaky, and everyone has been inattenive and one letter is different, so it’s a Lnck check and we didn’t notice I’m afraid it’s simply the normal Narrow check :P

119

A quick calculation shows that producing the above data for the last three conversions is about 8400 times more likely given a probability of 55% than it would be if the probability were 50%. So it seems highly unlikely that the probabilities have been set to 50% this entire time. But, as Gillsing points out, the devs could have tweaked them mid-way through.

120

@Guy
You’re probably the most statistics-calculus savvy person that has checked in on this thread as far as I can tell. So might I ask you a favor?

What would be the best way to assign &quotconfidence values&quot to the derived averages?
If you feel particularly magnanimous, feel free to add your calculations to the &quotcalc&quot tab in the updated &quotSource Spreadsheet&quot from my post at the top of the previous page. (If the spreadsheet web-interface is too cumbersome to bear then just send me the xlsx and I’ll upload it again.)
.
.
.
edited by dharthoorn on 9/12/2014