Different probabilities of encounters

This is where you spade, shovel and sickle. Or is it reap?
Post Reply
User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Different probabilities of encounters

Post by Flolle » Sun Jan 27, 2008 8:29 am

As many of you know, there are some areas in which one or more encounters have higher probabilities to appear than all others. The Irritating Series of Random Encounters combat would be an example for this. Originally, I thought Jick made sure those encounters would happen twice as often by having two identical versions of the encounter be possible to roll. This way queue effects wouldn't reduce appearance rates.

But here comes the problem: While reading this thread and doing some test with the Area Simulator afterwards, it appears as though that at least the Astronomers from the Hole In The Sky don't follow this route and simply only have a twice as high appearance rate which in turn can and will get reduced by queue effects over time.

The implications of this are that turn or stat calculations for areas with such encounters cannot be safely calculated since we don't know how they are handled.

These areas are (italic means there are enough numbers from that area):
- hole in the sky(Astronomer)
- hunted bathroom (Don't Hold a Grudge or Having a Medicine Ball, depending on how you look at it)
- giant castle (wheel stat adventure)
- airship (Irritating Series of Random Encounters)
- south of the border (a chewy encounter)
- knob laboratory (Knob Goblin Mad Scientist)
- knob kitchens (Knob Goblin Chef)
- knob harem (Knob Goblin Harem Girl)
- degrassi knoll (Gnollish Gearhead)
- icy peak (Snow Queen happens less often)
- hippy camp on the verge of war (sergeant)
- frat camp on the verge of war (sergeant)

Since this is very turn intensive spading and I don't have any multis, I'm asking whether you guys could help me getting numbers for this. All you need to do is run turns in the above areas and post the number of all encounter appearances. If it is possible to skip a choice adventure, do not do it, since this would only complicate information gathering (with the only exception of the giant castle wheel). I'm guessing that 1000 turns per area in total are probably enough.

Also, I don't know what of this spading might be useful for QN's project to spade whether the queue somehow changed with NS13. So it's probably best if he could chime in on this himself.


Data:

Knob Kitchens:

Code: Select all

Flolle:
Knob Goblin Chef:          154(64.7%)   175(63.2%)   184(64.3%)   145(59.7%)   199(65%)     85(56.7%)
Knob Goblin Master Chef:    84(35.3%)   102(36.8%)   102(35.7%)    98(40.3%)   107(35%)     65(43.3%)
total:                     238(100%)    277(100%)    286(100%)    243(100%)    306(100%)   150(100%)

Kujjie:
Knob Goblin Chef:          127(63.5%)   208(69.3%)
Knob Goblin Master Chef:    73(36.5%)    92(30.7%)
total:                     200(100%)    300(100%)

Total:
Knob Goblin Chef:         1277(63.85%)
Knob Goblin Master Chef:   723(36.15%)
total:                    2000(100%)

Conclusion: The true appearance rates of the two monsters in the Kitchens are 67% and 33%. These appearance rates then in turn get affected by the queue and end up being in a 64.4% to 35.6% ratio (differences are due to a not big enough sample size).


Knob Laboratory:

Code: Select all

Knob Goblin Mad Scientist:           97(68.8%)   207(73.7%)   213(73.4%)   219(71.8%)
Knob Goblin Very Mad Scientist:      44(31.2%)    74(26.3%)    77(26.6%)    86(28.2%)
total:                              141(100%)    281(100%)    290(100%)    305(100%)

Total:
Knob Goblin Mad Scientist:          736(72.4%)
Knob Goblin Very Mad Scientist:     281(27.6%)
total:                             1017(100%)

Conclusion: The true appearance rates of the two monsters in the Laboratory are 75% and 25%. These appearance rates then in turn get affected by the queue and end up being in a 71.1% to 28.9% ratio. This ratio isn't really showing here, but this dataset was just confirmation for a bigger dataset with a total of 8000 turns.


Hole in the Sky:

Code: Select all

Astronomer:                 31(11.4%)    35(11.8%)    40(12.8%)    34(12.3%)
Axe Wound:                  26(9.5%)     25(8.4%)     30(9.6%)     30(10.8%)
Beaver:                     28(10.3%)    30(10.1%)    32(10.3%)    24(8.7%)
Box:                        26(9.5%)     31(10.4%)    32(10.3%)    27(9.7%)
Bush:                       23(8.4%)     27(9.1%)     32(10.3%)    26(9.4%)
Camel's Toe:                28(10.3%)    35(11.8%)    31(9.9%)     28(10.1%)
Flange:                     28(10.3%)    31(10.4%)    27(8.7%)     30(10.8%)
Honey Pot:                  20(7.3%)     19(6.4%)     34(10.9%)    25(9%)
Little Man in the Canoe:    33(12.1%)    36(12.1%)    32(10.3%)    32(11.6%)
Muff:                       30(11%)      28(9.4%)     22(7.1%)     21(7.6%)
total:                     273(100%)    297(100%)    312(100%)    277(100%)

Total:
Astronomer:                140(12.1%)
Axe Wound:                 111(9.6%)
Beaver:                    114(9.8%)
Box:                       106(9.1%)
Bush:                      108(9.3%)
Camel's Toe:               126(10.9%)
Flange:                    123(10.6%)
Honey Pot:                  98(8.5%)
Little Man in the Canoe:   133(11.5%)
Muff:                      101(8.7%)
total:                    1159(100%)

Conclusion: The Astronomer has a twice as high appearance rate with a rejection rate of 25% (or something equal in functionality). The queue then does the rest. Simulation showed that the appearance rate should be around 12.5% which is reasonably close to the numbers found. All other monsters seem to have the same appearance rate (9.7% from simulation), with maybe the exception of the Little Man in the Canoe combat, but that one is probably just a statistical outlier (that's just a guess though).

BTW, if an encounter roll rejection is only possible once, the appearance rate of Astronomers would be higher by 0.2%, so this dataset would need to be considerably bigger to see how rejections are handled exactly.


Icy Peak:

Code: Select all

Seal Lubber:
Knott Yeti:           71(40.6%)    55(40.4%)
Upgraded ram:         65(37.1%)    53(39%)
Snow Queen:           39(22.3%)    28(20.6%)
total:               175(100%)    136(100%)

Total:
Knott Yeti:          126(40.5%)
Upgraded ram:        118(37.9%)
Snow Queen:           67(21.5%)
total:               311(100%)



South of the Border:

Code: Select all

A Chewy Encounter(tamarind):           31(14.7%)    34(15.2%)    33(14.5%)    30(12.8%)    29(13.2%)    29(12.7%)    23(13.1%)
A Chewy Encounter(pickle):             29(13.7%)    29(13%)      36(15.9%)    36(15.4%)    37(16.8%)    33(14.5%)    23(13.1%)
A Chewy Encounter(jabanero):           36(17.1%)    31(13.9%)    24(10.6%)    22(9.4%)     27(12.3%)    27(11.8%)    23(13.1%)
A Chewy Encounter(lime-and-chile):     31(14.7%)    24(10.8%)    30(13.2%)    30(12.8%)    31(14.1%)    29(12.7%)    24(13.7%)
Finger-Lickin'... Death:               46(21.8%)    48(21.5%)    54(23.8%)    59(25.2%)    49(22.3%)    55(24.1%)    39(22.3%)
La Farmacia de Suenos:                 38(18%)      57(25.6%)    50(22%)      57(24.4%)    47(21.4%)    55(24.1%)    43(24.6%)
total:                                211(100%)    223(100%)    227(100%)    234(100%)    220(100%)    228(100%)    175(100%)

Total:
A Chewy Encounter(tamarind):          209(13.8%)
A Chewy Encounter(pickle):            223(14.7%)
A Chewy Encounter(jabanero):          190(12.5%)
A Chewy Encounter(lime-and-chile):    199(13.1%)
Finger-Lickin'... Death:              350(23.1%)
La Farmacia de Suenos:                347(22.9%)
total:                               1518(100%)

Conclusion: Most likely a 3(A Chewy Encounter)-1(Finger-Lickin'... Death)-1(La Farmacia de Suenos) ratio. Simulation showed that that would result in appearance rates of 52.6%, 23.7% and 23.7%, which is reasonably close to the found numbers. In this model, the A Chewy Encounter adventure is seen as one single encounter. Which version of A Chewy Encounter you get will be decided by a four-sided die roll (1d4 for the nerds ;)) or something of equal functionality (1-1-1-1 distribution).


Giant Castle:

Code: Select all

Keep On Turning:                         32(23.9%)    31(26.1%)    34(23.6%)    31(26.3%)    42(30.9%)    41(29.5%)    31(40.8%)    27(20.5%)
What are the Odds?:                      40(29.9%)    33(27.7%)    37(25.7%)    34(28.8%)    43(31.6%)    39(28.1%)    24(31.6%)    44(33.3%)
In a Black Room, with Black Curtains:    37(27.6%)    32(26.9%)    38(26.4%)    34(28.8%)    38(27.9%)    40(28.8%)    20(26.3%)    38(28.8%)
Outage, Brief Candle:                    57(42.5%)    54(45.4%)    69(47.9%)    50(42.4%)    55(40.4%)    60(43.2%)    32(42.1%)    50(37.9%)
Total:                                  134(100%)    119(100%)    144(100%)    118(100%)    136(100%)    139(100%)     76(100%)    132(100%)

Keep On Turning:                        269(27.0%)
What are the Odds?:                     294(29.5%)
In a Black Room, with Black Curtains:   277(27.8%)
Outage, Brief Candle:                   427(42.8%)
Total:                                  998(100%)

Conclusion: The distribution in the Castle is 2-1-1-1, which - thanks to the queue - results in a 43.5% appearance rate for the wheel stat noncombat and a 28.25% chance of appearance for each of the other two noncombats if the wheel is skipped every time.


Haunted Bathroom:

Code: Select all

Ascension 1:
Having a Medicine Ball:     52(34.7%)    54(35.1%)    44(32.8%)    36(27.3%)    41(27.7%)
Don't Hold a Grudge:        98(65.3%)   100(64.9%)    90(67.2%)    96(72.7%)   107(72.3%)
total:                     150(100%)    154(100%)    134(100%)    132(100%)    148(100%)

Ascension 2:
Having a Medicine Ball:     38(34.9%)    52(35.9%)    39(30.0%)    35(28.5%)    42(28.2%)    43(31.2%)
Don't Hold a Grudge:        71(65.1%)    93(64.1%)    91(70.0%)    88(71.5%)   107(71.8%)    95(68.8%)
total:                     109(100%)    145(100%)    130(100%)    123(100%)    149(100%)    138(100%)

Spent 200 turns somewhere else and then came back:
Having a Medicine Ball:     68(34.2%)    45(32.6%)    33(28.9%)    37(29.1%)
Don't Hold a Grudge:       131(65.8%)    93(67.4%)    81(71.1%)    90(70.9%)
total:                     199(100%)    138(100%)    114(100%)    127(100%)

Spent 400 turns somewhere else and then came back:
Having a Medicine Ball:     47(31.3%)
Don't Hold a Grudge:       103(68.7%)
total:                     150(100%)

Ascension 3:
Having a Medicine Ball:     52(31.1%)    51(35.7%)    44(37.3%)    46(38.0%)    38(31.1%)    34(30.4%)    40(30.1%)    42(28.8%)    54(35.3%)    54(35.3%)
Don't Hold a Grudge:       115(68.9%)    92(64.3%)    74(62.7%)    75(62.0%)    84(68.9%)    78(69.6%)    93(69.9%)   104(71.2%)    99(64.7%)    99(64.7%)
total:                     167(100%)    143(100%)    118(100%)    121(100%)    122(100%)    112(100%)    133(100%)    146(100%)    153(100%)    153(100%)

A different character:
Having a Medicine Ball:      6(26.1%)    22(31.0%)    34(33.7%)    28(32.2%)    30(29.7%)    30(27.8%)    38(34.9%)
Don't Hold a Grudge:        17(73.9%)    49(69.0%)    67(66.3%)    59(67.8%)    71(70.3%)    78(72.2%)    71(65.1%)
total:                      23(100%)     71(100%)    101(100%)     87(100%)    101(100%)    108(100%)    109(100%)


Total:
Having a Medicine Ball:   1265(32.0%)
Don't Hold a Grudge:      2682(68.0%)
total:                    3947(100%)

Muscle grudges:            707(26.4%)
Myst grudges:             1294(48.2%)
Moxie grudges:             681(25.4%)

Conclusion: The numbers look a little strange, but I'm not willing to throw anymore turns at this. Seems like a 2-2-1-1 distribution for Medicine Ball, Myst grudge, Muscle grudge and Moxie grudge in that order. Note that both Medicine Ball and Myst grudge are entered as two distinct encounters that are handled differently by the queue (same as the Astronomers in the pre-change HitS that were seen as two different combats by the queue). This would theoretically result in appearances rates of 33.3% for Medicine Ball and Myst grudge and 16.7% for the two left over grudes.


Airship:

Code: Select all

Irritating Series of Random Encounters:    36(19.3%)    45(24.6%)    38(22.0%)    40(23.3%)    30(18.5%)    35(19.7%)    43(23.9%)    36(20.5%)    40(23.1%)
Spunky Princess:                           28(15.0%)    22(12.0%)    28(16.2%)    26(15.1%)    23(14.2%)    32(18.0%)    24(13.3%)    31(17.6%)    24(13.9%)
Burly Sidekick:                            30(16.0%)    32(17.5%)    24(13.9%)    24(14.0%)    30(18.5%)    26(14.6%)    35(19.4%)    26(14.8%)    27(15.6%)
Protagonist:                               32(17.1%)    29(15.8%)    28(16.2%)    25(14.5%)    29(17.9%)    28(15.7%)    26(14.4%)    28(15.9%)    28(16.2%)
Quiet Healer:                              29(15.5%)    26(14.2%)    27(15.6%)    26(15.1%)    25(15.4%)    31(17.4%)    26(14.4%)    28(15.9%)    28(16.2%)
MagiMechTech MechaMech:                    32(17.1%)    29(15.8%)    28(16.2%)    31(18.0%)    25(15.4%)    26(14.6%)    26(14.4%)    27(15.3%)    26(15.0%)
total:                                    187(100%)    183(100%)    173(100%)    172(100%)    162(100%)    178(100%)    180(100%)    176(100%)    173(100%)

Total:
Irritating Series of Random Encounters:    343(21.7%)
Spunky Princess:                           238(15.0%)
Burly Sidekick:                            254(16.0%)
Protagonist:                               253(16.0%)
Quiet Healer:                              246(15.5%)
MagiMechTech MechaMech:                    250(15.8%)
total:                                    1584(100%)

Conclusion: Most likely a 2-1-1-1-1-1 distribution for Irritating Series of Random Encounters and the other combats. After queue effects, this results in a 22.5% appearance rate for Irritating Series of Random Encounters and 15.5% appearance rates for all other combats.
Last edited by Flolle on Sun Jul 06, 2008 4:40 pm, edited 45 times in total.

User avatar
QuantumNightmare
AFH
Posts: 1827
Joined: Mon Jan 29, 2007 10:28 pm
Location: Montreal

Post by QuantumNightmare » Sun Jan 27, 2008 9:13 am

Prior work on the subject:
MMGdW wrote:Well, who would have thought. I still have some logfiles from spicedxxcider from the kitchen (690 turns). It's from a time when the queue was still 3 adventures long. I had a look at these logs with the following result:

Kitchen (690 turns):
Master Chef 279 (40.4%)
Chef 411 (60.6%)

Assuming a suppression factor of 0.75 and a queue of length 3 one can now compare these to two scenarios:
a)two identical chefs and one master chef, all with equal probability
b)two adventures with unequal chances (2/3 vs 1/3)

In case a), in the long run, all adventures are equally likely giving a total chance of 66.6% for the normal chef and 33.3% for the master chef.
In scenario b) the normal chef has a higher chance and is thus more suppressed than the master chef. Numerical simulations shows that overall we expect the normal chef to appear 61% of the time vs. 39% for the master chef.

This means that most likely the kitchen contains exactly 2 adventures with chances split 2/3 vs. 1/3. Now before you try to optimize your queue usage, keep in mind that the queue is now 5 adventures long. This seriously delutes the effect seen here, because it is very unlikely to have only one of the adventures in the queue. The figure I got for the long queue is a suppression to 64%, which is rather miniscule.

This also means that the effect might be difficult to detect in the Hits, where the expected difference between the two scenarios is also about 2%. Well, maybe I can come up with a more usefull measure than the total encounter rate.

If anyone has date on the Hits, I'l gladly look at it.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Wed Feb 13, 2008 5:09 pm

Trivial Updates wrote:The Hole in the Sky no longer has two different (identical) versions of the Astronomer in it.

Radioshow Transcript wrote:There were basically 2 copies of the astromomer… were started with 6 constellations, then added 3 more, but we didn’t want to reduce the chance of encountering an astronomer. So we put in a 2nd astronomer with only a 50% chance of finding that one. This messed with the items that affected fighting a given monster – Harold’s bell, Transcendant Olfaction, etc. I’m not sure if this will make it more likely or less likely that you fight an astronomer… now it is 2 copies of the same monster, with 25% chance of rejecting it. We may need to adjust the odds later.

Kinda contradicting. But a test with the Area Simulator(adding this took, like, 2 min) showed that without a queue the radioshow version takes ~42 turns and with a queue ~38 turns, so I guess I'll go with it.

The radioshow quote is also interesting in the regard that I didn't know that it was possible for normal encounters to have rejection rates. Besides making things possibly more complicated to emulate, does anybody want to take a guess whether this might be part of the mechanism behind the new Friar's part 1 and Cyrpt (and Airship?)? As in slowly decreasing rejection rates the longer you stay in the area.

Brulak
AFH
Posts: 945
Joined: Sat Apr 14, 2007 2:51 pm
Location: Hamburg, Germany

Post by Brulak » Thu Feb 14, 2008 1:31 am

Flolle wrote:The radioshow quote is also interesting in the regard that I didn't know that it was possible for normal encounters to have rejection rates.


Teleportitis without plus sign? Prior to reading about this rejection percentage, I thought the game would just reroll an un-signed teleportitis.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Thu Mar 06, 2008 6:05 pm

Please disregard the conclusions in this post since it can't be safely said whether the numbers found on the wiki are valid or not.

While looking through this, I found some numbers that I could use for this project.



What I've found:


Hole in the Sky:
In a 300 turn dataset, Astronomers appeared 12.3% of the time. Not a big enough dataset, but I think it shows a trend. After testing this a little with the Area Simulator (I know, it's not the same as mathematical proof, but queue math is rather complicated, so this will have to do), it seems like the two Astronomers are handled the same by the queue. That means that whenever one of the Astronomers is in the queue, the queue thinks the other is in it too.

This is probably a matter of the queue saving the ID of a monster and both Astronomers having the same ID. Essentially, this would mean that the current implementation of the two Astronomers would be equal to one Astronomer with a twice as high appearance rate. Testing with that results in a 12.1% apprearance rate of Astronomers, or ~45.5 turns avg with a standard deviation of ~11.4 turns to see 6 Astronomers. That would partially explain the problems with this bounty.

As I noted, 300 turns isn't conclusive enough, but I have to admit that the numbers do seem to add up. Still needs more numbers.


Knob Laboratory:
In a >8000 turn dataset, Knob Goblin Mad Scientist appeared 71.1% of the time. This works out to a 3:1 appearance rate ratio between both monsters in the area (namely 75% to 25%) which then gets subjected to the queue. Testing said the expected outcome for the 3:1 ratio after queue effects would be a 71.1% appearance rate for Knob Goblin Mad Scientist.

The amount of data is more than enough, the numbers add up. This is conclusive.


Knob Kitchens:
966 turn dataset, more or less perfect 66.7%(Knob Goblin Chef) to 33.3%(Knob Goblin Master Chef) distribution. I'll go with the two monsters of the same kind approach here, by which I mean two Knob Goblin Chefs (which are handled differently by the queue) and one Knob Goblin Master Chef are in the encounter roll queue of the Knob Kitchens.

The dataset is big enough. Conclusive. Edit: No, not conclusive, a 7:3 appearance rate ratio is as viable. Final edit: This has been spaded by AFH members now. 64%/36% appearance rates were found which point to the real appearance rates being 67%/33% which then get subjected to the queue.


Icy Peak:
Kinda surprising, huh? Well, I hadn't realised that the Icy Peak has encounters with different probabilities, but it is the case. A 663 turn dataset which relatively conclusively showed a 38%(Yeti) to 38%(Ram) to 24%(Queen) ratio. As with the Laboratory, testing showed that a simple 2:2:1 appearance rate distribution would - after queue effects - come out to exactly the distribution that had been found through spading.

The dataset is big enough. Numbers add up. Conclusive.


Any questions? Did I do something wrong? Will you only believe me if I calculate everything by hand? >.>
Last edited by Flolle on Wed Mar 26, 2008 9:05 pm, edited 4 times in total.

greycat
AFH
Posts: 359
Joined: Sat Sep 22, 2007 9:42 pm

Post by greycat » Thu Mar 06, 2008 7:48 pm

Flolle wrote:Hole in the Sky:
In a 300 turn dataset, Astronomers appeared 12.3% of the time. Not a big enough dataset, but I think it shows a trend. After testing this a little with the Area Simulator (I know, it's not the same as mathematical proof, but queue math is rather complicated, so this will have to do), it seems like the two Astronomers are handled the same by the queue.


Unfortunately, I think you'll need to discard this data and start over. There was a recent-ish "trivial update" in which they announced the death of the second Astronomer.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Thu Mar 06, 2008 8:41 pm

greycat wrote:Unfortunately, I think you'll need to discard this data and start over. There was a recent-ish "trivial update" in which they announced the death of the second Astronomer.

I'm aware of that. See the third post in this thread.

The second Astronomer I'm referring to is the functional identical one that they talked about in the radio show which was done after the change. Also, there was a thread in the main forum in which Hellion more or less confirmed that two identical Astronomers do exist. I merely made a guess on how these two Astronomers are handled by the queue based on the encounter rates from the dataset and testing with the Area Simulator (which, I admit, isn't as good as doing the actual calculations by hand) on my part.

The dataset was posted on the 26th of February 2008, which was after the change to the Hole in the Sky (14th of February or earlier). Thus, I think the data is valid.

Sorry if I came off rather dickish, it's just that I pay very much attention to this encounter probability stuff, so it's very unlikely for me to overlook changes which are officially announced. If you notice other things, don't be afraid to post, it is very much appreciated! (even if I fail to convey that with my posts >.>)

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Fri Mar 07, 2008 3:52 pm

Flolle wrote:Knob Kitchens:
966 turn dataset, more or less perfect 66.7%(Knob Goblin Chef) to 33.3%(Knob Goblin Master Chef) distribution. I'll go with the two monsters of the same kind approach here, by which I mean two Knob Goblin Chefs (which are handled differently by the queue) and one Knob Goblin Master Chef are in the encounter roll queue of the Knob Kitchens.

The dataset is big enough. Conclusive.

Addendum: Well, maybe not totally conclusive. Appearance rates of 70%(Knob Goblin Chef) and 30%(Knob Goblin Master Chef) would after queue effects result in a 67.2% to 32.8% ratio. Any idea how these two can be differentiated without having to collect >5000 turns of data? :?

Edit: I did a little more reading, and it looks like some old data from days when the queue was 3 items long says that the 7:3 ratio might actually be what is going on in the Kitchens.

To be more precise, a 5565 turns dataset says that the appearance rates are ~64.2% and ~35.8% in the Kitchens. Testing said that a 7:3 ratio would turn into 63.9% and 36.1% with a 3 item queue. The data was posted on the 5th of September 2006, one day before the queue was changed from 3 to 5 items. Since this data is so old, I won't conclude that the 7:3 ratio is the mechanic in the Kitchens yet, but I think it is at least a pointer that something might be up here.


Edit edit: Since data collection in this thread showed 64/36 appearance rates in the Kitchens, I can only assume that either Yiab put in the wrong date on his data or the change to the queue was done a few days earlier before it was finally announced.
Last edited by Flolle on Tue Mar 18, 2008 3:12 pm, edited 1 time in total.

User avatar
KujjieKujjieKoo
Spade Ninja
Posts: 2240
Joined: Mon Apr 30, 2007 8:50 pm
Location: !Chicago
Contact:

Post by KujjieKujjieKoo » Mon Mar 10, 2008 12:49 pm

Some additional data on the Kitchens:

Test 1: 200 adventures in Bad Moon
Knob Goblin Chef (127) (63.5%)
Knob Goblin Master Chef (73) (36.5%)

Test 2: 300 adventures
Knob Goblin Chef (208) (69.3%)
Knob Goblin Master Chef (92) (30.7%)

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Mon Mar 10, 2008 1:31 pm

Ok, added data from Kujjie and me about the Knob Kitchens to the first post. The data contradicts the data from the wiki page that I linked (must either be an extrem outlier or faulty data collection), but it goes hand in hand with historic data from the Kitchens. See the second post in this thread as to why that is the case.

I think about two more data points of the Kitchens are needed (for a total of ~2000 turns) before the spading can be called finished on that location.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Sun Mar 30, 2008 8:29 pm

I just looked through my Castle data which I collected over the last few days and while the sample size has to get twice as big before I draw conclusions for that area, I can say something definitive on something else: Skipped choice adventures are in fact added to the queue, which is what had been theorized before. If it weren't the case, the Castle wheel would have occured much more often (>80% in relation to the number of other noncombats).

Yay for killing two birds with one stone! :)

User avatar
QuantumNightmare
AFH
Posts: 1827
Joined: Mon Jan 29, 2007 10:28 pm
Location: Montreal

Post by QuantumNightmare » Mon Apr 14, 2008 8:18 am

What is the probability of getting a quartet ballroom adventure at varying amounts of combat rate modifiers? I use a weird formula to predict what it will be in my leveling sheet, but I don't think it's very accurate. Can your area simulator run turns at all combat rate modifiers from +25% combats to -29% noncombats, and see what the quartet rate will be in each case?

We can then make an accurate and ugly quartet predictor to include in the leveling spreadsheet, since the queue itself isn't calculated explicitly.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Mon Apr 14, 2008 9:04 am

Just for confirmation since I don't want to run numbers for nothing, the Quartet can only appear every 20th encounter at the earliest, even if it is skipped, right?

User avatar
QuantumNightmare
AFH
Posts: 1827
Joined: Mon Jan 29, 2007 10:28 pm
Location: Montreal

Post by QuantumNightmare » Mon Apr 14, 2008 10:20 am

Correct. The original spading was done by choosing the same quartet adventure over and over, effectively skipping the noncombat.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Mon Apr 14, 2008 12:10 pm

First of all, I just noticed that the forum update ate my Bathroom data. I'll try to see whether I saved it somewhere. Otherwise I'll have to parse it out of my mafia logs again.

Alright, ran the numbers. I don't know whether this is exactly what you had in mind, so I'm including the complete results, too.

Quartet appearances in relation to total turns spent in the Ballroom (I always did 10000000):

Code: Select all

-30%: 465015 (4.65%)
-29%: 463872 (4.64%)
-28%: 462778 (4.63%)
-27%: 461445 (4.61%)
-26%: 460153 (4.6%)
-25%: 458813 (4.59%)
-24%: 457321 (4.57%)
-23%: 455916 (4.56%)
-22%: 454374 (4.54%)
-21%: 452659 (4.53%)
-20%: 450838 (4.51%)
-19%: 448923 (4.49%)
-18%: 447020 (4.47%)
-17%: 444657 (4.45%)
-16%: 442549 (4.43%)
-15%: 439929 (4.4%)
-14%: 437395 (4.37%)
-13%: 434544 (4.35%)
-12%: 431568 (4.32%)
-11%: 428227 (4.28%)
-10%: 424763 (4.25%)
-9%: 420942 (4.21%)
-8%: 416877 (4.17%)
-7%: 412593 (4.13%)
-6%: 407812 (4.08%)
-5%: 402901 (4.03%)
-4%: 397314 (3.97%)
-3%: 391043 (3.91%)
-2%: 385113 (3.85%)
-1%: 378430 (3.78%)
0%: 370873 (3.71%)
1%: 363659 (3.64%)
2%: 355277 (3.55%)
3%: 346698 (3.47%)
4%: 337456 (3.37%)
5%: 327389 (3.27%)
6%: 316860 (3.17%)
7%: 305718 (3.06%)
8%: 293478 (2.93%)
9%: 280181 (2.8%)
10%: 266026 (2.66%)
11%: 250772 (2.51%)
12%: 233938 (2.34%)
13%: 216084 (2.16%)
14%: 195455 (1.95%)
15%: 173559 (1.74%)
16%: 147394 (1.47%)
17%: 118890 (1.19%)
18%: 85090 (0.85%)
19%: 45890 (0.46%)
20%: 0 (0.0%)


Complete results. Note that Quartet appearances are not added to total encounters seen because the Quartet is skipped:

Code: Select all

-30%:
Floating Platter of Hors d'Oeuvres: 1743860 (17.44%)
Tapdancing Skeleton: 1743213 (17.43%)
Zombie Waltzers: 1745860 (17.46%)
Curtains: 4767067 (47.67%)
Quartet: 465015 (4.65%)
Total: 10000000

-29%:
Floating Platter of Hors d'Oeuvres: 1778456 (17.78%)
Tapdancing Skeleton: 1779569 (17.8%)
Zombie Waltzers: 1776778 (17.77%)
Curtains: 4665197 (46.65%)
Quartet: 463872 (4.64%)
Total: 10000000

-28%:
Floating Platter of Hors d'Oeuvres: 1815436 (18.15%)
Tapdancing Skeleton: 1812682 (18.13%)
Zombie Waltzers: 1813069 (18.13%)
Curtains: 4558813 (45.59%)
Quartet: 462778 (4.63%)
Total: 10000000

-27%:
Floating Platter of Hors d'Oeuvres: 1848097 (18.48%)
Tapdancing Skeleton: 1847996 (18.48%)
Zombie Waltzers: 1848392 (18.48%)
Curtains: 4455515 (44.56%)
Quartet: 461445 (4.61%)
Total: 10000000

-26%:
Floating Platter of Hors d'Oeuvres: 1883854 (18.84%)
Tapdancing Skeleton: 1883185 (18.83%)
Zombie Waltzers: 1884059 (18.84%)
Curtains: 4348902 (43.49%)
Quartet: 460153 (4.6%)
Total: 10000000

-25%:
Floating Platter of Hors d'Oeuvres: 1918064 (19.18%)
Tapdancing Skeleton: 1916486 (19.16%)
Zombie Waltzers: 1917059 (19.17%)
Curtains: 4248391 (42.48%)
Quartet: 458813 (4.59%)
Total: 10000000

-24%:
Floating Platter of Hors d'Oeuvres: 1953718 (19.54%)
Tapdancing Skeleton: 1949204 (19.49%)
Zombie Waltzers: 1952939 (19.53%)
Curtains: 4144139 (41.44%)
Quartet: 457321 (4.57%)
Total: 10000000

-23%:
Floating Platter of Hors d'Oeuvres: 1987882 (19.88%)
Tapdancing Skeleton: 1986412 (19.86%)
Zombie Waltzers: 1986818 (19.87%)
Curtains: 4038888 (40.39%)
Quartet: 455916 (4.56%)
Total: 10000000

-22%:
Floating Platter of Hors d'Oeuvres: 2020694 (20.21%)
Tapdancing Skeleton: 2021923 (20.22%)
Zombie Waltzers: 2021646 (20.22%)
Curtains: 3935737 (39.36%)
Quartet: 454374 (4.54%)
Total: 10000000

-21%:
Floating Platter of Hors d'Oeuvres: 2055959 (20.56%)
Tapdancing Skeleton: 2056940 (20.57%)
Zombie Waltzers: 2052428 (20.52%)
Curtains: 3834673 (38.35%)
Quartet: 452659 (4.53%)
Total: 10000000

-20%:
Floating Platter of Hors d'Oeuvres: 2091188 (20.91%)
Tapdancing Skeleton: 2088249 (20.88%)
Zombie Waltzers: 2090271 (20.9%)
Curtains: 3730292 (37.3%)
Quartet: 450838 (4.51%)
Total: 10000000

-19%:
Floating Platter of Hors d'Oeuvres: 2124151 (21.24%)
Tapdancing Skeleton: 2124248 (21.24%)
Zombie Waltzers: 2125807 (21.26%)
Curtains: 3625794 (36.26%)
Quartet: 448923 (4.49%)
Total: 10000000

-18%:
Floating Platter of Hors d'Oeuvres: 2159560 (21.6%)
Tapdancing Skeleton: 2158422 (21.58%)
Zombie Waltzers: 2158809 (21.59%)
Curtains: 3523209 (35.23%)
Quartet: 447020 (4.47%)
Total: 10000000

-17%:
Floating Platter of Hors d'Oeuvres: 2194711 (21.95%)
Tapdancing Skeleton: 2194898 (21.95%)
Zombie Waltzers: 2193486 (21.93%)
Curtains: 3416905 (34.17%)
Quartet: 444657 (4.45%)
Total: 10000000

-16%:
Floating Platter of Hors d'Oeuvres: 2228636 (22.29%)
Tapdancing Skeleton: 2226642 (22.27%)
Zombie Waltzers: 2227979 (22.28%)
Curtains: 3316743 (33.17%)
Quartet: 442549 (4.43%)
Total: 10000000

-15%:
Floating Platter of Hors d'Oeuvres: 2262027 (22.62%)
Tapdancing Skeleton: 2263112 (22.63%)
Zombie Waltzers: 2260699 (22.61%)
Curtains: 3214162 (32.14%)
Quartet: 439929 (4.4%)
Total: 10000000

-14%:
Floating Platter of Hors d'Oeuvres: 2297099 (22.97%)
Tapdancing Skeleton: 2295291 (22.95%)
Zombie Waltzers: 2296829 (22.97%)
Curtains: 3110781 (31.11%)
Quartet: 437395 (4.37%)
Total: 10000000

-13%:
Floating Platter of Hors d'Oeuvres: 2331021 (23.31%)
Tapdancing Skeleton: 2329819 (23.3%)
Zombie Waltzers: 2329739 (23.3%)
Curtains: 3009421 (30.09%)
Quartet: 434544 (4.35%)
Total: 10000000

-12%:
Floating Platter of Hors d'Oeuvres: 2364518 (23.65%)
Tapdancing Skeleton: 2364175 (23.64%)
Zombie Waltzers: 2363049 (23.63%)
Curtains: 2908258 (29.08%)
Quartet: 431568 (4.32%)
Total: 10000000

-11%:
Floating Platter of Hors d'Oeuvres: 2399625 (24.0%)
Tapdancing Skeleton: 2397604 (23.98%)
Zombie Waltzers: 2398126 (23.98%)
Curtains: 2804645 (28.05%)
Quartet: 428227 (4.28%)
Total: 10000000

-10%:
Floating Platter of Hors d'Oeuvres: 2432243 (24.32%)
Tapdancing Skeleton: 2431256 (24.31%)
Zombie Waltzers: 2432086 (24.32%)
Curtains: 2704415 (27.04%)
Quartet: 424763 (4.25%)
Total: 10000000

-9%:
Floating Platter of Hors d'Oeuvres: 2464593 (24.65%)
Tapdancing Skeleton: 2467941 (24.68%)
Zombie Waltzers: 2465468 (24.65%)
Curtains: 2601998 (26.02%)
Quartet: 420942 (4.21%)
Total: 10000000

-8%:
Floating Platter of Hors d'Oeuvres: 2500627 (25.01%)
Tapdancing Skeleton: 2499678 (25.0%)
Zombie Waltzers: 2500825 (25.01%)
Curtains: 2498870 (24.99%)
Quartet: 416877 (4.17%)
Total: 10000000

-7%:
Floating Platter of Hors d'Oeuvres: 2532805 (25.33%)
Tapdancing Skeleton: 2533648 (25.34%)
Zombie Waltzers: 2532076 (25.32%)
Curtains: 2401471 (24.01%)
Quartet: 412593 (4.13%)
Total: 10000000

-6%:
Floating Platter of Hors d'Oeuvres: 2566628 (25.67%)
Tapdancing Skeleton: 2568045 (25.68%)
Zombie Waltzers: 2566839 (25.67%)
Curtains: 2298488 (22.98%)
Quartet: 407812 (4.08%)
Total: 10000000

-5%:
Floating Platter of Hors d'Oeuvres: 2597724 (25.98%)
Tapdancing Skeleton: 2599494 (25.99%)
Zombie Waltzers: 2602584 (26.03%)
Curtains: 2200198 (22.0%)
Quartet: 402901 (4.03%)
Total: 10000000

-4%:
Floating Platter of Hors d'Oeuvres: 2633407 (26.33%)
Tapdancing Skeleton: 2633500 (26.34%)
Zombie Waltzers: 2633706 (26.34%)
Curtains: 2099387 (20.99%)
Quartet: 397314 (3.97%)
Total: 10000000

-3%:
Floating Platter of Hors d'Oeuvres: 2667747 (26.68%)
Tapdancing Skeleton: 2667057 (26.67%)
Zombie Waltzers: 2668208 (26.68%)
Curtains: 1996988 (19.97%)
Quartet: 391043 (3.91%)
Total: 10000000

-2%:
Floating Platter of Hors d'Oeuvres: 2700973 (27.01%)
Tapdancing Skeleton: 2699651 (27.0%)
Zombie Waltzers: 2700413 (27.0%)
Curtains: 1898963 (18.99%)
Quartet: 385113 (3.85%)
Total: 10000000

-1%:
Floating Platter of Hors d'Oeuvres: 2731686 (27.32%)
Tapdancing Skeleton: 2734704 (27.35%)
Zombie Waltzers: 2733442 (27.33%)
Curtains: 1800168 (18.0%)
Quartet: 378430 (3.78%)
Total: 10000000

0%:
Floating Platter of Hors d'Oeuvres: 2767222 (27.67%)
Tapdancing Skeleton: 2766544 (27.67%)
Zombie Waltzers: 2764873 (27.65%)
Curtains: 1701361 (17.01%)
Quartet: 370873 (3.71%)
Total: 10000000

1%:
Floating Platter of Hors d'Oeuvres: 2796494 (27.96%)
Tapdancing Skeleton: 2799988 (28.0%)
Zombie Waltzers: 2798099 (27.98%)
Curtains: 1605419 (16.05%)
Quartet: 363659 (3.64%)
Total: 10000000

2%:
Floating Platter of Hors d'Oeuvres: 2829711 (28.3%)
Tapdancing Skeleton: 2832550 (28.33%)
Zombie Waltzers: 2830842 (28.31%)
Curtains: 1506897 (15.07%)
Quartet: 355277 (3.55%)
Total: 10000000

3%:
Floating Platter of Hors d'Oeuvres: 2862222 (28.62%)
Tapdancing Skeleton: 2865185 (28.65%)
Zombie Waltzers: 2861451 (28.61%)
Curtains: 1411142 (14.11%)
Quartet: 346698 (3.47%)
Total: 10000000

4%:
Floating Platter of Hors d'Oeuvres: 2894952 (28.95%)
Tapdancing Skeleton: 2893171 (28.93%)
Zombie Waltzers: 2894857 (28.95%)
Curtains: 1317020 (13.17%)
Quartet: 337456 (3.37%)
Total: 10000000

5%:
Floating Platter of Hors d'Oeuvres: 2928036 (29.28%)
Tapdancing Skeleton: 2924941 (29.25%)
Zombie Waltzers: 2925321 (29.25%)
Curtains: 1221702 (12.22%)
Quartet: 327389 (3.27%)
Total: 10000000

6%:
Floating Platter of Hors d'Oeuvres: 2957132 (29.57%)
Tapdancing Skeleton: 2955054 (29.55%)
Zombie Waltzers: 2959982 (29.6%)
Curtains: 1127832 (11.28%)
Quartet: 316860 (3.17%)
Total: 10000000

7%:
Floating Platter of Hors d'Oeuvres: 2988906 (29.89%)
Tapdancing Skeleton: 2990826 (29.91%)
Zombie Waltzers: 2987212 (29.87%)
Curtains: 1033056 (10.33%)
Quartet: 305718 (3.06%)
Total: 10000000

8%:
Floating Platter of Hors d'Oeuvres: 3021221 (30.21%)
Tapdancing Skeleton: 3018810 (30.19%)
Zombie Waltzers: 3018602 (30.19%)
Curtains: 941367 (9.41%)
Quartet: 293478 (2.93%)
Total: 10000000

9%:
Floating Platter of Hors d'Oeuvres: 3049709 (30.5%)
Tapdancing Skeleton: 3050999 (30.51%)
Zombie Waltzers: 3048628 (30.49%)
Curtains: 850664 (8.51%)
Quartet: 280181 (2.8%)
Total: 10000000

10%:
Floating Platter of Hors d'Oeuvres: 3077253 (30.77%)
Tapdancing Skeleton: 3082622 (30.83%)
Zombie Waltzers: 3081085 (30.81%)
Curtains: 759040 (7.59%)
Quartet: 266026 (2.66%)
Total: 10000000

11%:
Floating Platter of Hors d'Oeuvres: 3109326 (31.09%)
Tapdancing Skeleton: 3108987 (31.09%)
Zombie Waltzers: 3110627 (31.11%)
Curtains: 671060 (6.71%)
Quartet: 250772 (2.51%)
Total: 10000000

12%:
Floating Platter of Hors d'Oeuvres: 3137775 (31.38%)
Tapdancing Skeleton: 3137784 (31.38%)
Zombie Waltzers: 3139682 (31.4%)
Curtains: 584759 (5.85%)
Quartet: 233938 (2.34%)
Total: 10000000

13%:
Floating Platter of Hors d'Oeuvres: 3168777 (31.69%)
Tapdancing Skeleton: 3165710 (31.66%)
Zombie Waltzers: 3165738 (31.66%)
Curtains: 499775 (5.0%)
Quartet: 216084 (2.16%)
Total: 10000000

14%:
Floating Platter of Hors d'Oeuvres: 3194421 (31.94%)
Tapdancing Skeleton: 3193353 (31.93%)
Zombie Waltzers: 3195947 (31.96%)
Curtains: 416279 (4.16%)
Quartet: 195455 (1.95%)
Total: 10000000

15%:
Floating Platter of Hors d'Oeuvres: 3220499 (32.2%)
Tapdancing Skeleton: 3218990 (32.19%)
Zombie Waltzers: 3224073 (32.24%)
Curtains: 336438 (3.36%)
Quartet: 173559 (1.74%)
Total: 10000000

16%:
Floating Platter of Hors d'Oeuvres: 3247616 (32.48%)
Tapdancing Skeleton: 3246948 (32.47%)
Zombie Waltzers: 3247122 (32.47%)
Curtains: 258314 (2.58%)
Quartet: 147394 (1.47%)
Total: 10000000

17%:
Floating Platter of Hors d'Oeuvres: 3271640 (32.72%)
Tapdancing Skeleton: 3271425 (32.71%)
Zombie Waltzers: 3271757 (32.72%)
Curtains: 185178 (1.85%)
Quartet: 118890 (1.19%)
Total: 10000000

18%:
Floating Platter of Hors d'Oeuvres: 3292501 (32.93%)
Tapdancing Skeleton: 3293206 (32.93%)
Zombie Waltzers: 3297480 (32.97%)
Curtains: 116813 (1.17%)
Quartet: 85090 (0.85%)
Total: 10000000

19%:
Floating Platter of Hors d'Oeuvres: 3315049 (33.15%)
Tapdancing Skeleton: 3316026 (33.16%)
Zombie Waltzers: 3314700 (33.15%)
Curtains: 54225 (0.54%)
Quartet: 45890 (0.46%)
Total: 10000000

20%:
Floating Platter of Hors d'Oeuvres: 3333560 (33.34%)
Tapdancing Skeleton: 3333762 (33.34%)
Zombie Waltzers: 3332678 (33.33%)
Curtains: 0 (0.0%)
Quartet: 0 (0.0%)
Total: 10000000

User avatar
QuantumNightmare
AFH
Posts: 1827
Joined: Mon Jan 29, 2007 10:28 pm
Location: Montreal

Post by QuantumNightmare » Mon Apr 14, 2008 12:56 pm

That's exactly what I was looking for, thanks! I put together a 6th order polynomial that generates the quartet odds based on your applied combat rate modifiers, which will be included in the next update of the leveling spreadsheet.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Thu May 15, 2008 2:14 pm

I just added a couple days worth of turns in the Bathroom.

Is it just me or does the Bathroom data look strange? It seems almost as if the encounter probabilities get changed after spending 600-700 turns in the Bathroom. The data was gathered during two ascensions; the first one was from samples 1 to 5 and the rest is from the second ascension.

Edit: I just changed the format a little to make things more obvious. The second ascension had some turns in bathroom before I started collecting data again.



Also, I guess I have to ask for a third time since I haven't gotten an answer yet: Is it possible to add the possibility to create tables inside posts? Not only would it help in this thread, but also in other threads of the commune. Or the contest forum, as I recently noted. I don't demand that this feature has to be added, I'd just like to know whether it will/can be done. (By which I mean: Even a "No." is a better answer than no answer at all.)
Last edited by Flolle on Fri May 16, 2008 1:06 pm, edited 2 times in total.

User avatar
lostcalpolydude
Forum Maestro
Posts: 2944
Joined: Wed Apr 23, 2008 6:54 pm

Post by lostcalpolydude » Thu May 15, 2008 2:34 pm

Flolle wrote:Also, I guess I have to ask for a third time since I haven't gotten an answer yet: Is it possible to add the possibility to create tables inside posts? Not only would it help in this thread, but also in other threads of the commune. Or the contest forum, as I recently noted. I don't demand that this feature has to be added, I'd just like to know whether it will/can be done. (By which I mean: Even a "No." is a better answer than no answer at all.)

From some searching I found http://www.phpbb.com/community/viewtopic.php?t=219269 which seems to be what you are asking about. I don't know if it's compatible with this version of phpBB (or even what version of phpBB this is).
bmaher: Softcore character
lostcalpolydude: Hardcore character (abandoned)

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Thu May 15, 2008 2:45 pm

Yeah, that would be more than enough for me.

Hellion
Pie of the Sky
Posts: 47
Joined: Fri May 16, 2008 3:15 pm
Location: Flatland

Post by Hellion » Fri May 16, 2008 3:34 pm

Flolle wrote:
Trivial Updates wrote:The Hole in the Sky no longer has two different (identical) versions of the Astronomer in it.

Radioshow Transcript wrote:There were basically 2 copies of the astromomer? were started with 6 constellations, then added 3 more, but we didn?t want to reduce the chance of encountering an astronomer. So we put in a 2nd astronomer with only a 50% chance of finding that one. This messed with the items that affected fighting a given monster ? Harold?s bell, Transcendant Olfaction, etc. I?m not sure if this will make it more likely or less likely that you fight an astronomer? now it is 2 copies of the same monster, with 25% chance of rejecting it. We may need to adjust the odds later.

Kinda contradicting. [...]


How so? Previously there were two astronomers, let's call them astronomer 1 and astronomer 2. There was no in-game way to tell these two apart; one was an exact duplicate of the other. However, astronomer 2 would be rejected and rerolled 50% of the time that it was selected.

Now there is only astronomer 1, which appears twice in the list of selectable monsters, and will be rejected and rerolled 25% of the time.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Fri May 16, 2008 4:42 pm

Heh, this thread is probably a bit confusing with me changing my model on how stuff works all the time in my first few posts ... :D

Trivial Updates wrote:[...] two different (identical) versions of the Astronomer [...]

Radioshow Transcript wrote:[...] now it is 2 copies of the same monster [...]

That's what was kinda confusing. Back then, I had a wrong idea about how encounter ratios are accomplished. After looking through some of the theories on the "On the Track" effect, I realised that it would be much easier to do it with a roulette wheel mechanic. As in every encounter gets one (or more) slot(s) and then a slot number is rolled. The encounter behind the slot will be the rolled encounter. Of course, this is ignoring the queue and other rejection rates for the sake of simplicity. I'm sure everyone else had this one figured out from the start, but I hadn't. >.>

Anyway, yes later on I understood how the two comments were meant to be understood. (if that makes any sense ... >.>)


And on another note, is the model I proposed for the Hole in the Sky correct? As in, the Astronomer has two "slots" but due to the queue the appearance rate is a little dampened. It seemed like you were the person that did the math on that one back when there was this talk about the Astronomer bounty and they said on the radio that the bounty was okay, even though my model would say that the bounty takes slightly too long.

Since you're a /dev and all, you probably can't answer this one anyway, but I mentioned this mostly just because I again wanted to say that I'm sorry if I came off as an ass in that radio show transcript thread. I really didn't mean to step on anybody's toes with my post. :oops:

Hellion
Pie of the Sky
Posts: 47
Joined: Fri May 16, 2008 3:15 pm
Location: Flatland

Post by Hellion » Fri May 16, 2008 7:11 pm

Flolle wrote:...is the model I proposed for the Hole in the Sky correct? As in, the Astronomer has two "slots" but due to the queue the appearance rate is a little dampened.


Well, I don't see your model in a quick re-read, but this is all stuff I figured out well before I became a dev, so I don't feel too constrained in talking about it. And you probably know most or all of this already anyway. So I'm just gonna ramble on for a bit.

There is a list of available encounters associated with a given area. (Actually, there's a list for combats (encounters) and non-combats (adventures), but they both work the same way (except you can't influence the list of adventures at all), so I'm going to talk about them interchangeably.)

When you're in area 1, the available encounter list contains encounters 1,2,3, and 4. In this case obviously you should have a 1/4 chance of receiving any specific encounter.

When you're in area 2, the available encounter list contains encounters 5, 5, and 6. Clearly your base chance of encounter 5 is 2/3 and encounter 6 is 1/3. With the 5-position-long queue, these chances are modified to more like 64/36 or so due to the more frequent occurance of the queue containing 55555 (a 13% chance to raise the odds of encounter 6 from 1/3 to approximately (1/3 + (2/3*3/4*1/3) = 1/2 (that's 2/3, the odds of choosing encounter 5, times 3/4, the chance of rerolling it since it's already in the queue, times 1/3, the chance of choosing adventure 6 on the reroll; I ignore the recursive nature of the selection process here) instead of 66666 (a 1/2% chance to raise the odds of encounter 5 from 2/3 to approximately 2/3 + (2/3*3/4*1/3) = 5/6) to tilt the balance slightly away from the unmodified percentages.

The Hole In The Sky went from A,1,2,3,4,5,6 (astronomer plus 6 others) to A1, A2,1,2,3,4,5,6,7,8,9; where A2 has a 50% chance of being discarded and rerolled on top of the regular queue mechanic. Then it switched to A1,A1,1,2,3,4,5,6,7,8,9; where A1 has a 25% chance of being rerolled on top of the regular queue mechanic.

(There are other conditionals than just a flat "X%" that can be put on an adventure, of course; I'll leave it to you to theorize about how/when that happens. :-) )

It seemed like you were the person that did the math on that one back when there was this talk about the Astronomer bounty and they said on the radio that the bounty was okay, even though my model would say that the bounty takes slightly too long.

Yep, that was me. My simulator code is undoubtedly a bit different than yours....

but I mentioned this mostly just because I again wanted to say that I'm sorry if I came off as an ass in that radio show transcript thread. I really didn't mean to step on anybody's toes with my post.


No apology necessary!

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Mon May 19, 2008 2:46 pm

Hellion wrote:Yep, that was me. My simulator code is undoubtedly a bit different than yours....

Well, your explanation looks exactly like I implemented it. Makes me wonder where I'm going wrong ... :?

/cue another half a year of sleepless nights thinking about KoL because the encounter rolling still isn't perfect. T.T

Hellion
Pie of the Sky
Posts: 47
Joined: Fri May 16, 2008 3:15 pm
Location: Flatland

Post by Hellion » Tue May 20, 2008 10:39 pm

Well, you'll be happy to know, since your simulator produced different results than mine, I wrote another one (since the first one was on another computer) and re-ran the tests.

This resulted, eventually, in finding the bug that was in the first one that led me to mistakenly believe that the bounty would take an average of 40 turns.

I have reported my mistake into the /dev channel and my corrected results, so perhaps we shall see a change ... eventually.

(If you're curious: my original code treated the two occurances of the astronomer as two unique monsters instead of 2 copies of the same monster, so the queue effect was not nearly as pronounced.)

User avatar
starwed
AFH
Posts: 263
Joined: Sun Mar 16, 2008 12:31 am

Post by starwed » Wed May 21, 2008 11:05 am

I have reported my mistake into the /dev channel and my corrected results, so perhaps we shall see a change ... eventually.

I can't help but notice that I encountered rather more astronomers than normal today... (5 in 20-30 adventures, including 2 right off the bat.)

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Wed May 21, 2008 11:42 am

Well, getting 5 Astronomers (with an empty queue at the start) takes 37.5 turns on average with a standard deviation of 10.4, so I guess that result wouldn't be totally out of the realm of possibility ...

(/me is doubting that they'd implement this change so fast when they are on a road trip ...)

Edit: Also, wai, the sleepness nights are scratched off of the road map! :D

Hellion
Pie of the Sky
Posts: 47
Joined: Fri May 16, 2008 3:15 pm
Location: Flatland

Post by Hellion » Wed May 21, 2008 1:07 pm

Jick said he'd probably remove the 25% conditional on the astronomers, but not until he gets back from his trip. So nothing has changed, yet.

salien
Pie in the Sky
Posts: 115
Joined: Fri Jan 11, 2008 1:50 pm

Post by salien » Wed May 21, 2008 2:06 pm

Couldn't that mess up the balance of stars/lines vs. charts (the ascension astronomer balance), though? I don't know if anyone's actually looked at that, and it certainly varies with a person's item bonus; but if the goal is just to fix the bounty, lowering the required items from 6 to 5 should do that without touching the ascension balance.

Actually, I went and did some rough calculations of the current ascension setup; on average right now, you should have encountered 3 astronomers after 25 adventures. In order to have all the stars/lines you need by that time, you want to run at least +78% item drops. If the 25% rejection rate on astronomers is removed, I'm not entirely certain how many adventures it will take to encounter 3 of them, but my initial estimate pushes it down to 20 adventures; at that point the minimum desired item drop bonus goes up to +131%. Not sure if that's desired or intended, but it seems that if the issue is just with the bounty and not how long the HitS takes, adjusting the bounty makes more sense than adjusting the astronomer.

feng shweez
Inscrutable Pi
Posts: 361
Joined: Wed Jul 11, 2007 10:41 am
Location: Boston, MA

Post by feng shweez » Wed May 21, 2008 2:11 pm

/me pokes Hellion to get the hobo bounty changed.

Unless you want me to make a...shudder forum post about it.
[url=http://img341.imageshack.us/img341/9392/nootrosforumslk9.gif][img]http://img341.imageshack.us/img341/9392/nootrosforumslk9.gif[/img][/url]

User avatar
lostcalpolydude
Forum Maestro
Posts: 2944
Joined: Wed Apr 23, 2008 6:54 pm

Post by lostcalpolydude » Wed May 21, 2008 2:20 pm

feng shweez wrote:/me pokes Hellion to get the hobo bounty changed.

Unless you want me to make a...shudder forum post about it.

I think the last time it came up, Jick finally just said that the Outskirts bounty is too fast, and the hobo bounty is the way it should be (since those are the two that were being compared). He didn't seem to care about changing anything, but if he did it might instead be making another bounty longer. I have no idea where this was posted, but that's what I remember.
bmaher: Softcore character
lostcalpolydude: Hardcore character (abandoned)

Hellion
Pie of the Sky
Posts: 47
Joined: Fri May 16, 2008 3:15 pm
Location: Flatland

Post by Hellion » Wed May 21, 2008 3:28 pm

The initial balance of astronomers was intended to be 1/7. Nuking the 25% flat reject rate returns the astronomers to their intended frequency, so I expect that's what Jick was thinking when he indicated his preference for that option.

And Feng, my recollection is the same as lostcalpolydude's; the expected "fix" for the hobo bounty will probably be to make the Knob bounty longer. So I'm not gonna poke at it much.

User avatar
Flolle
Fie the Pie
Posts: 848
Joined: Mon Apr 02, 2007 10:13 am

Post by Flolle » Sun Jul 06, 2008 4:59 pm

I finalised the Bathroom data. The numbers still look strange to me, but I'm not willing to throw more turns at this than I already did. If someone else wants to further check it out, be my guest.

I also added Airship data.

Post Reply