Search odds condensed (results)

From The Shartak Wiki
(Redirected from Search odds discussion)
Jump to: navigation, search

Results are from Search odds condensed. Results are rounded to the nearest 1%.

Note Feel free to leave any analysis or thoughts, this initial data is from trial 1 and 2 of the Search odds condensed. Any data that was left with nothing found was left out this time, but will be used for the next set of results. Feel free to add to the data to be able to gleen more from the results.

Contents

Open issues

  1. Are the odds different for differing jungle types? -- fitzcarraldo|T 11:16, 3 March 2006 (GMT)
  2. Need to check mountain caves versus trails, otherwise they will all fall under mountains. -- fitzcarraldo|T 11:16, 3 March 2006 (GMT)
  3. At this time data from being within huts are considered d0 resuls. As well eventual jungle squares with trees have not been broken out. need to break out. -- fitzcarraldo|T 11:16, 3 March 2006 (GMT)
  4. Do results differ on d0 within village limits to d0 just outside?
  5. If there is a tree on a jungle square, can those squares be cleared of jungle? does it have any effect on results? -- fitzcarraldo|T 11:16, 3 March 2006 (GMT)

Analysis

More data is needed, if you would like to help, please enter data at Search odds condensed.

Further analysis is coming, but feel free to add your thoughts.

I'll put current find rates here shortly, using 95% confidence intervals. See Statistical analysis for a brief overview (or to argue about) data analysis. See How searching works for discussion about underlying game mechanics (meta-stuff that frames data collection and data analysis).

Current Resource Hut Data, June 1, 2006

OutsiderMedicalHut1497.PNG

OutsiderAmmoHut394.PNG

The overall find rate of about 36% appears to be about equally split between key locational items (FAK in the Medical Hut; Rifle, Bullet, BulletBox in the Ammo Hut) and everything else (items found in both Medical and Ammo Huts: water, beer, machete, blunt machete, knife, sharpening stone, gold coin, GPS unit). When eyeballing the two graphs shown above, we see the same general profile for the shared items. This suggests the following intriguing hypothesis:

The find rates for each non-key item (such as a GPS unit) is exactly the same in the Medical Hut as in the Ammo Hut. The chance of finding a First Aid Kit in the Medical Hut is exactly equal to the chance of finding one of (a Rifle or Bullet or Bullet Box) in the Ammo Hut.

This suggests a find rate of ~6% for bottles of water; ~3% for machetes and knives; ~2% for beer, sharpening stones, blunt machetes, and gold coins; and ~1% for gps units. In the Medical Hut, the find rate for first aid kits is ~17%, while in the Ammo Hut, there is a ~10% find rate for bullets, ~4% for boxes of bullets, and ~3% for rifles.

35% versus 1-in-3

We have completed so many searches in the Outsider Medical Hut that the 1-in-3 find rate is no longer supported by the data: 404 finds in 1120 yields p = 36.1% ±1.4%. Here are some of the current possibilities to investigate:

  • Find rate of 35% (7 in 20)
  • Find rate of 36% (9 in 25)
  • Find rate of 36.67% (11 in 30)
  • Find rate of 37.5%, or 5 in 8
  • Someone forgot to identify themselves as Scavenging when logging data
  • Bad luck
  • Change in medical hut search odds

In my opinion, the last possibility is likely -- hundreds of searches in the medical hut were logged without a GPS discovery, and there were no anecdotal cases of GPS discovery in the medical hut. Many players were complaining in-game about GPS scarcity. Shortly before the trading huts were introduced, GPS finds started appearing in the search data, and now almost every 60 AP search yields 1-2 GPS units.

Comparison Charts, April 2006

The comparison includes two types of data, both the probability of finding nothing (orange) and the number of searches performed (maroon)

Data comparison


Comparative Jungle Analysis

There seems to be so far no linear difference between the jungle types that would demonstrate the concept that it could be easier to find items in d0 than in d9. But, in fact, the data seems to illustrate that the best area to find something is in d4. Assuming it is not a case of RNG luck, then perhaps the concepts driving the data is this:

If one clears everything, then there is little left to find. Alternately if one clears very little, the obstructions make it difficult to find anything.

The data doesn't seem to be erroneous, as the slope of the line, from d4-d1, indicates the other data correlates to this effect. But, the number of searches is really not great enough to ascertain that the data is very accurate. Also, the data d5-9 does not seem to correlate to a bell curve effect. In the end, it will simply take more searching to verify.

NOTE, any special cases where the data set was in ruins or a village, or with a tree or hut in the area have not been included with the jungle data. The special cases and combinations will come later.

Comparative d0 Analysis: Jungle versus Village

I suspect that the Jungle search odds are different from the Outsider Village search odds. In the data on d0 squares through April 7, I see 58 Jungle finds in 471 searches (12%), whereas there are 72 Village finds in 296 searches (24%). Also the found items are very different:

  • Jungle = 12 berries, 10 mangos, 14 bananas / 7 machetes, 1 knife, 2 rifles, 3 FAKs
  • Outsider Village = 0 berries, 1 mango, 3 bananas / 20 machetes, 9 knives, 7 rifles, 9 FAKs

The outsider village searches look similar to a mix of Ammo Hut and Medical Hut. --Tycho44 05:49, 16 May 2006 (BST)

Beach and Grassland

Beach
Grassland


Beach
Grassland


Mountain, Swamp, River, Water, Deep Water

Mountain
Swamp
Water


Jungle

This shows all the jungle data combined d0-10

Jungle results combined



Jungle d0-10

Jungle d0
Jungle d1
Jungle d2
Jungle d3
Jungle d4


Jungle d5
Jungle d6
Jungle d7
Jungle d8
Jungle d9
Jungle d10




updated

fitzcarraldo|T 12:45, 10 April 2006 (BST)

Update: Added contrast of Jungle vs Village finds. --Tycho44 05:49, 16 May 2006 (BST)

Personal tools
Namespaces

Variants
Actions
Navigation
Toolbox