Followers 3

# How to tell if a cache is missing: some dnf math

## Recommended Posts

Let's say you are a cache owner and have hidden a reasonably easy to find cache (d2 or less, single stage, visible if you look for it from the right angle) and someone logs a single dnf:  What are the chances the cache is actually missing?  Based on my own experience, I'd say the odds are roughly 40:60 (It's probably there).  This is based on my own actual find rate of 94% and the subsequent results of the 6% I couldn't find (for every 10 dnfs I have, 6 of them are found subsequently and 4 of them are either never found again, archived or repaired/replaced).  So for me, I generally assume if I couldn't find the cache; it was probably there.  My average find is a d2 cache, so I think it applies to this group.

Ok so a few weeks go by and there is a second dnf.  Now the odds are tilting towards the cache may be missing:  You multiply the probabilities 60% times 60% which gives you odds of 36%, by the third dnf (.6 times .6 times .6) the odds fall to 22% (one chance in five) and by the the fourth dnfs shows up it's down to less than 15% .  Now you might consider some mitigating factors (team dnfs, newbies) but it's pretty clear that by the time you get to five consecutive dnfs on a d2 or less cache (10% chance it's there) it's time for you to go check on it.  Add to that that some (perhaps many) dnfs are never logged and you should plan on doing a maintenance run...

Evil hides aside (they are designed not to be found) one can have a pretty clear idea of the odds of the cache being find-able based on these numbers.  I think this supports the use of an algorithm such as the "Cache health score" to nudge CO's to go check on the cache.  It also supports "armchair" posting of NM logs in my view for CO's who ignore the obvious as the dnfs pile up.  It's the number of consecutive dnfs that matter, not who dnfed.  No one can find what isn't there.  Arguing that "you must go look for my cache before I'll check on it" it really kind of odd, I think.

To put it another way:  For a cache with nothing wrong with it,  s single dnf will happen roughly one log out of 20, (6% dnf rate)  two consecutive dnfs roughly once every 40 logs) but three consecutive dnfs will only happen about once every 80 logs.  Chances are really pretty good that a cache with a 95% find ratio that wracks up three or more consecutive dnfs is gone...

If you are a CO I think this is a reasonable system to use.  The easier the cache and the higher it's find ratio, the more likely this to be accurate.  Very few things in life have better than 98% odds of being true...

edexter

Edited by edexter
context

19 minutes ago, edexter said:

Let's say you are a cache owner and have hidden a reasonably easy to find cache (d2 or less, single stage, visible if you look for it from the right angle)

Of course, in real life, any system used will need to accommodate other caches as well: greater than D2, or multi-stage, or cleverly hidden, or whatever.

I'm not sure I would consider two DNF logs on the same cache to be completely independent. A previous DNF log may be discouraging to some cachers, resulting in them putting less effort into finding the cache, or giving up that much sooner than when the previous log was a find. I know there have been times this happened to me.

I also look up the profile of every cacher that makes a DNF on any of my hides. If they are new cachers (few finds, recent membership), I'm more likely to brush it off than a veteran cacher.

Conditional probabilities - make the model a little more complex, but perhaps a little more accurate too.

My rule is to make a run after 3 DNFs unless the cache is made to be difficult.

Edited by Mineral2

19 minutes ago, Mineral2 said:

I also look up the profile of every cacher that makes a DNF on any of my hides. If they are new cachers (few finds, recent membership), I'm more likely to brush it off than a veteran cacher.

Depends. If it's 2 DNFs in a row, on 2 different days, or the descriptions suggest that they looked in the right place, there might be a problem.  I may contact the last person who DNF'd for more details. I'll go check my cache if I think there's a chance that the cache is missing.  My caches are nearby (within a 40-minute drive). I have one D4 cache that regularly gets a DNF but rarely 2 in a row. 3 in a row almost always means there's a problem (it has gone missing 4 times in 14 years).

If only it were so simple in real life. There are many things that can lead to a DNF - some that I've had on my caches include presence of muggles, swarms of mosquitoes, approaching storm, failing light, climb too tough, etc. None of the DNFs with reasons like that are likely to imply the cache is missing.

So here are some stats from my own hides. I've only included traditionals and my EC since other factors usually dominate the D rating of puzzles and multis.

D1: #1 - 0 DNFs from 15 finds

D1.5: #1 - 0 DNFs from 240 finds, #2 - 2 DNFs from 38 finds, #3 - 0 DNFs from 27 finds, #4 - 0 DNFs from 5 finds

D 2.0: #1 - 10 DNFs from 46 finds (twice 2-in-a-row), #2 3 DNFs from 39 finds (once 2-in-a-row), #3 - 1 DNF from 44 finds, #4 6 DNFs from 15 finds (twice 2-in-a-row)

D 2.5: #1 - 3 DNFs from 44 finds (once 2-in-a-row)

Only two of those DNFs have been due to a missing or inaccessible cache. Can you tell which ones?

And by the way, the D1.5 with the 2 DNFs was my EC and no, it wasn't missing or in need of maintenance.

As others have mentioned, attempts aren't independent. DNFs beget DNFs as some searchers assume before they even start that it's likely missing if the last log was a DNF. You can see that from my stats, where caches with only 3 DNFs in total have had two in a row. Likewise when a tough cache that's had a long string of DNFs is finally found, word gets out and suddenly there's a flurry of finds (GCT2V1 is a good example of this).

7 hours ago, edexter said:

I think this supports the use of an algorithm such as the "Cache health score" to nudge CO's to go check on the cache.

You are ignoring DNFs with the text like "Found it!" or "Oops, wrong side of fence", etc. That's the main problem with using a DNF count in an algorithm. (It doesn't help that the CHS algorithm is more sensitive to DNFs than the counts you're describing.)

7 hours ago, edexter said:

It also supports "armchair" posting of NM logs in my view for CO's who ignore the obvious as the dnfs pile up.  It's the number of consecutive dnfs that matter, not who dnfed.  No one can find what isn't there.  Arguing that "you must go look for my cache before I'll check on it" it really kind of odd, I think.

Yes, I agree completely that reading DNFs can lead to a perfectly reasonable conclusion that the cache likely isn't there, in which case actually going to GZ is both pointless, because it won't produce any additional information, and a waste of time, because there's no cache to find. If no one posts an NM under those conditions, no NM will never be posted. Counting DNFs is not always reliable, so I hope no one posts NMs based on DNF count alone.

28 minutes ago, dprovan said:

I hope no one posts NMs based on DNF count alone.

I would post an NM based how obvious it is that the container is likely missing using things like the D-rating, hint, past logs, photo gallery, previous DNFs.

As a CO I wouldn't care if someone logged an NM for any reason. I'd rather it was a mistaken NM then no NMs posted for fear of irritating me, the CO. It's easy to read the NM log then post an OM immediately with an explanation, or go check the cache if the NM log makes it seems possible that my cache needs attention.

14 hours ago, edexter said:

If you are a CO I think this is a reasonable system to use.  The easier the cache and the higher it's find ratio, the more likely this to be accurate.  Very few things in life have better than 98% odds of being true...

That might be reasonable for easy-to-find easy-to-reach caches with low DNF rates and statistically-signifcant numbers of finds, but it's a different story for the higher D/T ones. There are:

• Few attempts on the cache - it might take years to get the find count into double figures, and trying to infer a probability of DNFs on a cache that's only had a handful of finds won't work.
• Many more possible ways for non-cache-related DNFs, particularly for multis where messing up a waypoint calculation can result in searching in the wrong place. Some will log a DNF if a high-T cache turned out to be too physically demanding for them - I've done that myself when I haven't been able to climb out to where the cache is.
• Much less chance of a remote cache just going missing by itself - a T4.5 is just as tough for muggles as it is for cachers - so the probability that a DNF is due to a missing cache is much lower than for an urban cache where muggling is a constant risk.

Counting DNFs or calculating DNF-to-find ratios on high D/T caches doesn't tell you much at all about whether the cache might be missing. I've had something like 30 DNFs across my higher D/T multis and puzzles and none have been due to a missing cache.

I pretty much agree with all of the above.  The "quality" of the dnfs matters (so you have to read them) and the stats really only apply to d2 or less caches.  Personally I find my dnf rate climbs right along with the d rating and, the higher the d rating, the more confident I am that I'm dealing with a dnf not a wnt (Was Not There).   I'm also more likely to hunt for a higher d cache that is also a higher t and these tend to have more real dnfs.  In other words a 1.5/1.5 with three consecutive dnfs and an overall find rate of 95% is likely missing while three consecutive dnfs on a d3.5/t3.5 with a 70% find rate suggests it is accurately rated (and likely still there).   I haven't run the numbers but I think in general, except for tree climbs, the higher the t rating the less likely a cache is to "go missing".  A tethered watertight cache on the ground not in a flood zone rated t3 or higher is is pretty unlikely to wander off.  While a roadside bison hanging in a cedar tree tends to have a short life span and a wet log (It's always the o-ring...)

dexter

## Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

×   Pasted as rich text.   Paste as plain text instead

Only 75 emoji are allowed.