EQDB Update - 1 Year Anniversary
Skywise
Mother Earth. I think I've done quite well.
The following does not reflect one full year of analysis, but shy
by two weeks for reasons explained on the EQDB website.
Of the 11,606 candidate quakes (global mag 4+) analysed so far, I
successfully predicted 10,706 of them, for a success rate of 92.245%.
Of the 3,685 predictions analyzed so far, 2,453 of them had at
least one hit, for a success rate of 66.567%.
Here's the analysis output from my confirmation program,
List Generated: 2006/ 9/27 2: 4:43.640
NEIC data: 2006/ 9/26 2:59:44.000
Magnitude range: 4.00 - 10.00
Distance range: 0 - 1000 kilometers
Predictions: 3685 (total number of predictions made so far)
Prediction Hits: 2453 (number of predictions with at least one
hit)
Total candidates: 11606 (total number of quakes within the magnitude
range)
Total hits: 10706 (number of candidate quakes within range of
a prediction)
Candidates missed: 900 (number of candidate quakes missed)
Hit/Candidate Ratio: 92.245%
Hit/Prediction Ratio: 66.567% (predictions with hits vs total predictions)
Brian
--
http://www.skywise711.com - Lasers, Seismology, Astronomy, Skepticism
Seismic FAQ: http://www.skywise711.com/SeismicFAQ/SeismicFAQ.html
Quake "predictions": http://www.skywise711.com/quakes/EQDB/index.html
Sed quis custodiet ipsos Custodes?
bettera...@gmail.com
or forecasted, that are the best upon yourself consideration?
Are you considering a radius of 500 Km of so, for the near future?
Alexis.
Thomas A. Russ
> Well, it's been one full year now that I've been throwing darts at
> Mother Earth. I think I've done quite well.
>
> The following does not reflect one full year of analysis, but shy
> by two weeks for reasons explained on the EQDB website.
>
> Of the 11,606 candidate quakes (global mag 4+) analysed so far, I
> successfully predicted 10,706 of them, for a success rate of 92.245%.
>
> Of the 3,685 predictions analyzed so far, 2,453 of them had at
> least one hit, for a success rate of 66.567%.
So, it would also appear from these results that each prediction had a
mean of 4.36 earthquakes satisfying the parameters of a prediction.
Without even looking at the predictions, that makes it appear that the
prediction are extremely non-specific.
Now, the other question that is left unanswered, is what the expected
success rate would be through pure chance. In other words, what if you
took the same predictions, but randomized their starting times. How
many of them would have also been successful?
--
Thomas A. Russ, USC/Information Sciences Institute
Skywise
@i3g2000cwc.googlegroups.com:
> Brian, It is possible for you to say us, 3 of the cases of "prediction"
> or forecasted, that are the best upon yourself consideration?
It appears English is not your native language, so forgive me
for not understanding your question....
> Are you considering a radius of 500 Km of so, for the near future?
This I do understand.
I have thought of applying my method to smaller scale regions,
such as western North America, or even just Southern California.
Of course, I would necessarily use much smaller circles. In
fact, I'd like to test the method to see how small of a circle
I can use and yet still achieve apparently good success rates.
My original goal was simply to blow the pants off everybody
else with an insanely high hit count. I think I've achieved
that.
I've also thought of ditching circles and using other boundary
shapes.
But alas, time has suddenly become precious for me, so it
may have to wait while I concentrate on other work.
Skywise
news:ymislid...@sevak.isi.edu:
> Skywise <in...@oblivion.nothing.com> writes:
>
>> Well, it's been one full year now that I've been throwing darts at
>> Mother Earth. I think I've done quite well.
>>
>> The following does not reflect one full year of analysis, but shy
>> by two weeks for reasons explained on the EQDB website.
>>
>> Of the 11,606 candidate quakes (global mag 4+) analysed so far, I
>> successfully predicted 10,706 of them, for a success rate of 92.245%.
>>
>> Of the 3,685 predictions analyzed so far, 2,453 of them had at
>> least one hit, for a success rate of 66.567%.
>
> So, it would also appear from these results that each prediction had a
> mean of 4.36 earthquakes satisfying the parameters of a prediction.
> Without even looking at the predictions, that makes it appear that the
> prediction are extremely non-specific.
Yes and no. I am very specific in the description of the predictions.
But no very specific given the paramaters of the prediction. Keep
in mind, however, that these parameters were chosen for a specific
reason, namely to show the futility of a specific quake forecasters
method.
> Now, the other question that is left unanswered, is what the expected
> success rate would be through pure chance. In other words, what if you
> took the same predictions, but randomized their starting times. How
> many of them would have also been successful?
That indeed is a very good question. An attempt was made at doing a
statistical anlysis of my method, but it resulted in a ridiculously
high significance, which is obviously wrong given how I do the
predictions. What that showed was that the analysis method was not
up to snuff and could not account for the way I spiked the punch.
It's no secret anymore. I quite literally cheat.
Felix Tilley
> It appears English is not your native language, so forgive me
> for not understanding your question....
She is posting from Chile:
whoarin 200.104.27.27
OrgName: Latin American and Caribbean IP address Regional Registry
OrgID: LACNIC
Address: Potosi 1517
City: Montevideo
StateProv:
PostalCode: 11500
Country: UY
ReferralServer: whois://whois.lacnic.net
NetRange: 200.0.0.0 - 200.255.255.255
CIDR: 200.0.0.0/8
NetName: LACNIC-200
NetHandle: NET-200-0-0-0-1
Parent:
NetType: Allocated to LACNIC
NameServer: NS.LACNIC.NET
NameServer: TINNIE.ARIN.NET
NameServer: NS-SEC.RIPE.NET
NameServer: SEC3.APNIC.NET
NameServer: NS2.DNS.BR
NameServer: NS1.AFRINIC.NET
Comment: This IP address range is under LACNIC responsibility for further
Comment: allocations to users in LACNIC region.
Comment: Please see http://www.lacnic.net/ for further details, or check the
Comment: WHOIS server located at whois.lacnic.net
RegDate: 2002-07-27
Updated: 2005-12-05
OrgTechHandle: LACNIC-ARIN
OrgTechName: LACNIC Whois Info
OrgTechPhone:
OrgTechEmail: whois-...@lacnic.net
# ARIN WHOIS database, last updated 2006-09-29 19:10
# Enter ? for additional hints on searching ARIN's WHOIS database.
Found a referral to whois.lacnic.net.
% Joint Whois - whois.lacnic.net
% This server accepts single ASN, IPv4 or IPv6 queries
% Copyright LACNIC lacnic.net
% The data below is provided for information purposes
% and to assist persons in obtaining information about or
% related to AS and IP numbers registrations
% By submitting a whois query, you agree to use this data
% only for lawful purposes.
% 2006-09-30 20:54:16 (BRT -03:00)
inetnum: 200.104.0/18
status: allocated
owner: VTR BANDA ANCHA S.A.
ownerid: CL-VPNS-LACNIC
responsible: Italo Sambuceti
address: Reyes Lavalle, 3340, 4th floor
address: 6760335 - Santiago -
country: CL
phone: +56 02 3101502 []
owner-c: ISO
tech-c: ISO
inetrev: 200.104.0/18
nserver: NS00.VTR.NET
nsstat: 20060928 AA
nslastaa: 20060928
nserver: NS01.VTR.NET
nsstat: 20060928 AA
nslastaa: 20060928
created: 20021223
changed: 20021223
nic-hdl: ISO
person: Italo Sambuceti Oyarzn
e-mail: isam...@VTR.CL
address: Reyes Lavalle, 3340, 4 th floor
address: 676-0335 - Santiago -
country: CL
phone: +56 02 3101609 []
created: 20020906
changed: 20021122
Skywise
news:05fth2tlumh47ghh5...@4ax.com:
> On Thu, 28 Sep 2006 05:01:01 -0000, in sci.geo.earthquakes, Skywise
> <in...@oblivion.nothing.com> wrote:
<Snipola>
>>I have thought of applying my method to smaller scale regions,
>>such as western North America, or even just Southern California.
>>Of course, I would necessarily use much smaller circles. In
>>fact, I'd like to test the method to see how small of a circle
>>I can use and yet still achieve apparently good success rates.
>
> as the size drops, I would expect your hit rate to drop. big areas
> make for bit hit rates.
Very true. So the challenge would be to place those smaller
circles very strategically in order to maintain a high hit
rate. Can it be done? That's the question.
<Snipola>
> Direct hit 100 points
> 1st 05% 95 points
> 1st 10% 90 points
> 1st 20% 80 points
> 2nd 20% 60 points
> 3rd 20% 40 points
> 4th 20% 20 points
> 5th 20% 10 points
> last 10% 05 points
> last 05% 01 point
> out of range 0 points
>
> And by using the three elements for prediction
>
> A set range of circles (this system could be used for variable circle
> sizes... (with a little work making the evaluation a bit more
> logarithmic)
This idea is intriguing, but this score would only be meaningful when
comparing the results of two predictions against each other. When more
quakes are involved total points alone is not enough. I mean, six
10-point scores is equal to one 60-point score. So which method is
better?
>>My original goal was simply to blow the pants off everybody
>>else with an insanely high hit count. I think I've achieved
>>that.
>
> you have.. (but you "played the system" and done it very honestly...
That part about honesty could be said to be the whole issue.
>>I've also thought of ditching circles and using other boundary
>>shapes.
>
> Oh, Five or Six sided?
I was thinking of irregular polygons, non symmetrical. Although
the ideal would be some sort of parallelogram not unlike the
watchboxes used by weather forecasters.
>>But alas, time has suddenly become precious for me, so it
>>may have to wait while I concentrate on other work.
>
> Work! oh, such a nasty word...
Well, due to work I've not as much free time as I once had. My
current quake related project is starting with a crash course
in statistics. I know simple stuff, but I need to use more
complex methods. I've got this idea that I may be able to turn
the project into a lesson on statistics while also achieving
the main goal.
Skywise
news:pan.2006.09.30....@cyberbromo.int:
> On Thu, 28 Sep 2006 05:01:01 +0000, Skywise wrote:
>
>
>> It appears English is not your native language, so forgive me
>> for not understanding your question....
>
>
> She is posting from Chile:
<Snipola>
Well, that explains that.
My comment wasn't meant in a bad way.
Alexis, if you are reading this, I hope you'll ask the question
again in a different way.
Skywise
> On Sun, 01 Oct 2006 05:20:45 -0000, in sci.geo.earthquakes, Skywise
> <in...@oblivion.nothing.com> wrote:
>
> Note: Crossposted to AEP...
>
> <Maybe you should have named it alt.predictions.earthquakes>
I had considered that, but went with the idea that it was
mimicking the way we say it, "earthquake predictions". But
I do realize the logic for reversing it like you mentioned.
Oh well....what's done is done. I think it will work once
Google picks it up. I hate to admit it, but Google is a
becoming an important part of Usenet.
<Snipola>
> Each Quake prediction would receive a single score. Each score would
> be based upon 100 points for position, 100 points for time, and 100
> points for Magnitude...
>
> Each prediction would receive a score... 0-300... and each score you
> be broken out with a P + T + M as well as a total... P=5 T=50 and
> M=5 would make a score of 55.
>
> The senseless claiming of hits with huge circles, wide time line
> windows and big spreads of magnitude would have to be disallowed, and
> a maximum spread would have to be in place...
OK, I think I see where you're goign with this. However, I
think maybe instead of rings, an algorithm that applies a
smooth curve would be better. Hmmmmm....and if the algorithm
were designed correctly, larger circles (or windows) would
automatically be weighted less.
Come to think of it, we may be coming full circle on this
idea and ending up with a standard statistical analysis.
<Snipola>
> Want to swap jobs...
<Snipola>
NO THANKS!!! I'll keep my five 10 hour days. :)
Skywise
> On Mon, 02 Oct 2006 04:15:44 -0000, in sci.geo.earthquakes, Skywise
> <in...@oblivion.nothing.com> wrote:
<Snipola>
>>Come to think of it, we may be coming full circle on this
>>idea and ending up with a standard statistical analysis.
>
> Maybe.. there has to be a method to include the extreme fuzziness.
I am currently giving myself a crash course in statistics. It's
just a matter of time needed to read the book.
Once done and I've accomplished my immediate goal, I want to
work on this idea of fuzzy statistics.
>><Snipola>
>>> Want to swap jobs...
>><Snipola>
>>
>>NO THANKS!!! I'll keep my five 10 hour days. :)
>
> and When I think I have it bad, I look at those people on rotating
> shifts...
>
> I have to see if I can get a look at the effiency test software the
> carrier uses to rate our proformances at work. Multiple Elements, and
> tends to be fuzzy, some items are rated hard and others are fuzzy...
> Lots of work with Delta-T and Forces, and distances...
>
> I have had a couple of event tapes pulled after incidences, and they
> found nothing I did was questionable or "wrong".
Wait. Do I recall correctly that you drive certain very large
and looooooong vehicles? And you work 12 hour days? Hmmmmm.....
That's kinda scary. Fatigue must be really difficult to fend off
with the monotony of the track.