Thursday, August 16, 2012

Apple Time Machine backups should not delete oldest, thin to monthly when space becomes short

The new version of the Time Machine backup capability in Mac OSX Mountain Lion has important improvements.

Most notably it transparently allows one to use multiple disks for backup. I use one backup disk at home and one at work. These allow me to be more confident that if one disk crashes I still have the other, and wherever I am working, I have an hourly backup to fall back on.

However here's a suggestion for the next improvement Apple could make: Currently the policy for maintaining backups is (quoting directly from the Time Machine control panel shown below):

  • "Hourly backups for the past 24 hours
  • Daily backups for the past month
  • Weekly backups for all previous months

The oldest backups are deleted when the disk becomes full."

My suggestion is this: When the disk becomes full, older backups should be thinned to monthly (up to three months ago) starting with the oldest. Weekly backups should always be retained for three months.

So if the disk becomes full, it should not delete the oldest backup, instead it should delete the second-oldest, and what was the third and fourth oldest, leave the what was the fifth oldest, and delete what was the sixth oldest, etc.

Why? Sometimes missing files can only be retrieved from deep history. With monthly backups far back in time, one has a pretty good chance of being able to retrieve such files. Weekly backups several years old are likely not needed.


In my experience I have retrieved lost work from Time Machine about 3 times a year. Not much. However, backups are all about insurance and peace of mind. Who knows when I will have a disk crash or do something catastrophic to my documents.

This post updates my earlier post on backups. Note that I no longer backup to Mobile Me since it has gone away. Some of my data is backed up on iCloud, although I personally feel it is critical to maintain one's own personal backup of cloud-stored data. I still back up using SuperDuper periodically in addition to Time Machine, although now I have two Time Machine disks, I will be updating my SuperDuper partitions less often.

Sunday, August 12, 2012

Fair London 2012 Olympic medals ranking - Final results

Here are my 'fair' Olympic medals rankings, based on the final results of the London 2012 Olympics, after the final medals were awarded on August 12th. This updates my earlier posts on August 8th and August 9th, which proved quite popular.

What have I done to make the rankings fairer:
  1. I have weighted the rankings by population.
  2. I have further normalized by UN development index. Developed countries have an index closer to 1, but developing countries, which have much less capacity to train athletes, are ranked lower, typically in the 0.5 range.
  3. I have weighted silver medals as worth two bronze medals, and gold medals as worth two silver medals. This seems considerably fairer than the two schemes widely used by the International Olympic Committee and the press, either ranking strictly by gold medals (breaking ties with silver and bronze), or else weighting all medals evenly.
The first table below gives my final rankings, ordered by this scheme. The raw score is column WM10M, and the ranking is column WMR.

Population and development index, used for normalization, are shown on the right. 

In all the columns in the table below, I have highlighted the top 20 countries (and ties) in the different rankings, so you can see how the different schemes affect the sequence. My preferred WM10M ranking is in blue. I have also marked the top 20 countries by population, and by development index, using grey.

Small island countries of Grenada and The Bahamas come first and second in WM10M score by virtue of winning a medal despite their extremely small population. Jamaica and New Zealand come third and fourth. Congratulations to those countries.

Column definitions
  • Gold, Silv, Bro - medal counts
  • GR: Ranking, based on Gold medals, with ties resolved by silver and bronze (top 20 orange)
  • TotM: Total medals (top 20 yellow)
  • WTotM: Total medals weighted using the 4 2 1 scheme described above (top 20 green)
  • G10M: Gold medals per 10 million, weighted by development index (purple)
  • M10M: Total medals per 10 million, weighted by development index (pink)
  • WM10M: 4-2-1 weighted medals per 10 million, weighted by development index (blue)
The second table below shows the same data, organized by population. The third table shows the same data, organized by the 'Gold medal' rankings used by the IOC.

No ranking will ever be totally fair. Maybe team sports, or sports which only give one medal such as modern pentathlon, should be given higher rankings for the medals. But there is no clear way to do come up with an unbiased ranking scheme if those adjustments are made. I think what I present is a good compromise.


Table 1: Rankings ordered by WM10M: 4-2-1 weighted medals per 10 million, weighted by development index.



Table 2: The same data ordered by size of country (PopM)

Table 3: The same data, ordered using the IOC's gold medal based rankings

Thursday, August 9, 2012

Olympic medals ranking August 9 by population and development index

(See here for my final update as of the end of the Olympics)

Yesterday's post was very popular, so here is an update one day later. The data below are as of mid day August 9th, 2012.

I have ordered countries explicitly here by my WM10M ranking, which is medals per 10 Million people, but normalized for UN development index, and where a silver counts double the weight of a bronze and a gold counts as double the weight of a silver. For explanations of the other columns, see yesterday's post.

Jamaica passed New Zealand to move into second place today.

I will post a final ranking at the end of the Olympics, and may do one or two between now and then.

A raw spreadsheet is here.


Wednesday, August 8, 2012

Olympic medals rankings normalized by country population and development index

(See here for my final update to this as of the end of the Olympics)

We all see China and the US jockeying for position at the top of the Olympic medal counts. But which countries are really doing best?

The table below orders countries by population (right column marked PopM). Coloured cells are the top 20 (and tied) countries according to various ranking criteria.

Orange is the 'gold medal' ranking, with ties resolved by silver and bronze medals. This is the ranking used by most rankings tables. Yellow is total medals. Green (WTot, WMR) is total medals weighted such that Gold has a value of 4, silver a value of 2 and bronze a value of 1. This is the system I prefer when population is not considered.

Purple (G10M, GMPR) is gold medals per 10 million of population , but normalized by UN development index (DevInd column). A country with a lower development index has fewer resources, so should be expected to be capable of training fewer athletes. Light orange (M10M, MPR) is total medals, again weighted by population normalized by UN development index. Finally blue (WM10M, WMR) is weighted total medals (as per the green rankings) weighted by normalized population.

Grenada does best in most rankings by virtue of its one gold medal and small population. New Zealand is doing best in countries with moderate population. Other countries doing well are Jamaica (due to their sprinters) and Slovenia.

Great Britain is the large country (by population) doing best. North Korea and Australia also stand out.

The US and China a quite far down, at 37 and 58 in my blue (WMR) rankings, suggesting that although they are doing well overall, they are not doing as well when population and development index are both factored in.

The table notably shows very populous countries with few medals: India and Indonesia stand out near the top.

The data is as of midday Wednesday August 8th. I intend to update it at the end of the Olympics. The original spreadsheet can be found here.

I wish the media would report data like this for all to see.