It’s an interesting point and I somewhat agree but I think the example takes it too far. The most important information is not the sort column but rather the country name. If I want to see who is leading, I don’t want to scan to the final column on the page to see the country names. In fact, if I just want to see which country is leading (and I don’t strongly care about the sort order chosen) then I might only look at the country names and ignore the detailed counts entirely. Whereas if I only see the total medal count or gold medal count column, then I basically have no information since I still need to scan to the last column to find the country names they are associated with.
So the proposed solution here is "use two tables." The rearrangement of the column order is somewhat arbitrary and unconventional. Seems we haven't solved the real problem.
I think to answer the question of "how should the scores be ordered," we simply need to ask "how do you win the Olympics?" Is it the most golds? Most total medals? Weighted sum of all medals? Order it that way, it's the only way it makes sense for a leaderboard to be ordered. If one does not "win the Olympics," then there's really no meaning to tease out of any potential leaderboard ordering, so order it however you'd like.
The Financial Times’ alternative medal table ranks countries not by their total medal haul but by the difference to the tally they are expected to achieve, according to an economic model that takes into account their economic, social and political characteristics:
1. Russian Olympic Committee
2. Australia
3. China
4. Italy
5. Netherlands
6. Taiwan
7. Germany
8. Cuba
9. Austria
10. New Zealand
That's a little silly, since in reality athletic skill is something acquired over time, but doesn't really require much economic input, relative to the overall economy.
> To predict the baseline total number of medals each country might be expected to win in any given summer Olympics, we updated a linear regression model produced by Julia Bredtmann, Carsten J Crede and Sebastian Otten ahead of the Rio Olympics.
> Mimicking the original work by Bredtmann, Crede and Otten, we built the same model and fitted it on data for each Summer Olympic year between 1992 and 2008, and then used it to predict the medals table at London 2012. The full model outperforms a naive model that uses only the number of medals a country won and the year of the Games.
> We then fitted the same model on data up to and including the 2016 Rio Games, and used this to predict medal counts for Tokyo 2020 (using the latest available data for each input, i.e forecasts for GDP per capita and population in 2021).
So they don’t have any tables describing their regression coefficients, but one of the params they use is lag-1 medals. For example, predicting 2020 medals using 2016 medals.
It wouldn’t surprise me if the vast majority of the model’s predictive power came from that, as opposed to GDP per capita, whether the country is Muslim (yes this is actually a predictor too!)
The whole point of a country medal table is to rank countries, not the countries' athletes. If a country is able to produce a better outcome by having a stronger economy and more social equity, that should be rewarded. Ranking outliers relative to predictions is only useful for figuring out which countries' doping programs to investigate.
I would choose weights for gold, silver, and bronze and then sort accordingly, this would be confusing which would make the consumer of information pay attention to more of the details.
Yes most readers know to look for a triangle icon or something similar to indicate which column is currently the sort column if there are multiple options. Changing all the column orderings does feel heavy handed. It’s like if spreadsheets always had to sort by column A, then column B, etc.
This post is really about display of static tables, not interactive ones.
But in the interactive scenario, you could imagine dragging a column heading left or right to change the sort order - I think it could be quite intuitive.
Don’t you think the most important column may not be the one you sort by? Most data will have an ID or label (country name in this case) that differentiates the row and gives it meaning. Putting this column last obscures the meaning of the rows.
I think the most important thing to convey straight off is the overall structure or order, and the importance of an individual column is secondary to that.
I don't think the most important column necessarily has to be on the left - in fact these tables already often have the rank to the left of the country name. There are many other ways to convey column importance, such as bolding the values.
The rank being on the left is like a row index. It helps you count without needing to count from the beginning of the column. For this reason, many data-oriented languages like R don’t even consider the row index to be part of the data.
If anything the ranking reinforces the idea that a given country is in a given position which is the overall structure and intent of the data (in my view). I’m not likely to look at the table to extract the fact that “7 gold medals is 17th place.” I’m going to see that “country X is 17th place.”
Not opposed to a ranking on the left. Rank and score have equal sorting order precedence (one is derived from the other) so based on what I'm suggesting a golf leaderboard would look like:
I choose the chaotic option of sorting by the score:
3 * gold + 2 * silver + bronze
Sorting by gold feels as incorrect to me as sorting by total medals. Sure, a bronze isn't as good as a gold, but sorting by gold is silver/bronze erasure.
3 * gold + 2 * silver + bronze + β1 / population + β2 / GDP
where β1 and β2 are some coefficients.
How do we choose these coefficients? Maybe try to run a simple regression to see how historically predictive pop and GDP are of a nations score w/o the GDP and pop terms.
Gold is the only one that matters. Including two more consolation places is completely arbitrary. (Why not have one and stop at silver? Why not have three and include copper after bronze?) It just benefits countries that can field larger teams. Number of second place finishes should only be used as a tie-breaker.
That is actually what pretty much every medals table, including the official table on olympics.com, already does (except in the USA because it's important to say we're number one, even if it's number one in third place finishes).
If one country managed to win all the silver medals, while each of the other countries managed to get one gold medal each, which country got the best results at the Olympics?
There isn't one perfect scoring system in this case, but disregarding everything but gold would completely ignore the value of second place, which is still a win over all other competitors but the first place
The silver medal is a consolation prize. The other countries would tie for first in the number of first place finishes, which is all that matters. Again, this is what most medals tables already do.
Again, silver and bronze are consolation prizes, so extrapolating even further from there is a fool's errand. Some sports (including most of the team sports) don't even allow countries to field multiple entries, with the strong implication that those other places don't matter.
Tournament-based sports at the Olympics don't even bother to define a ranking after fourth, and many sporting events outside the Olympics don't even bother determining ranking after second and only go as far as second because it's needed to determine who is first.
To quote the Big L, "that's just, like, your opinion man".
I, for one, think we do an incredible disservice to the athletes that perform to not acknowledge other places.
When an athlete manages to beat an Olympic or World record and comes in second by a hair, I think its pretty silly to state "other places don't matter".
First place is better than second is better than 8th is better than 11th.
If these places didn't matter, why would we bother keeping record of standings?
Because that way of thinking encourages sending hundreds, thousands, or millions of competitors and because (as shown by sports that don't allow one country to win multiple medals) the organizers of the sports don't think those other places matter either. We don't try to determine the world rankings of people who failed to qualify for the Olympics, nor should we try to determine the rankings of teams that fail to qualify from group play into elimination rounds of the Olympics, nor should we try to rank teams that are eliminated from elimination rounds that haven't played each other. They're top athletes, but the point of each Olympic sport is to determine the best. Any additional ranking computed beyond that is purely arbitrary.
If you think a silver at the olys is a "consolation prize" then you have never spent a moment in any type of real athletic competition, let alone been among the best of the best on the planet.
The level of competition does not change the definition of the phrase. If you think it does, you have never used a dictionary before, let alone understood my comments.
It might have been arbitrary many decades ago to decide the cut-off to be at 3, but the athletes have been since optimizing for those arbitrary rules, i.e. getting somewhere above that cut-off.
Instead of maximizing for first place, which requires a different strategy, i.e going „all in“.
Author here, agree that it would be a bit clearer if the multi-row values were vertically centered or otherwise visually highlighted. Unfortunately I didn't find the time to implement this for this post.
A different way of thinking about it would be considering each Olympic trial as a voter and use gold/silver/bronze as 1st/2nd/3rd in a ranked-choice vote. Each "contest" effectively votes on a country.
It feels wrong to me to view it this way. The first column of a table should be the item, and the remaining columns should be the details of the item. This implies 77 total medals is the item, and the detail of that is the US. It just parses wrong in my head.
I almost wonder if the proper order isn't some sort of % of medals won vs participants for each country, so countries that have few athletes in the competition but all medal in most of the events they're in show highly.
Then again, I'm willing to admit I have no clue what people that look at these charts are after.
The rank goes with the label. Who got first? Tiger Woods got first. It isn't helpful to know that someone shot a 68 and got first without knowing who it is.
In your examples for the Olympics, the name is the very last column. So I would need to scan through Rank, Gold Medals, Silver Medals, Bronze Medals, and Total Medals to find the name that the Rank belongs with. Compare this to Rank and Silver Medals which have nothing to do with each other. You can be rank 1 with 0 silver medals. If sorting by Total Medals you could have 0 gold medals and be in rank 1. So the rank belongs with the name. The most basic question everyone answers with the table is “which country is in first?”. Some people may wonder “how many silver medals did the first place country get?” and obviously the point of a table is to allow them to find this information too, but it’s of much lower interest to most readers. For example, you’d never find a table that just showed Rank and Silver Medals (or even Rank and Gold Medals) but it’s common to see a table that only shows Rank and Country or in the case of golf Rank and Player Name. Who won the Masters? It wasn’t the score that won, it was the player
Sort by total and use color to highlight the leader in each specific medal count (i.e. China's gold count should be given a gold color treatment, US for silver/bronze).
> Once again, the table flows naturally from left to right. The most critical column is leftmost(number of gold medals).
Hm, I don't think so. The most critical bit of information is the country, the Olympics are a contest of countries. To wit, I'm sure many publications will order the table so that it best pleases their audience, i.e. places their country the highest.
When I, and I would bet most, people are looking at these lists they're looking for the top three countries, or their own, the exact reason for each country's place is far less material than the
ranking system that the table maker chose.
I've no doubt many people look at a medal table, look for their own country's name, confirm it is near the top, and close the tab without examining it further.
But I'm not sure that is the use case we should be optimizing for!
Looks like New Zealand is on top, then Australia, Netherlands and Great Britain (there might be stronger countries down the list, I just calculated the first scores).
China is last, the US second to last and then Germany.
United States 0.508836076
People's Republic of China 0.117310443
Russian Olympic Committee 0.754847645
Great Britain 1.48721921
Japan 0.752177356
Australia 2.996845426
Germany 0.734939759
Italy 0.928689884
France 0.745600954
Netherlands 2.604166667
New Zealand 7.551020408
I wonder what it says about countries/societies sport system that are near to each other, like France, Germany, Japan and Russia at 0.75
71 comments
[ 2.7 ms ] story [ 133 ms ] threadI think to answer the question of "how should the scores be ordered," we simply need to ask "how do you win the Olympics?" Is it the most golds? Most total medals? Weighted sum of all medals? Order it that way, it's the only way it makes sense for a leaderboard to be ordered. If one does not "win the Olympics," then there's really no meaning to tease out of any potential leaderboard ordering, so order it however you'd like.
> Mimicking the original work by Bredtmann, Crede and Otten, we built the same model and fitted it on data for each Summer Olympic year between 1992 and 2008, and then used it to predict the medals table at London 2012. The full model outperforms a naive model that uses only the number of medals a country won and the year of the Games.
> We then fitted the same model on data up to and including the 2016 Rio Games, and used this to predict medal counts for Tokyo 2020 (using the latest available data for each input, i.e forecasts for GDP per capita and population in 2021).
Model: https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1740-...
It wouldn’t surprise me if the vast majority of the model’s predictive power came from that, as opposed to GDP per capita, whether the country is Muslim (yes this is actually a predictor too!)
But in the interactive scenario, you could imagine dragging a column heading left or right to change the sort order - I think it could be quite intuitive.
I don't think the most important column necessarily has to be on the left - in fact these tables already often have the rank to the left of the country name. There are many other ways to convey column importance, such as bolding the values.
If anything the ranking reinforces the idea that a given country is in a given position which is the overall structure and intent of the data (in my view). I’m not likely to look at the table to extract the fact that “7 gold medals is 17th place.” I’m going to see that “country X is 17th place.”
- USA: 167
- CHN: 164
- ROC: 109
- GBR: 96
- JPN: 95
Which I think makes the ranking the same as total medals for the top 5.
EDIT: Spreadsheet is here https://cryptpad.fr/sheet/#/2/sheet/view/yFp7KdPHA2hC1cBOL0A...
https://www.bloomberg.com/graphics/tokyo-2020-summer-olympic...
The US got a lot of recent medals which increases the lead
How do we choose these coefficients? Maybe try to run a simple regression to see how historically predictive pop and GDP are of a nations score w/o the GDP and pop terms.
New Zealand is at the top, then Australia, the Netherlands and GB (probably some others, I only calculated the number for the top of the rankings).
That is actually what pretty much every medals table, including the official table on olympics.com, already does (except in the USA because it's important to say we're number one, even if it's number one in third place finishes).
There isn't one perfect scoring system in this case, but disregarding everything but gold would completely ignore the value of second place, which is still a win over all other competitors but the first place
Those were the calculations for video games winter games/summer games in the 1980. Why not use that?
Ranking each country by the sum of the normalised reciprocal of their placements in events.
I think that would be pretty cool.
Ranking by gold or ranking by total medals in my mind remain terrible ways to rank.
Tournament-based sports at the Olympics don't even bother to define a ranking after fourth, and many sporting events outside the Olympics don't even bother determining ranking after second and only go as far as second because it's needed to determine who is first.
I, for one, think we do an incredible disservice to the athletes that perform to not acknowledge other places.
When an athlete manages to beat an Olympic or World record and comes in second by a hair, I think its pretty silly to state "other places don't matter".
First place is better than second is better than 8th is better than 11th.
If these places didn't matter, why would we bother keeping record of standings?
I just don't get the "Gold or last" mentality.
Instead of maximizing for first place, which requires a different strategy, i.e going „all in“.
gold - $815
silver - $445
bronze - $5
https://en.as.com/en/2021/07/23/olympic_games/1627028063_790...
I almost wonder if the proper order isn't some sort of % of medals won vs participants for each country, so countries that have few athletes in the competition but all medal in most of the events they're in show highly.
Then again, I'm willing to admit I have no clue what people that look at these charts are after.
https://www.espn.com/golf/leaderboard
I don't understand what you mean. Why does the rank go with the label, but the score doesn't?
> It isn't helpful to know that someone shot a 68 and got first without knowing who it is.
But you do know - the name is right there next to the score.
That "their" country is as high on the list as possible, and better than "neighbours".
Hm, I don't think so. The most critical bit of information is the country, the Olympics are a contest of countries. To wit, I'm sure many publications will order the table so that it best pleases their audience, i.e. places their country the highest.
When I, and I would bet most, people are looking at these lists they're looking for the top three countries, or their own, the exact reason for each country's place is far less material than the ranking system that the table maker chose.
But I'm not sure that is the use case we should be optimizing for!
Removing that part, none of it makes much sense (if you only care about individual athletes or disciplines, you’ll be fine with annual world events)
Looks like New Zealand is on top, then Australia, Netherlands and Great Britain (there might be stronger countries down the list, I just calculated the first scores).
China is last, the US second to last and then Germany.
I wonder what it says about countries/societies sport system that are near to each other, like France, Germany, Japan and Russia at 0.75Where do you see Russia on that list?
Do you think state-sponsored doping should be ignored since sanctioning it is a mild inconvenience for you?