Smaller is better charts

[XWAUForceflow]

Okay, this is totally random, but whenever I see one of LTTs (or actually many others) reviews where they have multiple charts that show performance where on some charts lower numbers are better I get annoyed that I always have to find that information first. Unless you know the tests it's just not intuitive and checking for the 'smaller numbers are better' info just takes too long. (Yes, yes, I am too old, so get off my lawn)

So I was wondering it if wouldn't make sense to instead of this:

 

You actually do this:

That way the way the charts design inherently shows what orientation one should look at and for me I feel it would be much more intuitive.

Really, totally random and completely off-topic, but I was just curious what others think.

[Lurking]

The best (regardless of if best is lowest or largest number) should be on top. That's the most intuitive and most reviews do that. The statement about lower/higher being better should still be there to be correct. But at first glance isn't needed 

 

What you show with the worst being on top I've never seen. And starting the bar from the right wouldn't help.

[tikker]

Having to find that information is inherent to charts. If you are talking about temperatures or render times, lower numbers are better. If you are talking about maximum sustained clock speeds or write speeds, higher numbers will be better. There isn't really a catch all for it.

 

Personally I find the second chart more confusing than the first one. How does it speed up legibility or understanding? It takes me more time to link the labels to the bars due to the white space between them cause by bars and labels using different axes, it's generally counterintuitive that the horizontal axis decreases towards the right and in the end it doesn't fundamentally change that you still need to look up "smaller numbers are better" to be able to understand that shorter bars  imply better results/performance.

[IkeaGnome]

9 minutes ago, tikker said:

Having to find that information is inherent to charts. If you are talking about temperatures or render times, lower numbers are better. If you are talking about maximum sustained clock speeds or write speeds, higher numbers will be better. There isn't really a catch all for it.

 

Personally I find the second chart more confusing than the first one. How does it speed up legibility or understanding? It takes me more time to link the labels to the bars due to the white space between them cause by bars and labels using different axes, it's generally counterintuitive that the horizontal axis decreases towards the right and in the end it doesn't fundamentally change that you still need to look up "smaller numbers are better" to be able to understand that shorter bars  imply better results/performance.

I think I see what OP was thinking here. I do agree that it makes it way more confusing than it already is or needs to be though. If you look "outside" the bars, it becomes a "reverse graph", and then the lower number that is better has a bigger "bar"

[tikker]

24 minutes ago, IkeaGnome said:

I think I see what OP was thinking here. I do agree that it makes it way more confusing than it already is or needs to be though. If you look "outside" the bars, it becomes a "reverse graph", and then the lower number that is better has a bigger "bar"

Hmm I see. It's an idea, but to me that adds more work to it as I now have to actively think about the negative space.

[Lurking]

Whatever the solution will be, it will be a solution in search of a problem.

 

It isn't that hard to understand that in some tests less or more is better. It is just arbitrary anyway. If you measure fps, higher is better. But you also could show milliseconds per frame, then lower would be better. 

 

Like a car shows a maximum speed (larger is better) and 0-60 number (smaller is better). No one sees an issue with that. It is selected for humans to understand since showing an average acceleration would be hard to comprehend 

 

All reviewers kind of follow the same scheme and it is easy to follow. No need to change to a more complex system.

[XWAUForceflow]

On 3/18/2023 at 5:25 PM, Lurking said:

The best (regardless of if best is lowest or largest number) should be on top. That's the most intuitive and most reviews do that. The statement about lower/higher being better should still be there to be correct. But at first glance isn't needed 

 

What you show with the worst being on top I've never seen. And starting the bar from the right wouldn't help.

The problem with that is that if you have multiple options (for example 10 GPUs) and they keep switching positions between graphs it's again harder to find what you are looking for. Personally I find that much more confusing especially if I want to compare a new GPU for another specific GPU in that list. Much harder to find those two if they keep switching positions.
I admit my graph implied a sorting that was not intended. Maybe a better sample would be this:

 

So those would be two slides following each other where the first one has values like FPS and the second one could be rendering times.

For me it would be very easy to pick out at one glance that on the first test the nVidia GPU won with a small lead over the AMD GPU and then transition directly to the second graph, still knowing which GPU is in which position and still focusing on the right side of the screen instantly seeing that here the roles are reversed.

 

Note that I am talking about graphs in the videos that are often shown for a short period of time. On an website where I control when to switch to the next graph that isn't an issue.

21 hours ago, Lurking said:

Whatever the solution will be, it will be a solution in search of a problem.

 

It isn't that hard to understand that in some tests less or more is better. It is just arbitrary anyway. If you measure fps, higher is better. But you also could show milliseconds per frame, then lower would be better. 

 

Like a car shows a maximum speed (larger is better) and 0-60 number (smaller is better). No one sees an issue with that. It is selected for humans to understand since showing an average acceleration would be hard to comprehend 

 

All reviewers kind of follow the same scheme and it is easy to follow. No need to change to a more complex system.

I don't understand your point. I don't want to change the tests or the numbers of the results. I only want to change the representation of those numbers. So the maximum speed graphs would start on the left and go to the right, and the acceleration bars would start from the right and go to the left. Everything else would remain the same.

Upside for me would be that I don't have to know that higher speed and lower acceleration is better because of the way the bars are shown. I simply know that every graph that is from left to right means higher is better and if it's from right to left lower is better.

Now obviously for those well known cases it's easy to know. But going back to the GPU tests there are often a lot of productivity tests where I don't really know what they measure, hence I do not know by the name of the test if larger or smaller is better.

 

22 hours ago, tikker said:

Hmm I see. It's an idea, but to me that adds more work to it as I now have to actively think about the negative space.

Yeah, I guess everybody is wired differently.

For me it would be easier to simply keep my focus to the right of the bar graph and simply see which graph starts farthest to the right. (Just like before I had to look which graph stopped farthest to the right) But that's just preference, there's really no right or wrong here. I was just curious what others thought of that proposal.

[Lurking]

29 minutes ago, XWAUForceflow said:

I don't understand your point. I don't want to change the tests or the numbers of the results. I only want to change the representation of those numbers. So the maximum speed graphs would start on the left and go to the right, and the acceleration bars would start from the right and go to the left. Everything else would remain the same.

Upside for me would be that I don't have to know that higher speed and lower acceleration is better because of the way the bars are shown. I simply know that every graph that is from left to right means higher is better and if it's from right to left lower is better.

Now obviously for those well known cases it's easy to know. But going back to the GPU tests there are often a lot of productivity tests where I don't really know what they measure, hence I do not know by the name of the test if larger or smaller is better.

You propose to not list the results in order of the result, but to keep the order of models listed the same. This will be highly un-intuitive for anyone not looking at a specific model. Like if an article is about 3 CPUs, those will be highlighted out of the maybe 20 listed. So it is easy to find the 3 in question. Of course, in each test they will be in a different position. This also wouldn't work if one test doesn't include all the same hardware. Like if the article is about 3 CPUs, they may not have legacy results for all hardware for all the tests (i.e. one benchmark didn't exist when older hardware was introduced and tested).

 

Maybe the reviewer could add a selection drop down for you to select "your" hardware and then that will be highlighted. But this would be a lot of effort and clutter and also wouldn't help if you want to look at multiple CPU/GPUs. You could use the AnandTech bench where you select 2 pieces of hardware to compare and that narrows down the list. 

[XWAUForceflow]

21 minutes ago, Lurking said:

You propose to not list the results in order of the result, but to keep the order of models listed the same. This will be highly un-intuitive for anyone not looking at a specific model. Like if an article is about 3 CPUs, those will be highlighted out of the maybe 20 listed. So it is easy to find the 3 in question. Of course, in each test they will be in a different position. This also wouldn't work if one test doesn't include all the same hardware. Like if the article is about 3 CPUs, they may not have legacy results for all hardware for all the tests (i.e. one benchmark didn't exist when older hardware was introduced and tested).

 

Maybe the reviewer could add a selection drop down for you to select "your" hardware and then that will be highlighted. But this would be a lot of effort and clutter and also wouldn't help if you want to look at multiple CPU/GPUs. You could use the AnandTech bench where you select 2 pieces of hardware to compare and that narrows down the list. 

I specifically wrote earlier that I am talking about graphs in videos, not on webpages. And in LTT videos graphs are often not sorted by who is best but by a fixed order so having the best on top is not a universal sorting either.

[Lurking]

9 minutes ago, XWAUForceflow said:

I specifically wrote earlier that I am talking about graphs in videos, not on webpages. And in LTT videos graphs are often not sorted by who is best but by a fixed order so having the best on top is not a universal sorting either.

That's a shame. Technically in a video the graph should be equally as logical as in an article. Even more so since you have less time to look at (but they can dynamically change which model they lighttight to match what is spoken about at the moment, GN does that). i don't watch LTT, but every channel i do watch orders by best on top and so on. 

 

It doesn't matter what you or I think. Each reviewer will do what THEY think is best or easiest for them. 

[tikker]

1 hour ago, XWAUForceflow said:

Note that I am talking about graphs in the videos that are often shown for a short period of time. On an website where I control when to switch to the next graph that isn't an issue.

You can pause the video and study the chart though.

1 hour ago, XWAUForceflow said:

The problem with that is that if you have multiple options (for example 10 GPUs) and they keep switching positions between graphs it's again harder to find what you are looking for. Personally I find that much more confusing especially if I want to compare a new GPU for another specific GPU in that list. Much harder to find those two if they keep switching positions.

I still think switching the second form will be highly confusing, since that is in a way "not how bar charts work". I do understand the point about units switching order. It will depend on what the goal of the chart is.

1 hour ago, XWAUForceflow said:

Upside for me would be that I don't have to know that higher speed and lower acceleration is better because of the way the bars are shown. I simply know that every graph that is from left to right means higher is better and if it's from right to left lower is better.

In your examples the axes also flip order, so you've just created a mirrored version which tells you the same information as the first chart, but in a mirrored way, so nothing has really changed about the "X  is better" problem.

1 hour ago, XWAUForceflow said:

Now obviously for those well known cases it's easy to know. But going back to the GPU tests there are often a lot of productivity tests where I don't really know what they measure, hence I do not know by the name of the test if larger or smaller is better.

This is a universal "problem". I put apostrophes there, because charts always need explanation. That is why they put "lower/higher is better" in there, such that the reader knows better what they are looking at. If one is unsure what the test itself is measuring, then it can be looked up or asked on a forum or similar. Having those references of why those benchmarks are chose are good to have and good to refresh every now and then, but it will get superfluous rather quick if every video they explain every benchmark.

1 hour ago, XWAUForceflow said:

Yeah, I guess everybody is wired differently.

For me it would be easier to simply keep my focus to the right of the bar graph and simply see which graph starts farthest to the right. (Just like before I had to look which graph stopped farthest to the right) But that's just preference, there's really no right or wrong here. I was just curious what others thought of that proposal.

I don't think the issue will change? The component order will still change if sorted by value even if the bars are on the right and if you keep component order the same then you may as well keep the normal form of a bar chart and the bar will simply change length.

38 minutes ago, XWAUForceflow said:

I specifically wrote earlier that I am talking about graphs in videos, not on webpages. And in LTT videos graphs are often not sorted by who is best but by a fixed order so having the best on top is not a universal sorting either.

There is no universal sorting. It depends on the message the chart means to send. If the goal is to show a comparison of various workloads then it can indeed be beneficial to keep the order of the components the same. If the goal is to show the fastest and slowest then it makes sense to sort by value and thus the order will change.