Linus ... buddy ... your graphs ... can we fix them please?

[Amias]

Graphs? I'm talking these comparative charts.

 

Example 1

Example 2

Two charts, same layout, format and information. Apart from this little Snippet

 

 

"Higher is better". "Lower is better."

 

Flicking through charts in your videos. Unless catching this tiny snippet, the data is confusing and doesn't clearly get the comparative information across. "Oh the i7 is super fast and powerful", next graph "oh the M1's good at this one" wait ...

 

Higher is better ... Lower is better ... got me again!

 

Two Graphs, showing the same information, only completely backwards. (ignore the titles)

Could we, as a suggestion. Do this instead?

 

Two graphs, two information sets, but the 'higher is better' is obvious and consistent with the latter 'lower is better'.

 

 

Immediately, we know the items 1-5 are ranked 5,4,3,2,1. The graphical information is consistent across slides. And it's easy to quickly compare and contrast.

 

Further right = better. Further left = worst. Consistency.

 

Thank you. 

[Coolmaster]

Their graphs are rarely ranked in order of best, usually it's somewhat arbitrary usually with the main object at the top. If you're making a video about the M1 you wouldn't want everyone focusing on the iMac or the PC you want people to focus on the M1 scores.

[piratemonkey]

hopefully you aren't quickly going through a review to see if you should buy a couple hundred dollar thing. but ya, their graphs could be better, but your solution is quite jarring honestly. especially with the pace they're going at, having to readjust to the bars being on the opposite would be too much

[Slottr]

This would be welcoming to see in tech across the board TBH.

 

I'm looking at you, 50 GPU roundup benchmarks

[WereCatf]

Spoiler

9 minutes ago, Amias said:

 

 

Mmmmno, I'm gonna vote a very definite, hard no on the lower graph there. That's way more confusing.

[eeeee1]

I would imagine that if you really cared that much about the results that you would look more carefully. also if you put how its ranked on the top, where does the title go? basic graph work 101. no matter where you go the title of what is being described should be at the top. whats the difference between a subheading being at the top or bottom?

[Amias]

8 minutes ago, Coolmaster said:

Their graphs are rarely ranked in order of best, usually it's somewhat arbitrary usually with the main object at the top. If you're making a video about the M1 you wouldn't want everyone focusing on the iMac or the PC you want people to focus on the M1 scores.

 

Aye, sorry. Not asking for the Y-Axis to be ranked in that order. 

 

I want the X-Axis to reflect left = worst right = better. So in the case of lower = better, you reverse the order of the X-Axis. 

 

2 minutes ago, eeeee1 said:

I would imagine that if you really cared that much about the results that you would look more carefully. also if you put how its ranked on the top, where does the title go? basic graph work 101. no matter where you go the title of what is being described should be at the top. whats the difference between a subheading being at the top or bottom?

Aye, apologies, I'm not advocating the title position change.

Should have made that more clear. I'm advocating for the x-axis option to follow the rule set of 'left = worst, right = better".

 

As for examining the graphs, kind of difficult when watching a video where they flick through them.

 

 

 

[minibois]

ehhh I don't find your 'fix' any better than what we have now. if anything, I find it worse.

to me, the graphs in the videos are only there to support the voice-over.

 

i.e. "just look how much better product X is compared to product Y"

*shows graph, where you can see a bar is twice as small/large on one vs. the other*

These graphs usually work well enough for that purpose. no need to complicate it with different directions too

[Amias]

4 minutes ago, minibois said:

These graphs usually work well enough for that purpose. no need to complicate it with different directions too

But that's the problem.

 

They are currently in different directions. I'm asking for them to be in one direction. 

[piratemonkey]

1 minute ago, Amias said:

But that's the problem.

 

They are currently in different directions. I'm asking for them to be in one direction. 

no your solution has bars originating from both left and right. LTT's current doesn't

[FloRolf]

Yeah, no, not really. 

Their graphs are fine. If people can't read it's their problem. 

[minibois]

Just now, Amias said:

But that's the problem.

They are currently in different directions. I'm asking for them to be in one direction. 

I meant different directions relative to the screen they are presented on, I didn't mean different directions meaning better/worse.

the better/worse is made apparent by the voice-over already.

 

data is meant to be able to be seen in a glance, understood after knowing the context and/or seeing what the data points mean.

 

looking at a right to left graph and understanding "more to the right = better" takes just as much effort as seeing "oh this measurement is in minutes and seconds, lower = better) in my opinion.

 

but again, the visual medium is less important that the spoken medium in this case, the voice-over provides the context to fully reading the graphs.

[LogicalDrm]

The first set of changes has some merit. But overall, LTT videos are rarely something I would check because of graphs. If I want detailed reviews, I look at other channels. I still wouldn't call most of LTT "reviews" reviews. Only when they have "Should you buy it?" said in the script, then I can take it as review. Rest of the time its overview. Or unboxing on SC.

 

The issue with graphs is actually deeper than just direction of scoring. As you can see from example, the order of items is same across slides used. This isn't always the case, sometimes they change order. And while from better to worst would work, just highlighting the items talked when editing would do same.

[akio123008]

31 minutes ago, LogicalDrm said:

just highlighting the items talked when editing would do same.

Don't they do that already sometimes?

 

Personally I agree with the OP that "lower is better" should have a different type of graph than "higher is better", because if you for some reason miss that phrase, you can easily get confused. There should be a graphical way to distinguish between the two types, not just some phrase below it.

 

However, I think the idea of the backwards graph proposed, as @WereCatfpointed out, is even more confusing.

 

What I would do to make "lower is better" graphs look different from "higer is better" graphs, is to rotate the latter 90 degrees. I could try and explain, but here's an image:

 

 

Whenever something is "lower is better" it's usually some amount of time it takes to do something, so I thought it'd be intuitive to have time on the horizontal axis, as that's usually how graphs that involve time are represented, which is why for that type of chart I'd use horizontal bars.

 

For a benchmark score or something like that, eg "higher is better" it makes more sense to have the magnitude on the vertical axis, hence why I'd use vertical bars for that type.

 

 

 

 

[LogicalDrm]

9 minutes ago, akio123008 said:

Don't they do that already sometimes?

If you mean by coloring the new/title ones, yes. But I mean flashing or using arrows to point out which part of the graph is being talked about. I don't remember LTT doing that. Could be that they are, I rarely "watch" the videos, usually they are on the background.

[minibois]

13 minutes ago, akio123008 said:

What I would do to make "lower is better" graphs look different from "higer is better" graphs, is to rotate the latter 90 degrees. I could try and explain, but here's an image:

I thought about that yesterday too and I think it's a good solution, except for the fact that they often have quite a bit of info to put in the charts, which may not always fit on the bottom of a vertical histogram.

[mariushm]

My nitpicks with such graphs ...

 

I'd rather have both M1 devices at the top ... why is that Intel mac mini between the m.1  devices ?

 

Would be useful to overlay on those bars the percentage, there's plenty of room ...  just say 100% or 1 to the lowest number,  so you can see right away  80% or 0.8 means 20% slower... or something like that.

 

 

Don't see the point of being so accurate and say there 27m05s... just say 27 minutes, nobody really cares that much and who complains can f... off.  5s is less than 0.3% of total time won't change the correctness of the chart

For example, 8:17 encoding time means 3.269 fps .... if you say 1620 seconds (27m), the result would be 3.259 ... so literally 0.01 fps difference.

 

 

 

[WereCatf]

39 minutes ago, akio123008 said:

What I would do to make "lower is better" graphs look different from "higer is better" graphs, is to rotate the latter 90 degrees.

Eh, I'd personally flip the metrics around: instead of "lower is better", I'd use "distance from the worst performer" -- if you had, say, values of 10ms, 20ms and 30ms to put on the graph and lower value was better, 30ms would be the worst value. The distance for 30ms from the performer would 0, thus a small or no bar. The second one would be 20ms, which would have a distance of 10ms, thus being a bar halfway the width of the graph. Then the last would be 10ms, ie. 20ms distance and the biggest bar -- POOF, you have just displayed the same data, but with bigger bars meaning better performance!

 

No need to change orientation of the graphs or use fancy colouring or anything like that, the result would be immediately understandable.

 

I don't have a clue how to use Libreoffice Calc, so you'll have to excuse the bad-looking graphs, but they do display my point anyway. Someone who actually knows what they're doing could make them look far better.

 

Direct comparison:

 

Relative to the worst performer:

[akio123008]

1 hour ago, WereCatf said:

POOF, you have just displayed the same data, but with bigger bars meaning better performance!

Yes but you have removed the linearity of it. the best performing product completed the task 3x faster, but in your new graph it's infinitely better.

 

Here's another example:

 

Imagine this with the worst product (A) being 130ms, the best one (B) being 20ms and some random one in the middle of the range (C) 95ms.

 

Now product B is objectively 95/20=4.75 times faster than product C. No doubt about that.

 

However, if you convert it with this "distance to the worst" concept, you'd get a value of 130 - 95 = 35ms for product C and a value of 130 - 20 = 110ms for product B. This means that the ratio now says that product B is 110/35 = 3.14 times faster than product C, which isn't correct.

 

Meaning that, yes you can see that product B is better, but the magnitude of this improvement is distorted, kind of ruining the purpose of a bar chart in the first place; if we only want to see what's better we may as well just look at the number and forget the scaled bars.

 

However, your point is valid, you can convert "lower is better" into a "higher is better" graph, you just used the wrong maths: what you need to do is to divide one by the value.

 

Another example: say we have a benchmark for encoding time, let's call the time that comes out T.

Now instead of displaying T we can display 1/T, giving us a higher value when T is lower, and maintaining the correct scale.

 

side note, you may know that 1/T = f in physics, with f being frequency, so we're effectively converting "time per encoding job" into "encoding jobs per second", which is equivalent

 

[akio123008]

59 minutes ago, LogicalDrm said:

But I mean flashing or using arrows to point out which part of the graph is being talked about. I don't remember LTT doing that. Could be that they are, I rarely "watch" the videos, usually they are on the background.

I don't watch them closely that much either, but I think at some point I have seen them circle some product in the graph that was being talked about. I can't point to an exact video where that's the case though.

[WereCatf]

41 minutes ago, akio123008 said:

However, your point is valid, you can convert "lower is better" into a "higer is better" graph, you just used the wrong maths: what you need to do is to divide one by the value.

 

 

Another example: say we have a benchmark for encoding time, let's call the time that comes out T.

Now instead of displaying T we can display 1/T, giving us a higher value when T is lower, and maintaining the correct scale.

Something akin to this, then? (No, I still don't know how to use this damn thing!)

[akio123008]

4 minutes ago, WereCatf said:

Something akin to this, then? (No, I still don't know how to use this damn thing!)

<image>

The use of a percentage is actually quite useful as well, given a frequency value wouldn't make much sense to people. Especially with something like reaction time which you used as an example here.

[WereCatf]

3 minutes ago, akio123008 said:

The use of a percentage is actually quite useful as well, given a frequency value wouldn't make much sense to people. Especially with something like reaction time which you used as an example here.

Yeah, I figured that'd be more easily digestible to most people.

[Moonzy]

I dont like relative graph because sometimes i wanna know the actual value (ie, if 2060 is 200% better than 1060 at 4k, but still runs at 30 fps, i wanna know it's 30 fps)

 

I quite like gamer nexus graph, best at top, and highlight the reviewed product + relevant competition

 

[Tristerin]

16 hours ago, piratemonkey said:

hopefully you aren't quickly going through a review to see if you should buy a couple hundred dollar thing. 

This.

 

While I understand OP's request, as it has a hair of merit in terms of people with OCD (not me), watch, digest and actually understand the info you are looking for.  Dont just assume bar graph = great or you will have more than just these problems on hand lol