Honestly, this “way of looking at it” just never occurred to me.
I never think “gee, I wonder what the ratio of my weak upsets to my strong upsets is?”.
I don’t see how that’s meaningful.
And … if there were a chorus of people saying “that’s how it should be” then sure.
But actually, there’s one person arguing for it, one person who said they like it how it is.
Not much to go on…
GaJ
“I never think “gee, I wonder what the ratio of my weak upsets to my strong upsets is?”. I don’t see how that’s meaningful.”
Me either, and I have no clue what that sentence even means… How can an upset be either strong or weak?
Again, here’s my rationale worded as simply as I’m able, because the way you phrased it still makes me think like there’s some miscommunication going on here:
I think it makes sense that stronger shades of colour should be designated for surprising (“upsetting”) game outcomes, whereas weaker shades of colour should be for unsurprising (“expected”) outcomes.
I agree. For me, the most interesting figures are the wins against stronger players and the losses against weaker players.
Ah hah. At last we get to the bottom of it.
Here’s source of the disconnect.
A pie chart is for visualising ratios.
Let me show you what I see when I look at the pie chart on my page:
… at a glance I can see that my next game should probably be vs someone weaker than me, and that I can probably expect my rank to still be going up because that will take my win loss ratio over 50.
So this is why they are the way that they are.
The whole motivation for the pie chart was to let people monitor these ratios and choose what kind of game to play next with this in mind if they want to.
I also don’t find it challenge to say:
“The (red) losses against weaker (weak colours) is small”.
I’m telling you all this just to show you what was in the mind of the developer when they did it this way, what the intended result is supposed to be.
And to show you what would be lost if it were to be changed.
I acknowledge you may say “I don’t care, I don’t actually see it that way and it never occurred to me to wonder about my ratio of games vs stronger to games vs weaker players - I don’t see any point on keeping an eye on that”.
At which point we’d be debating which feature we want knowing what each other cares about 
GaJ
I think there might still be some miscommunication here…
As (I think) I’ve been saying this whole time, I get that this is why you like the pie chart set up the way that it is - and I’m not asking you to change it! If you swap the shades of red without changing the arrangement of the pie chart, then
You won’t lose any benefit, while the order as I knew it might be restored…
Side note that obviously in the bigger scheme of things, this isn’t a big deal (cough why can no one explain how OGS-modded Glicko 2 works… cough byo-yomi period display is still messed up… cough one-off site donations… cough). As much as anything, I just like the discussion and trying to get to a mutual understanding of the issue
Yes, I totally value the fact that you are carefully explaining and working towards mutual understanding - it’s been awesome. (And probably would have been dealt with in 10 iminutes over a beer 
) Kinda rare to be able to have this sort of conversation in a forum
I’m sorry its taking me ages to see what you’re trying to get at. 
And who knows - if I finally appreciate it properly, an improvement might come that can make everyone happier, and that I can actually affect because it’s client side 
Are you able to hack the pictures to show what you want? I can’t for the life of me see how it will work out that if the intensity of the shades does not represent the player’s strength, then how does the pie chart show the ration of games vs stronger to games vs weaker opponents? 
The middle diagram I posted shows how I visualise the ratio of games vs strong to games vs weaker - by looking at the ratio of weak colour to strong colour.
I should also say that I can see that this isn’t ideal for the bar chart. Full disclosure: I never find myself looking at that thing, I almost dropped it out competely when I was making the changes It doesn’t speak interesting information to me, but I know that this is “just me”.
GaJ
I don’t have a way to edit images right now, so I’ll have to try and explain with words again…
If you swap the shades of red without changing the arrangement of the pie chart, then
I guess you’re saying that swapping the shades of red would muddle the clarity of strong-vs-weak to the point where it’s not intuitive to recognize (despite the captions, mind you) that “games against stronger” are grouped on the right and “games against weaker” on the left? If so, wouldn’t it be possible to add a black marker line on the actual pie chart (with an explanation of what it means) as you did with your middle graphic?
Yes, that’s it.
In fact “the whole point” of carefully chosing the meaning of the colours and saturations was so that these two important ratios could be clearly seen without having to draw cluttery lines on the pie chart - the relevant data stands out directly based on the intuitive meaning of the colours (weak - game vs weak player, red - loss, green-win, strong - game vs strong player) 
The new information here is that users are saying “I would like unexpected results to stand out by saturation”.
I need three-dimensional colours now 
GaJ
Welp, at least we’ve come to a full mutual understanding at last…
I just feel like it’d still be relatively easy to recognize “strong-vs-weak” from the pie-chart lay-out (which does have captions to explain what signifies what) whereas “upset-vs-expected” can’t currently be easily gleaned (especially on the bar chart, from which the previous saturation code originated).
Shrugs
I see it. Let me think.
GaJ
To me it’s not about the ratio but more the colour labelling as these are the games that you’d most likely want to review. Won games against stronger opponents and lost games against weaker opponents are more likely than not going to have a higher percentage of teachable moments, and so while it might not be necessary to see how many of these games exist, it would be nice to be able to see quickly at a glance (possibly with some graded colouring in the history table?) which games are the standout ones and which are the more expected results that may have less useful information to glean from a review.
Sorry Lisa if this is the point where our agreement diverges, but these are my personal thoughts on the topic.
Nope, I absolutely share your thoughts, well put. I guess if standout games could be highlighted in your game history somehow, I’d accept that as a consolation prize. Although ofc I still feel like having them represented in the bar and pie charts would only have benefits, with “strong-vs-weak” still easily visible from the layout…
Yes. Highlighting the standout results in the history table is a good idea.
Well, it’s good that we are closing in on what the real feature that is needed is
There’s no real reason to see “unexpected” results in the pie chart (which is about ratios) or in the bar chart (which is about performance over time).
Actually, what’s needed is a massively overhauled game history, where we can sort and filter things like ranked, not ranked, teaching games, annulled games, unexpected results etc.
This has been on my own wish-list for a while, and I’ve thought about “doing something”, but I’m not sure how far it can go without server-side support…
GaJ
I haven’t participated here before due to laziness Also, I can’t easily draw pictures here.
My ideal pie chart would be like:
(saturated red / bordeaux) (saturated green)
games lost agaist weaker opponent games won agaist stronger opponent
(diluted red / bordeaux) (diluted green)
games lost agaist stronger opponent games won agaist weaker opponent
rationale:
Hope this helps. Again, just my opinion, but I do find the current color choice (and pie chart ordering) quite unintuitive.
This is basically a refinement of Sarah Lisa’s request, and it clearly is gaining some support.
The difficulty with it is that it focusses on the use of the pie chart for something that it is not intended (or suitable) for: finding upset games.
In turn, the reason for this is that there isn’t a good tool for finding these things.
We definitely should be able to find upset games, for the reasons BHyden elaborated.
But the solution for finding upset games does not lie in messing up the pie chart from showing what it is supposed to show - specifically: “win/loss ratio”, and “games played vs weaker/vs stronger ratio”.
Those two ratios are very meaningful and connected to the topic of “how well is the rating system working” and “how good is my rating likely to be at the moment” IE they’re not just “random maybe interesting” ratios: there is a specific purpose for monitoring them (in fact, more than one).
I think the proper solution for finding upset games lies in the game history table.
It might turn out to be awkward if that solution calls for colours like the rating graph/pie chart, but it’s far from obvious that this will be the case. So I see that as a bridge to cross later.
The first bridge to cross is to start working on a solution for a better game history table, which is a rather daunting task, having looked at the code. Transforming that into something that is filterable and searchable and column selectable (which is what you’d want) and including the extra information, like win/loss type etc … that’s a project.
My feeling is that people who want a better took for finding upset games should look to that and work on that, rather than disrupting the current feature.
GaJ
Just a shame that the previous tool intended for representing upsets (along with the pie chart’s current intended use cases) has been disrupted (the bar chart)…
All snide aside, I would argue that
Hi GaJ,
I agree that my approach is very similar to what Sarah Lisa proposes and I understand that as the developer of this feature you should have the ultimate say on it. However I want to address one point in your previous answer. You said my proposal focuses on using the pie chart to find upset games, but that is not how I see it - my use of the pie chart (as well as the progress graph and table) is to answer the “how well am I doing” question.
It’s just, we seem to have different intuitions about answering that - to me, the upset games are the most interesting ones, because I can win against 25k players as often as I want and I can challenge sdk players as often as they’ll allow me but I have no expectation of winning these games. So, the games I won against stronger players, or lost against weaker players, end up being the interesting ones - “how am I doing” is the central question, “upset games” is just the angle I am answering that question from. To me the upset games are just as interesting in aggregate (to answer “how am I doing”) as they are as a per-game attribute (to identify games I should have another look at).
Well, it’s clear that there is this big multi-faceted question “How well am I doing?”
And it’s clear that there are many aspects to answering this.
One is to look at your rank value over time - the rating graph. At least that works properly
Another is to look at some interesting ratios. The pie-chart offers this.
Another is to look at the relative size of wins and losses that you have had over time - the bar charts do that.
Now to some problems.
“The pie chart doesn’t show (clearly) the ratio of unexpected games to … other ones”.
I still think this is a red herring. Even if we accept that “the amount of these in aggregate is interesting”, I still don’t think that there is a meaningful ratio. It’s only “are there lots or not?”. The problem isn’t that you can’t see the ratio that you want, the problem is that the colours aren’t making an aggregate stand out in the way you want.
Here’s the real problem: the bar chart no longer makes aggregations of unexpected games stand out.
That’s because the bar chart doesn’t use stand-out colours specifically for unexpected games anymore.
My guess is that if it did, then you would be happy having it glaring at you that “in Jun you had a blob (aggregate) of unexpected games” in saturated colours, and would be less insistent to see these “called out” in the pie chart.
Key challenges are finding a solution that works with:
For this reason, I’m not rushing to change it in the way you’re asking for. It’s not so much that as developer I have the ultimate say. Actually that’s not even true, because as owner and benevolent dictator anoek has ultimate say But I don’t want to assert any sort of ‘developer right’.
Rather, I’m asserting “interested in finding the right broad solution in the face of all the inputs” right
And I’m not sure we’re there yet.
GaJ
I haven’t read exactly everything in this thread.
I mostly use the pie chart to check my opponent’s “personality”, to see things like this person hates loosing so he only plays weaker opponents, this one likes playing stronger people (to improve?), and I am balanced.
This is why I like it the way it is, since it currently is perfect for this.