This version is a big improvement, thanks @MrAlex. I particularly like the percentile, that is useful information. I suggest two tweaks:
add 1 decimal place to percentile
put percentile above highest rank.
Would it be possible to add some comparative data to the game info box? I’m referring to this:
It would be great if the tally of victories in games between the two players could be added here, for instance:
Yi Zheng vs alemitrani
3 previous victories 0
This is information that can be found on the website but it is awkward to find and often involves much clicking and scrolling. If the app could offer it as mentioned above it would be a big plus.
The comparative statistics section could also include rank, highest rank and percentile, to give a clear and comprehensive understanding of who is who!
Another thing I noticed about the update is that the rating history graph has been replaced by what looks like a fitted curve:
I think there is a bug when calculating the score after a game against the AI.
According to which rule is the score calculated, when playing against the AI?
In my last few games against the AI I was under the impression that the score was calculated as territory + live stones + captured stones, which results in very weird scores.
Japanese rules count territory + captured stones, Chinese rules count territory + live stones. Counting live stones AND captured stones results in weirdly-high scores and give more weight to capturing than normal go.
Screenshot of a game against the ai. The scores are way too high.
Black
Territory: 25
Captures: 2
Live stones: 17
Total Japanese: 27
Total Chinese: 42
Weird total: 44
White
Territory: 19
Captures: 5
Live stones: 20
Komi: 5.5
Total Japanese: 29.5
Total Chinese: 44.5
Weird total: 49.5
In particular, with Chinese rules the sum of Black and White ‘s score should always be equal to the area of the board, plus komi, minus unplayed dame.
Here the area of the board is 9x9 = 81 and the komi is 5.5, so the sum of Black and White’ s scores should not be more than 86.5.
With Japanese rules the scores should typically be much, much smaller than with Chinese rules, unless there was a supernatural amount of ko fights and under-the-stones play.
Would love to do that, but I don’t have the data. I used the data from the forum post about percentiles, and it was only given in full percentile increments. I mean I could interpolate between those markers, but I think that would be too imprecise. If I can get more detailed data, I’ll happily update. FYI, this is the link to the spreadsheet where I crunched the number (red line is OGS, blue line is Lichess): Copy of OGS v Lichess - Google Sheets
Yep, it’s a planned feature, but due to the way the server data is structure that requires an extra call to the backend. I’ll get around to do it at some point.
Yeah, I noticed I’ve broken it, but when I went to fix it I went down the rabbit hole and ended up rewriting it completely. The new version will allow you to switch between a few time intervals (e.g. 3 months, one year etc) and also see the exact rank at a particular point in time. I have a few things to iron out before I release it though.
Thanks for bringing me to my attention, I think you are right, there is something weird there. When I first implemented that I did not have a good understanding of the different scoring system and I think I made a dog’s breakfast of it. Nowadays, mostly due to teapoweredrobot’s guidance, I think I understand slightly more so I may revisit this. I think the bot insists on Chinese scoring, but I forgot I shouldn’t be counting captures in this case.
Thanks for your reply @MrAlex, glad to hear the tally feature is planned and that the graph will be fixed.
I see. That’s problematic because the real percentiles will change continuously as the population playing Go on OGS evolves. To be implemented well this feature would need to refresh the underlying data at regular intervals. If this can’t be done it might be better to stick to rank / rating and let people work out their own percentile if they want to. At least, the app should mention the source of the data used and the date of the last time it was uodated.
I don’t have the app, but it was mentioned to me from a beginner that it might be nice to be able to replay moves in the learn to play section to see why a move worked or didn’t work.
Is there any playback options in that part?
(Suppose you just trial and error a solution and then want to figure out why you got it right)
Not sure if it’s been mentioned, but I seem to have no ability to scroll – ironically this really only comes up when there’s a new release when there’s a list of bugfixes and such, which gets cut off on the bottom of the screen.
I’ve seen percentile mentioned as a feature, but I don’t see them on the stats page? Perhaps this is also locked behind a scroll?
Nope, no replay I’m afraid. I need to improve on that. They can replay the whole section though…
Ugh… can I please ask what device are you using so that the dialog does not fit on the screen?
That would be impossible to do in the app without the stats being computed on the server beforehand. I’ve asked Anoek and Benjito if he was inclined to implement such a thing, but the reply has been along the line “yeah, that would be nice to have, but we don’t have the development capacity for it now”. FYI, this is what lichess has Weekly Rapid rating distribution • lichess.org
New version coming soon. This time I went down the rabbit hole with the chart in the stats: I broke it by mistake last release, but instead of spending 5 minutes to fix it, I spent 5 days rewriting it almost from scratch to give it a more modern look Hope you guys like the result!
New chart (do remember to tap on it!)
Fixed AI scoring bug (thanks @ArsenLapin1 for pointing it out). The AI will now use Japanese scoring rules (I chose Japanese because automatch uses Japanese and I don’t want newbies to get confused).
It’s a Pixel 2 XL. I’ll try to grab screenshot of the next update.
A new one also: it seems when attempting to place an illegal move in analysis, 1) the crosshairs appear in the analysis screen, and 2) that’ll become the selected move upon exiting. Dangerous since the ‘submit’ and ‘exit analysis’ buttons are in the same place; so you’d misplace a stone if you accidentally double tap on ‘exit’
The updated chart is very nice, I like the way it gives you timescale options and details for specific points in time. Thanks for making these improvements @MrAlex.
What if I said that the app often crashes again?
My graphs seem to be wrong, e.g. 5Y seems to consider only 6 games (and +42 ELSO since 16th may 2022). “All” starts on 8th of March 2022. 1Y on the 11th of May, but the graph is different from the previous one.
Lol… god knows what’s going on in there… Trouble is, in order to calculate those stats I have to fetch all ranked games and do a bit of calculations on them… something in there it fails due to the sheer size of the dataset. Not sure how to approach it.
Glad you liked it.
That’s something that would need fixing… what kind of invalid move are we talking about? I don’t think KOs are enforced in analysis and playing on top of other stones should not be possible. Suicide moves then?
It’s a bit taunting to have this game always show at the top of the list, even though it’s the least active of my games, since my opponent has been away for some time.
What is the exact order used to sort the games? Would it be possible to move games where the opponent is in holidays to the bottom of the list?
Hope you guys like the result!
