Nicely done! :D Interesting how the Sentinels always die.
So is this all going to be generated from the digital game, then? Nothing for us to input? Mostly because I don't want to have to write down every little thing that happens in an Oblivaeon game. :B
Yeah, Lynkfox and I couldn't come up with a reasonable way of adding it into the existing Google form, plus there would probably never been enough data to make analysis useful. There have already been 20 times as many games with the OblivAeon heroes in digital as in physical, just half a month after they came out. Even as it is I don't have the greatest confidence in the reliability.
Says something that the Primary Objective is played more then all other shield combined
It’s the easiest shield, and the mode is already difficult?
Hm, TakeWalker actually seems to have caught a bug. VGBM posted a screenshot in Discord that shows him winning a game with The Sentinels still active--although the screen was bugged because they'd replaced the Adamant Sentinels and was showing the HP of all eight of them. It's possible it's the same bug, though.
When will the stats for these new heroes and environments show up for the other game play modes? I’m not trying to press you on getting that done immediately as I’m sure what you put together for OblivAeon mode was quite the undertaking. Just curious moreso to see those other stats.
They're already there.
I don’t find them going to the lite stats page https://mindwanderer.net/sotm/app-stats-lite.html
Really interesting! Do the win rates at the bottom count unfinished games as losses or is that win rate the win rate of concluded games?
Especially interesting to see that only roughly 1 in 10 ObliVeon games are played to conclusion- I suppose that isn’t too surprising- the writing is usually on the wall long before you’ve actually lost and it’s a long game if you are fairly certain of defeat.
I’ve heard this before. Most of the games we’ve played, we believed we were going to lose. However, we still haven’t lost since it was released. It’s pretty amazing how fast the heroes can recover if they fight hard enough. ;-)
(Mind you, I get it. For me, I don’t find OblivAeon as enjoyable of a game because it’s torturously painful to play through. 
I’ll stick with traditional or even team games for the most part!)
Argeed.
I've only won 2 out of about 10 Oblivaeon digital games, but in both of the wins there were several moments when I thought I was FOR SURE going to lose, but played them out anyway. And then I won!
Do you have something cached maybe? Because they’re there.
That must be it as if I use a private browser session they do show up. I’ve not had any kind of caching issue with that page before and the full stats page had them which is why this struck me as odd.
I’ve found this to be true in Sentinels from the beginning. This is really a game where it’s not over 'till it’s over!
Oh, definitely- it’s very rare for me to quit a game even if it’s looking bad, as they can be some of the more interesting games because you’re on the edge of losing. I would imagine though due to the more ‘instant’ nature of digital play that there are definitely people who just rerack the game if it’s too much of a slog, and I guess it isn’t too shocking that OblivAeon mode increases that rate - it’s definitely a 3 course dinner rather than a snack, and time is precious. I’ve been enjoying the mode, but I’ve still only played it 4 times since release (2 wins, 2 losses) despite playing multiple digital solo games a day since then - it’s a commitment!
Hi MindWanderer: I was looking at the "Details" pages recently, and I was surprised by the number of hero x villain combinations where there was a chi-square p-value of less than 0.05.
I recognize it would not be easy to do, but I wonder if it would be possible to run the model once with the first order interaction terms? I ask because with a lot more data than you may have had at first, some of those interaction terms may now have a significant relationship to the overall difficulty. This would mean that the estimated difficulty in your calculator is over/under-estimating in some circumstances.
Or maybe not. The chi-square test you are using on the deviations on the details page seems very sensitive; there are many cases where there are just a handful of games but a significant p-value.
Anyway, as usual thanks for all your work, and I understand if what I am asking for is impossible. But I would love to see the co-efficients and p-values from those interaction terms at least once.
I'd love to do that, but unfortunately I don't have the computing power. On the user-submitted data, the program takes over half an hour to run, most of which is reformatting the JSON data I get into dummy-coded dataframes. On the video game data, when I initially ran the full data set, it took over 8 hours. Fortunately, now I have that, so I just have to add a month's worth at a time, but recoding everything to add the interaction terms would take the program... what, probably a month to run at that point? And a dataframe that large wouldn't even fit in memory, even if it were mostly zeros. It probably could be done, but not using anything like my current method, and I wouldn't know how.
Good grief! I withdraw the request.
Sounds like a highly parallelizable process. Would there be any benefit to breaking the task into smaller subtasks that could be executed by forum members?
The recoding probably could be, but that wouldn’t help with the issue of the dataframe just being too big to work with without setting up a Hadoop cluster or something, which is way beyond my level of expertise and resources. Even if it were, it would take a lot of work and organization, and frankly, to me, it’s not worth the amount of time it would take. Plus it would require distributing Handelabra’s raw data to a bunch of people I don’t know, and I don’t have permission to do that.