I couldn't find whether this was addressed before, but is there a way to filter people based on Isotropish level? (Something like L30+ instead of 5000+)
Not yet, but probably soon.
For me this gets dangerously close to forcing people not using external tools to start doing this. Someone who doesn't use Salvager and doesn't visit gokosalvager.com cannot even check his Isotropish level. Goko Pro level can be easily checked from within the client.
Ok, gonna rant here. And I'm only speaking for myself, not for michaeljb or the other contributors to Salvager:
I'm somewhat sympathetic to the troglodytes who still don't use the extension. The extension works pretty hard a little bit at being friendly to them, with features like #vp?, auto-automatch, and auto-kick explanation messages (currently beta). But there are limits to how much I'll deny to people who use my software for the benefit of those who don't.
If I implement kick-by-iso, the auto-kick messages will refer people to isotropish.com. Isotropish ratings change far slower than Goko ratings, so it's not like they'd have to be constantly checking either.
Personally, I really want this feature because seeing Pro and Isotropish ratings side by side keeps reminding me of what a joke Goko's Pro system is. I really hate putting in 5000+ and getting a level 20 opponent. That's just not going to be a fun game. But do I kick the guy? It seems like a dick move, since he satisfies the criterion I listed.
For mid- and low-range players, the problem is far, far worse. Just look at this godawful mess:
Majacian is thinking he's a little better than (V)_*,,,*_(V) because he's got a higher Pro rating. (V)_*,,,*_(V) is probably getting kicked from some games that Majacian can enter. Meanwhile, (V)_*,,,*_(V) is actually 32 levels higher than Majacian. That's like Banker Bot thinking it's a little better than Stef.
The problem gets worse the lower you go because Goko's worst tweaks to TrueSkill have their largest effect on low-rated players. But even near the top, the volatility is a huge problem. I briefly overtook SCSN on Goko Pro after a lucky run yesterday... Isotropish still had him as several levels above me, where he belongs.
And before you ask, yes, Isotropish is right and Goko Pro is wrong. Absolutely no doubt. Someone else can collect the data and analyze the win prediction accuracy of both systems... I'm not even going to bother.
I also agree with soulnet that it's really the user's choice. If you think it's rude, don't use it. Call me an arms dealer if you must... I'm big on personal liberty.
I assume you won't mind a counter-rant.
You have NO evidence that Isotropish is a better indicator of skill than the Dominion Online (hereafter DO, aka Goko/MF, whatever) pro leaderboard.
To be fair, you can't actually get good evidence, since you can't calculate predictions off of the DO ratings, since they don't put out the information necessary. But to still be fair, you don't bother to do this either, and are explicitly too lazy to here, so it's not a particularly strong point in isotropish's favor, either. A LOT of what you describe in the page you link to isn't actually issues with the rating system, but issues with just other bugs with their software (e.g. it gets the winning player wrong, the 0 difference on some games, etc.). All your examples here, you don't actually have any evidence that your system is right in comparison - you say things like "person X thinks he's better, but he's much worse", except you don't accept this without assuming your conclusion that your system is better to start with.
Is TrueSkill a good system for Dominion? Well, maybe. It's a fine thing to try at least. But you know, that doesn't mean that THIS IMPLEMENTATION is any good. It seems to me that this implementation is actually pretty crummy - it moves way, way, way too slow. And this jives with some people's thoughts, because they just 'know' that player X is better than Y, and this conventional wisdom, which was true at some point, continues to be propagated, even if there's good evidence against it from recent results, because your system moves incredibly slowly. I'm pretty sure your parameters are just set to be far too stodgy. How can this be? Well, you don't actually have any data analyzing the predictions to tell you what good parameters are. But beyond this, why are your parameters set to what they are? Well, it's based on perceptions from the iso days. And the perceptions, which could have been wrong to start with, were based on a day-to-day increase in uncertainty level, which you got rid of. So.... these parameters are significantly stodgier than it was, even if those were correct. And I tend to think (though admittedly, this is just my gut feeling), that they were a little bit too stodgy to start with.
But the biggest issue I have is definitely that you're still going off of the 'level' thing, and not just rating rating. Since you're cutting off inactive players, I really really don't think this is needed.
So ok, it's obviously not that I think that the DO ratings are particularly good, and I definitely don't have any evidence other than my anecdotal feeling that they're better than yours. But for Pete's sake, you have zero evidence the other way either...
Author
Topic: Goko Dominion Salvager Discussion (Read 603624 times)