Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[WanderingWinder]

Someone stuck in a line with 'to be filled in later' yesterday, fairly early in the day US, and never filled it in. I'm going to assume that it was forgotten and there's nothing doing now, but if anyone wants to have that game included, give a shout.

[SCSN]

That was me. It was a game against Lespeutere, he won, but I don't remember the seating nor the ending condition.

I was going to fetch the log, but they weren't working yesterday and still aren't.

[lespeutere]

That was me. It was a game against Lespeutere, he won, but I don't remember the seating nor the ending condition.

I don't know which seat I won from, either, I'm pretty sure it was on provinces (condition 2), though.

[WanderingWinder]

I actually had the condition 2 in. I'm going to stick lespeutere in the first players' seat and re-run the ratings. Should be re-updated in a couple minutes.

[Watno]

BTW, would it be possible to show the ratings from previous days somewehere as wll, so we can look at the developments as well?

did you miss this?

[SCSN]

BTW, would it be possible to show the ratings from previous days somewehere as wll, so we can look at the developments as well?

did you miss this?

They are here!

[WanderingWinder]

I'd still enjoy seeing the W-L-D record in the table. I know you can "estimate" how many games someone played based on their uncertainty, but I have no basis on which to understand about how many games people have played based on their uncertainty.

I understand and acknowledge this, but I'm not planning on taking the time to do this any time soon.

BTW, would it be possible to show the ratings from previous days somewehere as wll, so we can look at the developments as well?

did you miss this?

No. Here, I don't really have a problem with doing this, and I have the (raw) ratings from every day saved, but there are three problems: First, they aren't pretty at all, and actually not so easy to read. Second, where am I going to post all these? Third, the time it takes.

Overall, I'm just not sure how much I am going to keep this stuff up. Goko's system seems to have gotten better (I still have problems with it, but it's seemingly not *so* far off what you normally get anymore; well, anyway it's still bad, but not terrible), and more important, the ten-ish minutes it takes me to do this every day are more taxing to my schedule than I would have expected. I'm going to still try and do it for now, but we'll see.

There are two things I am still really interested in doing - one is finding the correct underlying curve for Win Expectancy in Dominion, or at least getting a pretty good approximation thereof. The second is quantifying first player advantage. The former I could probably actually do best off of a month (one not soon after a set release, because that is going to make bigger skill changes, which we don't want for this) of old iso data, now that I think about it, if I could get that in a reasonable format (probably a CSV is easiest, the only concern I have is with people's names, but I assume there should be a workaround for this). The first-player thing is going to be a lot harder on iso data. I would most prefer to do it after we have the other, but even what I have off of the goko data now, from this project, should be enough to get at least a start, whenever I can find the time to sit down and grind through it.

[rrenaud]

Automate.

If things don't break for some unforeseen reason, you should have to spend 0 minutes per day updating the rankings.

[WanderingWinder]

Automate.

If things don't break for some unforeseen reason, you should have to spend 0 minutes per day updating the rankings.

It IS automated. However, I still have to copy/paste/save a handful of files and lists, create a new spreadsheet, lock the old spreadsheet, run the program... all of which ends up taking, oh, about ten minutes every day.

[rrenaud]

That's not automated, unless you've cloned yourself ;P

[WanderingWinder]

That's not automated, unless you've cloned yourself ;P

Well, the actual rating of games is automated. I don't want to have to learn how to automate the process of selecting parts of files off of a piece of a file on the internet, copying a file into a part of a post here, copying part of a post here into part of another post here.... besides which I almost certainly won't be running these long enough to where all those ten minutes would add up to enough time to offset this, nor do I think it's even possible based on the copy-ability of things.

[eHalcyon]

Possibly you could recruit someone from the forums to be a code monkey. Volunteers?

[WanderingWinder]

Possibly you could recruit someone from the forums to be a code monkey. Volunteers?

Not gonna happen.

[eHalcyon]

Possibly you could recruit someone from the forums to be a code monkey. Volunteers?

Not gonna happen.

Why not?  If it's a worry about keeping your algorithm secret, well, you said you already have the actual calculations automated and it's just grabbing the results that you are doing manually.  Someone could code that part for you without needing any sensitive information -- just a script that grabs the relevant files from Goko, mines the information and outputs a document that is easier for you to pop into your spreadsheet.

[WanderingWinder]

Possibly you could recruit someone from the forums to be a code monkey. Volunteers?

Not gonna happen.

Why not?  If it's a worry about keeping your algorithm secret, well, you said you already have the actual calculations automated and it's just grabbing the results that you are doing manually.  Someone could code that part for you without needing any sensitive information -- just a script that grabs the relevant files from Goko, mines the information and outputs a document that is easier for you to pop into your spreadsheet.

Everything you describe them being able to do here literally saves me nothing. Well, I suppose I could get them to type the proper name of a file so that I could just download and save it rather than copy/paste/name/save, but that's about it. And let me tell you, those five seconds for me are not worth it.

[ragingduckd]

Goko's system seems to have gotten better (I still have problems with it, but it's seemingly not *so* far off what you normally get anymore; well, anyway it's still bad, but not terrible), and more important, the ten-ish minutes it takes me to do this every day are more taxing to my schedule than I would have expected.

I think there's something wrong with my database.  It's telling me that you posted this 16 hours after hitting #1 for the first time, after averaging 14 games per day in the previous week.

But, I mean... that would have to be a joke, right? 

[WanderingWinder]

Goko's system seems to have gotten better (I still have problems with it, but it's seemingly not *so* far off what you normally get anymore; well, anyway it's still bad, but not terrible), and more important, the ten-ish minutes it takes me to do this every day are more taxing to my schedule than I would have expected.

I think there's something wrong with my database.  It's telling me that you posted this 16 hours after hitting #1 for the first time, after averaging 14 games per day in the previous week.

But, I mean... that would have to be a joke, right? 

If you want to say I'm being self-serving, just come out and say it, man.

That I hit number one for the first time isn't relevant. That I've been playing more probably is. And actually, that I've been playing a lot probably is probably *why* I hit number one, as well - probably my biggest gripe with the system still is that it has the Skill-some multiple of uncertainty thing, which when I play so much, is going to vault me up there. But playing more gets me a better feel for the system. And it's still flawed beyond this - it's too easy to get where, like right now, my rating isn't changing nearly enough with every game, I still see some 5000s who are quite bad whilst most aren't, and I still think that a 2000 point swing, based on what I've seen from the play levels, shouldn't mean an expected win rate of the ~5:1 that it seems to be. I still think it is a bad system, in short, but not such an unusually terrible one as it was.

Now, I don't know what database you're sitting on (if you have such a DB, why not share some of this information with us?), but I presume if it has all that, you can also see that I've been running exceptionally good over the past couple days, and so it shouldn't really be seen as ego-centric if I feel that my results, objectively, should have shot up at least near the top. I certainly can't keep this pace up, as I've been getting lucky, but based on what the system can tell, I ought to be pretty high up.

But the biggest thing is to not do this thing backhandedly - I can take accusations to the face.

[ragingduckd]

I wasn't trying to be opaque.  It's either pretty douchy or just a remarkable faux-pas.

My database is just a scraping of the top-100 leaderboard for the past couple weeks.  I'm planning to match the rating changes against the game results to see what I can deduce about the Goko rating system.  I'm happy to share access to the database if anybody wants it.

[WanderingWinder]

I wasn't trying to be opaque.  It's either pretty douchy or just a remarkable faux-pas.

I don't know where you go off being so insulting, dude. What did I ever do to you? You're being entirely illogical here - I don't even understand what your point is supposed to be - I made a rating system because I was unhappy with how I was being ranked (at the time, I had been pretty firmly in second behind astrosity (or however he spelled it) for quite a decent spell, and everyone knew that he was exploiting the no-penalty-to-leave glitch, so I actually HAD effectively been number one)? This was supposed to stroke my ego... how? And I quit once I got to the top of Goko because... why? Wouldn't I have made some kind of deal of me topping my own rankings for the one day that happened? Or of topping the Goko ones? I mean, yeah, I'm not emotionless, and I am competitive, and I *do* like to do well, I don't deny any of that. But from the beginning, I have always said that it would be far preferable to have the actual site's rankings be good than to have a separate list. There are loads of reasons for this.

Or are you saying that I think the rating are more accurate because I'm doing well on them? Well, gee, I won something like 15 or 20 games in a row, which is absurdly lucky, I grant, but you know, if you did that, I would say you'd better be up at number one following that as well - this is a pretty clear indication of quality play, and if you think this is arrogant of me to say, well I don't know what to tell you. I know I can't maintain that - and I haven't, already. The big point of my increased satisfaction with the ratings isn't whether I'm going up or down - it is by *how much*. By tracking this, I can calculate (albeit only roughly) what win% the system expects me to have from a given rating differential. Having played 15000ish games, I have a pretty decent feel of how often I should be expecting to win against certain opponents. Certainly the system doesn't match up exactly, but then no system - including my own - entirely has. I still think Goko's has a fairly exploitable bias, worse than it could. But it's a bit hard to tell how much of the rating is the skill estimate, and how much is the uncertainty, and this is really annoying.

The biggest point I'm trying to make here, though, is that you still haven't answered my question: What is your point? Because if it's just to say that I'm being 'pretty douchy', this isn't a point so much as an insult. Maybe I should rephrase it: What positive contribution did you expect these posts of yours to make? How would they improve... anything?

My database is just a scraping of the top-100 leaderboard for the past couple weeks.  I'm planning to match the rating changes against the game results to see what I can deduce about the Goko rating system.  I'm happy to share access to the database if anybody wants it.

What this says to me is name and rating - that's what is actually on the leaderboard. However, if you can actually make the statement from your previous post on data, you must have more than that - otherwise, you couldn't know how many games I've played. Also, how often are you doing the scraping? I mean, I doubt it could be continuous, yet you supposedly have to-the-hour precision.

[Fabian]

rd,

Welcome to the forum!

[ragingduckd]

That's fair, WW.  My post wasn't constructive at all.  Here's a more useful (and civil) version:

You should keep your system going.  Even if strong players are now drifting to the top of the leaderboard, Goko's system still has a lot of problems.  You already have 21 people -- most of them top players -- who feel strongly enough about this to manaully track and send you their game results.  I think you have some obligation to these folks for helping you get started.  There are also many others (including me) who would be interested if they didn't have to submit their results manually.

If updating your ratings every day is a strain, then update them once a week.  Better still, take rrenaud's advice and automate the data entry.  One easy solution would be to have people mail in their formatted results.  Someone like rrenaud or nutki could easily write a script to aggregate that data and forward it to you.

An even better solution would be a fully automated system that reads the Goko logs and updates a website like councilroom.  That's a bigger project, but this forum is full of people with strong technical skills and not enough to do.  There must be someone you could trust to help without stealing your intellectual property.

I'm actually working on automated log parser myself.  I plan to it connect to a much more modest version of councilroom -- or to councilroom itself if rrenaud et al are going to keep it running and could integrate my code.  I could probably provide the data you need without too much more work.  Incidentally, my current database is just an scrape of Goko's top 100 ratings at 60-second intervals, enough to identify games (as rating changes) but not to provide the data you need.

If you're willing to keep your system running, and if getting the game data more easily would make the difference, I'd be happy to help.  If you'd rather work with someone else, there are a handful of others on the forums who appear to be working on similar projects.  Serakfalcon already has his on github: https://github.com/serakfalcon/DominionCoding/

[Watno]

There are people trying to get Councilroom running again, and they would probably be glad for help. See here: http://forum.dominionstrategy.com/index.php?topic=4185.msg92606#msg92606
It would be really awesome to have Councilroom running again.

[ftl]

I'm working on it. Goko logs are different than iso logs and some information which was obvious in iso is less so on goko, so log parsing is taking a while to get working.

[WanderingWinder]

So here's the deal. I am not going to maintain this (well, big surprise, I've already lapsed). What you (AI/rd) can't see is that there definitely aren't 21 people using this. Well, there weren't, more precisely. With much regularity, there were like 4-5. I think it's tedious for most people. It's certainly tedious for me. Sure, int he perfect world where I could just totally automate everything, it wouldn't be an issue at all. But that is not something that I'm going to be able to do anytime soon. Indeed, it's all very nice for you to say that I should continue this, but you aren't the one who is bearing any of the costs. Actually you aren't bearing any of the reward, either, from what I can see, so I don't really get what your point is, but whatever. This is a free service that I was providing entirely voluntarily, and I basically thus don't need any reason to stop providing it. It wasn't actually worth all that much anyway, as a rating pool largely based on so few players is pretty miserable. By far the best thing about it was just getting the data, but you've gone a long way in doing this, but on a bigger scale, which in large part obsoletes this.

The point is that it is tedious for me, and while you can say 'do it more efficiently', this doesn't make me able to do so.

Anyway, I am getting busier in RL, and this is just a very easy cut to make.


I would still like to work out what a 'fair' curve for Win Expectancy should be for the game, and maybe I will pour some more time into this at some point, but it probably won't be until August, at least, and more likely December or later.

[ragingduckd]

What you (AI/rd) can't see is that there definitely aren't 21 people using this. Well, there weren't, more precisely. With much regularity, there were like 4-5.

You're quite right.  I had no idea, and I wouldn't have pushed you if I'd known.  My apologies.