Mar 06, 2003 12:33
IBL Juice!
This is an old log from IBL ..... Its pretty funny which is why I'm reposting it.
[20:45] <Jolo> the most fair starting thing is really to put everybody at the same level
[20:45] <Sticks> but that didn't happen
[20:46] <Jolo> but that would prob add 3 or 4 weeks to the time it takes
[20:46] <Jolo> we'd barely have fair ratings after 10 wks
[20:46] <Jolo> this is why when you rank college teams, you do a pre-season ranking
[20:46] <Jolo> like for basketball or football
[20:46] <SuicidJky> what bout 5 points in order of signup
[20:46] <Jolo> that's pretty much the same thing too
[20:46] <Jolo> i intentionally spread them out to about 1/3 of the K value
[20:46] <Jolo> there's a reason for that
[20:47] <Jolo> if you play an equal guy and beat him, you gain 1.5 K points
[20:47] <Jolo> that's just how the formula works, no magic there
[20:47] <Jolo> err 0.5 K
[20:47] <Sticks> the magic K, mysterious
[20:47] <SuicidJky> 1.5 would be alot more jumping
[20:47] <Jolo> it's just a constant
[20:47] <Sticks> I know
[20:48] <Jolo> do you guys know about slope?
[20:48] <Sticks> but its unknown to us
[20:48] <Sticks> of a line or...
[20:48] <Jolo> right, but not a line
[20:48] <Jolo> it's nonlinear :)
[20:48] <Jolo> let me see if i can come up w/ an analogy
[20:48] <Jolo> u guys know the scales in grocery stores for produce?
[20:48] * FireIce knows about slope...
[20:48] <Sticks> yeah..
[20:49] <Jolo> like if you throw a watermelon in one, it bounces around?
[20:49] <Jolo> but eventually it stops bouncing and gives you the right weight
[20:49] <Jolo> K is kinda like how stiff the spring is
[20:49] <Jolo> if you have a loose spring, it will bounce all over the place
[20:49] <Jolo> make it hard to read
[20:49] <Jolo> and takes a long time to converge to the final weight
[20:49] <Sticks> loose = big K or small K?
[20:50] <Jolo> loose = big K
[20:50] <Jolo> (it's actually a really shitty analogy but i can't think of a better one right now)
[20:50] <Sticks> no, I get it
[20:50] <Jolo> so generally you don't want a really big K
[20:50] <Jolo> it looks like stuff is happening, very exciting
[20:50] <Jolo> but in reality you can't tell how fscking heavy the melon is
[20:51] <Jolo> to make things worse
[20:51] <Jolo> imagine if you had 30 scales
[20:51] <Jolo> all hooked to each other
[20:51] <Jolo> all bouncing
[20:51] <Jolo> that's what this system is
[20:51] <Jolo> so every time you play a game, you throw 4 melons in 4 of the scales
[20:52] <Jolo> everything bounces around and finally settles
[20:52] <Jolo> except then you throw 4 more melons in
[20:52] <Jolo> melons = players
[20:52] <Sticks> so then what would the starting points mean in this analogy?
[20:52] <Jolo> heh
[20:52] <Jolo> imagine if you could calibrate the scale a bit
[20:52] <Jolo> like for heavy melons vs. a single orange
[20:53] <Jolo> but pretend you're blind and have no idea wtf kinda fruit you're measuring
[20:53] <Jolo> your best guess is probably to have a single average calibration
[20:53] <Jolo> but then you'd throw a melon on and tilt the scale like crazy
[20:53] <Jolo> and the orange barely moves the needle
[20:53] <Jolo> eventually it'll all sort out
[20:53] <Jolo> but it takes longer to figure out
[20:53] <Jolo> but if you knew up front that you had melons vs. oranges
[20:54] <Jolo> you could use the heavy scale for the melons
[20:54] <Jolo> etc.
[20:54] <Sticks> but there will be someone getting shafted
[20:54] <Jolo> yup, that is always true
[20:54] <Sticks> like, using a heavy scale for an orange
[20:54] <Jolo> yup
[20:54] <Sticks> excellent...much clearer
[20:54] <Jolo> you'll pick up an orange that looks a melon
[20:54] <Jolo> has the color of a melon
[20:54] <Jolo> turns out it's not heavy at all
[20:54] <Jolo> you throw it in the heavy scale and it's all wrong
[20:54] <Jolo> but you'd notice that pretty quickly
[20:54] <Jolo> "wtf, that's no melon"
[20:55] <Jolo> now imagine if you had a pile of fruit sitting there
[20:55] <Jolo> that refuses to be weighed
[20:55] <Jolo> or maybe they get weighed once and then run away
[20:55] <Jolo> you'll never know for sure how heavy they are
[20:55] <Sticks> hehe
[20:55] * Min chuckles
[20:55] * Min saves this
[20:55] <Jolo> or maybe they'll get weighed wrong and appear real heavy
[20:55] <Sticks> running fruit
[20:55] <Jolo> but the thing is, who cares
[20:56] <Jolo> if you're just trying to make a fruit salad
[20:56] <Jolo> with the right mix of fruits
[20:56] <Sticks> they'll all end up dead
[20:56] <Jolo> so there's this orange that got mistaken for a melon once
[20:56] <Jolo> and after getting overweighted, it rolled away
[20:56] <Jolo> maybe on a dark and rainy day, it'll sit around among the other oranges
[20:56] <Jolo> "dood, i'm telling ya, i weighed as much as a melon"
[20:57] <Jolo> i mean, maybe that impresses some oranges
[20:57] <Jolo> but the melons are like, yeah, whatever, bring it if you dare
[20:57] <Min> wow
[20:57] <Min> I'm gunna save this
[20:57] <Jolo> but anyway after some time
[20:57] <FireIce> put it on the website :)
[20:58] * Min nods
[20:58] <Jolo> it'll be hopefully obvious that who's a melon and who's an orange
[20:58] *** underdog has quit IRC (Ping timeout: 181 seconds)
[20:58] <Jolo> and the thing is, there's nothing wrong with being an orange
[20:58] <Jolo> oranges taste good too
[20:58] <Sticks> except for that person that was weighed as an orange but was really a melon
[20:58] <Jolo> and when boy oranges and girl oranges get together
[20:58] <Sticks> he was to work his way up
[20:58] <Jolo> they make good juice
[20:58] <Jolo> but the thing is, at least they get to play with other oranges
[20:58] <Jolo> instead of getting squashed by melons
[20:59] <Jolo> which is not fun for either the melon or the orange
[20:59] <Jolo> right, if you're misclassified, you gotta work up or down
[20:59] <Jolo> that's the part that takes time
[20:59] <Jolo> if i got the math right this time, hopefully somebody playing every week will settle down after 3-5 wks
[20:59] <Jolo> depends on how consistent they are of course
[20:59] <Jolo> if an orange changes color all the time
[20:59] <Jolo> it'll be hard to tell if it's an orage or a melon
[21:00] <Jolo> those fruit will bounce around longer and may never settle down
[21:00] <Jolo> but thankfully that's pretty rare
[21:00] <Sticks> unfortunatly, this formula wasn't around when the games were made according to the old point scales in week1
[21:00] <Jolo> actually it was
[21:00] <Sticks> therefore, things look twisted
[21:00] <Jolo> it's just been 5.5 years
[21:00] <Jolo> i had to go back and rtfm
[21:00] <Sticks> well, the exact K value
[21:01] <Jolo> the manual that i wrote back then
[21:01] <Jolo> yeah, well K was appropriate for back then at 104 and change
[21:01] <Jolo> 104.35 iirc
[21:01] <Jolo> but there were fewer mac players etc.
[21:01] <Jolo> and we also had a better sense of their real skill
[21:01] <Jolo> cuz everybody had been playing everybody else for years already
[21:01] <Jolo> in retrospect, K=100 was way too big for you guys
[21:02] <Jolo> also, having essentially only 3 ratings was bad too
[21:02] <Jolo> we get these silly 13 way ties
[21:02] <Jolo> ties are not good for rating people
[21:02] <Jolo> or fruit
[21:02] <Jolo> it's hard to tell oranges from melons if they're the same weight
[21:02] <Jolo> every time somebody won, they'd shoot way up
[21:03] <Jolo> which looks real exciting to that guy i'm sure
[21:03] <Jolo> but it's really a mess
[21:03] <Jolo> b/c when you go up 10 ranks, that means you've skipped potentially 2 divisions
[21:03] <Jolo> nobody improves that much that quickly
[21:03] <Jolo> so then you get overmatched and crushed 0-3
[21:03] <Jolo> and you drop right back down again
[21:03] <Jolo> it's like the scale that never stops bouncing, up down up down
[21:04] <Jolo> heh, for that you blame Min
[21:04] <Jolo> hmm
[21:04] <Min> ack
[21:04] <Min> don't blame me
[21:04] * Min hides
[21:04] <Jolo> anyway
[21:04] <Min> <Min> with all the good people having 1800
[21:04] <Jolo> but it doesn't matter if people are mis ranked
[21:05] <Min> and well
[21:05] <Jolo> they'll move soon enough
[21:05] <Min> then jolo used my listing
[21:05] <Min> which was pretty random
[21:05] <Min> and seperated them all by 25 points
[21:05] <Jolo> in dark's case, he still dropped 191 points after wk 1
[21:05] <Min> each
[21:05] <Jolo> versus like 220 i think in the old scheme
[21:05] <Jolo> so both ways he got punished
[21:05] <Min> which put unknown way at the top n'stuff
[21:05] <Jolo> but in the new scheme he dropped 8 ranks
[21:05] <Jolo> which is enough to make sure he's moving down
[21:05] <Jolo> without matching him up against newbies
[21:06] <Jolo> unfortunately this didn't take effect until too late
[21:06] <Jolo> so last night he got matched against newbies
[21:06] <Jolo> if i could turn back time and redo the math between week 1 and 2
[21:06] <Jolo> Dark would've probably ended up at the bottom of div1 instead of middle of div2
[21:06] <Jolo> and chances are div1 would've been a lot closer
[21:06] <Jolo> as well as div2
[21:07] <Sticks> I'm still troubled by the initial rank, but I'm glad the beginning is over with
[21:07] <Jolo> the spreading out is also intentional
[21:07] <Jolo> that way you avoid those uncertainties that come with ties
[21:07] <Jolo> it also accelerates the process a bit
[21:08] <Jolo> good players start higher
[21:08] <Jolo> unless they are grossly overrated
[21:08] <Jolo> in which case they can move hundreds of points at a time downward
[21:08] <Jolo> you can theoretically win K points every game
[21:08] <Jolo> if you are like #35 and #34 and beat #1 and #2
[21:08] <Jolo> so even if we screw up completely and put the best player at the bottom
[21:09] <Jolo> he can win 3K or 3*75 = 225 points per division
[21:09] <Jolo> if he plays twice a week, that's 450 pts
[21:09] <Sticks> I suppose that the actual rank matters more than the pts
[21:09] <Jolo> 2 wks later he'd be on top again
[21:09] <Jolo> riiiight
[21:09] <Jolo> now yer getting it
[21:09] <Jolo> it's the _order_
[21:09] <Sticks> if someone gets misplaced
[21:09] <Jolo> that's a pretty fine point
[21:09] <Sticks> I mean
[21:10] <Jolo> the ratings do matter in the math
[21:10] <Jolo> but the point is to sort people out for fair divisions
[21:10] <Jolo> and for that it don't matter what the point spread is
[21:12] <Sticks> I'm satisified
[21:14] <Min> heh
[21:14] <Min> good!
[21:15] <SuicidJky> i am also satisified
[21:15] <SuicidJky> as long as u promise no more big changes =)
[21:15] <Sticks> it not the rank, its the position
[21:16] <Sticks> you went up 3 the first week, then 0
[21:16] <Sticks> you got put in the right place
[21:16] <Sticks> I went up 4 the first week, then 1 more next week
[21:17] <Sticks> I'm almost to my "spot" as I slow down
[21:17] <Sticks> smegs, on the other hand, was overanked
[21:17] <Sticks> rather, ranked well
[21:17] * SuicidJky makes mental note
[21:17] <Sticks> he came out even for two weeks of play
[21:17] <SuicidJky> loading 16 players on one computer is not smart
[21:19] <FireIce> [23:13] <SuicidJky> as long as u promise no more big changes =)
[21:19] <FireIce> I think Jolo made it clear that he reserves the right to make any changes he wants later on
[21:21] <Jolo> shit
[21:21] <Jolo> spent hrs talking about fruit
[21:21] <Jolo> forgot to give my cat her insulin :(
[21:21] <FireIce> :)
[21:21] <FireIce> ouch
[21:21] <Jolo> SuicidJky: well never say never (re big changes)
[21:21] <Jolo> but obviously i would only do it if it was a total disaster
[21:21] <Jolo> like if i completely fuxed up somehow this time
[21:22] <Jolo> and if i did, you'd know about it
[21:22] <Jolo> b/c people would be bouncing all over and mismatched like crazy
[21:23] <Jolo> btw feel free to use the ibl forms
[21:23] <Jolo> you can play hypothetical games
[21:23] <Jolo> to get a better feel of how the ratings work
[21:23] <SuicidJky> hmmmm
[21:23] * SuicidJky goes playing
[21:23] <Jolo> oops
[21:24] <Jolo> i may have fuxed up
[21:24] <Jolo> this doesn't change the rankings
[21:24] <Jolo> but the verbose game results i pasted might have been based on the old K
[21:24] <Jolo> the rankings were calculated separately offline by a diff program that i did update
[21:24] <Jolo> lemme check
[21:25] <Jolo> yeah fux me
[21:25] <Jolo> i'll need to fix that a bit
[21:25] <Jolo> the ratings and rankings on the top half of http://www.lgm.com/bolo/ibl/ratings.html will not change
[21:26] <Jolo> it's just the bottom half, the stuff where "so and so beat so and so and lost X points"
[21:26] <Jolo> that's a bit off
[21:26] <FireIce> yeah jolo
[21:26] <FireIce> there's a slight error
[21:26] <Jolo> i think
[21:26] <FireIce> maybe not in ranking
[21:26] <Jolo> unless i fuxed both up, heh
[21:26] <Jolo> i know the bottom is fuxed
[21:26] <Jolo> nobody should've gained or lost more than K=75 pts
[21:26] <Jolo> and some did
[21:26] <FireIce> but in my change...it says I lost 15 points week 2 when i really gained points
[21:26] <Jolo> heh fux me
[21:27] <FireIce> but the overall score is still right
[21:27] <FireIce> it just says i lost points when i didn't
[21:27] <FireIce> no biggie
[21:27] <Jolo> err wtf?
[21:27] <FireIce> i mean it says i lost points when i show a point increase in the point total
[21:27] <Jolo> in week 2 you won +33 and +119 in the bottom half
[21:27] <Jolo> in the top half it added up to +115
[21:27] <Jolo> which i think is right
[21:28] <FireIce> FireIce 0 1685 -15 1a2+1b3;3-
[21:28] <Jolo> yup in fact 115 is 3/4 of the bigger sum
[21:28] <FireIce> io mean after week 1
[21:28] <Jolo> oh that's week 1
[21:28] <Jolo> lemme check that
[21:28] <FireIce> yeah
[21:28] <Jolo> o that's ok
[21:28] <FireIce> new
[21:28] <FireIce> 14 Galt 0 1615
[21:28] <Jolo> b/c i didn't use the web forms at all for week 1
[21:28] <Jolo> i had to do that painfully by hand
[21:28] <FireIce> then i have -15 points, but it increase to 1685
[21:29] <FireIce> oh
[21:29] <FireIce> oh ok
[21:29] <Jolo> week 2 i basically re-reported the results
[21:29] <Jolo> using the web form
[21:29] <FireIce> so you had moved up ranking to like 1700 ..
[21:29] <Jolo> which is what generates that report story
[21:29] <FireIce> ah ok
[21:30] <Jolo> i can prob change the offline program to make that story too
[21:30] <Jolo> it's all there, just need to add some printf's
[21:30] <Jolo> i might do that just for my sanity
[21:31] <Sticks> creating a spread sheet of who you won/lost to really helps understand where youa re
[21:35] <Jolo> i'll just put a note on the week 2 page
[21:35] <Jolo> i ain't gonna redo the bottom half
[21:36] <Jolo> i need to sleep more than i need that fixed :)
[21:36] <Jolo> later guys
[21:36] <Jolo> hope this clears things up
[21:36] <Jolo> if you have any Qs i'd be happy to explain more
[21:36] <FireIce> bye jolo
[21:36] <Jolo> just be real careful about bitching at Min
[21:36] <Jolo> he too is spending a lot of time and energy on this
[21:36] <Jolo> he's barely gotten to play at all
[21:37] <Jolo> so be real careful to draw the line between asking questions and bitching
[21:37] <Jolo> one is constructive, the other pisses people off :)
[21:37] <Jolo> ttfn
[21:37] *** Jolo (jyl@bishop.mc.duke.edu) has left #ibl
[20:45] <Jolo> the most fair starting thing is really to put everybody at the same level
[20:45] <Sticks> but that didn't happen
[20:46] <Jolo> but that would prob add 3 or 4 weeks to the time it takes
[20:46] <Jolo> we'd barely have fair ratings after 10 wks
[20:46] <Jolo> this is why when you rank college teams, you do a pre-season ranking
[20:46] <Jolo> like for basketball or football
[20:46] <SuicidJky> what bout 5 points in order of signup
[20:46] <Jolo> that's pretty much the same thing too
[20:46] <Jolo> i intentionally spread them out to about 1/3 of the K value
[20:46] <Jolo> there's a reason for that
[20:47] <Jolo> if you play an equal guy and beat him, you gain 1.5 K points
[20:47] <Jolo> that's just how the formula works, no magic there
[20:47] <Jolo> err 0.5 K
[20:47] <Sticks> the magic K, mysterious
[20:47] <SuicidJky> 1.5 would be alot more jumping
[20:47] <Jolo> it's just a constant
[20:47] <Sticks> I know
[20:48] <Jolo> do you guys know about slope?
[20:48] <Sticks> but its unknown to us
[20:48] <Sticks> of a line or...
[20:48] <Jolo> right, but not a line
[20:48] <Jolo> it's nonlinear :)
[20:48] <Jolo> let me see if i can come up w/ an analogy
[20:48] <Jolo> u guys know the scales in grocery stores for produce?
[20:48] * FireIce knows about slope...
[20:48] <Sticks> yeah..
[20:49] <Jolo> like if you throw a watermelon in one, it bounces around?
[20:49] <Jolo> but eventually it stops bouncing and gives you the right weight
[20:49] <Jolo> K is kinda like how stiff the spring is
[20:49] <Jolo> if you have a loose spring, it will bounce all over the place
[20:49] <Jolo> make it hard to read
[20:49] <Jolo> and takes a long time to converge to the final weight
[20:49] <Sticks> loose = big K or small K?
[20:50] <Jolo> loose = big K
[20:50] <Jolo> (it's actually a really shitty analogy but i can't think of a better one right now)
[20:50] <Sticks> no, I get it
[20:50] <Jolo> so generally you don't want a really big K
[20:50] <Jolo> it looks like stuff is happening, very exciting
[20:50] <Jolo> but in reality you can't tell how fscking heavy the melon is
[20:51] <Jolo> to make things worse
[20:51] <Jolo> imagine if you had 30 scales
[20:51] <Jolo> all hooked to each other
[20:51] <Jolo> all bouncing
[20:51] <Jolo> that's what this system is
[20:51] <Jolo> so every time you play a game, you throw 4 melons in 4 of the scales
[20:52] <Jolo> everything bounces around and finally settles
[20:52] <Jolo> except then you throw 4 more melons in
[20:52] <Jolo> melons = players
[20:52] <Sticks> so then what would the starting points mean in this analogy?
[20:52] <Jolo> heh
[20:52] <Jolo> imagine if you could calibrate the scale a bit
[20:52] <Jolo> like for heavy melons vs. a single orange
[20:53] <Jolo> but pretend you're blind and have no idea wtf kinda fruit you're measuring
[20:53] <Jolo> your best guess is probably to have a single average calibration
[20:53] <Jolo> but then you'd throw a melon on and tilt the scale like crazy
[20:53] <Jolo> and the orange barely moves the needle
[20:53] <Jolo> eventually it'll all sort out
[20:53] <Jolo> but it takes longer to figure out
[20:53] <Jolo> but if you knew up front that you had melons vs. oranges
[20:54] <Jolo> you could use the heavy scale for the melons
[20:54] <Jolo> etc.
[20:54] <Sticks> but there will be someone getting shafted
[20:54] <Jolo> yup, that is always true
[20:54] <Sticks> like, using a heavy scale for an orange
[20:54] <Jolo> yup
[20:54] <Sticks> excellent...much clearer
[20:54] <Jolo> you'll pick up an orange that looks a melon
[20:54] <Jolo> has the color of a melon
[20:54] <Jolo> turns out it's not heavy at all
[20:54] <Jolo> you throw it in the heavy scale and it's all wrong
[20:54] <Jolo> but you'd notice that pretty quickly
[20:54] <Jolo> "wtf, that's no melon"
[20:55] <Jolo> now imagine if you had a pile of fruit sitting there
[20:55] <Jolo> that refuses to be weighed
[20:55] <Jolo> or maybe they get weighed once and then run away
[20:55] <Jolo> you'll never know for sure how heavy they are
[20:55] <Sticks> hehe
[20:55] * Min chuckles
[20:55] * Min saves this
[20:55] <Jolo> or maybe they'll get weighed wrong and appear real heavy
[20:55] <Sticks> running fruit
[20:55] <Jolo> but the thing is, who cares
[20:56] <Jolo> if you're just trying to make a fruit salad
[20:56] <Jolo> with the right mix of fruits
[20:56] <Sticks> they'll all end up dead
[20:56] <Jolo> so there's this orange that got mistaken for a melon once
[20:56] <Jolo> and after getting overweighted, it rolled away
[20:56] <Jolo> maybe on a dark and rainy day, it'll sit around among the other oranges
[20:56] <Jolo> "dood, i'm telling ya, i weighed as much as a melon"
[20:57] <Jolo> i mean, maybe that impresses some oranges
[20:57] <Jolo> but the melons are like, yeah, whatever, bring it if you dare
[20:57] <Min> wow
[20:57] <Min> I'm gunna save this
[20:57] <Jolo> but anyway after some time
[20:57] <FireIce> put it on the website :)
[20:58] * Min nods
[20:58] <Jolo> it'll be hopefully obvious that who's a melon and who's an orange
[20:58] *** underdog has quit IRC (Ping timeout: 181 seconds)
[20:58] <Jolo> and the thing is, there's nothing wrong with being an orange
[20:58] <Jolo> oranges taste good too
[20:58] <Sticks> except for that person that was weighed as an orange but was really a melon
[20:58] <Jolo> and when boy oranges and girl oranges get together
[20:58] <Sticks> he was to work his way up
[20:58] <Jolo> they make good juice
[20:58] <Jolo> but the thing is, at least they get to play with other oranges
[20:58] <Jolo> instead of getting squashed by melons
[20:59] <Jolo> which is not fun for either the melon or the orange
[20:59] <Jolo> right, if you're misclassified, you gotta work up or down
[20:59] <Jolo> that's the part that takes time
[20:59] <Jolo> if i got the math right this time, hopefully somebody playing every week will settle down after 3-5 wks
[20:59] <Jolo> depends on how consistent they are of course
[20:59] <Jolo> if an orange changes color all the time
[20:59] <Jolo> it'll be hard to tell if it's an orage or a melon
[21:00] <Jolo> those fruit will bounce around longer and may never settle down
[21:00] <Jolo> but thankfully that's pretty rare
[21:00] <Sticks> unfortunatly, this formula wasn't around when the games were made according to the old point scales in week1
[21:00] <Jolo> actually it was
[21:00] <Sticks> therefore, things look twisted
[21:00] <Jolo> it's just been 5.5 years
[21:00] <Jolo> i had to go back and rtfm
[21:00] <Sticks> well, the exact K value
[21:01] <Jolo> the manual that i wrote back then
[21:01] <Jolo> yeah, well K was appropriate for back then at 104 and change
[21:01] <Jolo> 104.35 iirc
[21:01] <Jolo> but there were fewer mac players etc.
[21:01] <Jolo> and we also had a better sense of their real skill
[21:01] <Jolo> cuz everybody had been playing everybody else for years already
[21:01] <Jolo> in retrospect, K=100 was way too big for you guys
[21:02] <Jolo> also, having essentially only 3 ratings was bad too
[21:02] <Jolo> we get these silly 13 way ties
[21:02] <Jolo> ties are not good for rating people
[21:02] <Jolo> or fruit
[21:02] <Jolo> it's hard to tell oranges from melons if they're the same weight
[21:02] <Jolo> every time somebody won, they'd shoot way up
[21:03] <Jolo> which looks real exciting to that guy i'm sure
[21:03] <Jolo> but it's really a mess
[21:03] <Jolo> b/c when you go up 10 ranks, that means you've skipped potentially 2 divisions
[21:03] <Jolo> nobody improves that much that quickly
[21:03] <Jolo> so then you get overmatched and crushed 0-3
[21:03] <Jolo> and you drop right back down again
[21:03] <Jolo> it's like the scale that never stops bouncing, up down up down
[21:04] <Jolo> heh, for that you blame Min
[21:04] <Jolo> hmm
[21:04] <Min> ack
[21:04] <Min> don't blame me
[21:04] * Min hides
[21:04] <Jolo> anyway
[21:04] <Min> <Min> with all the good people having 1800
[21:04] <Jolo> but it doesn't matter if people are mis ranked
[21:05] <Min> and well
[21:05] <Jolo> they'll move soon enough
[21:05] <Min> then jolo used my listing
[21:05] <Min> which was pretty random
[21:05] <Min> and seperated them all by 25 points
[21:05] <Jolo> in dark's case, he still dropped 191 points after wk 1
[21:05] <Min> each
[21:05] <Jolo> versus like 220 i think in the old scheme
[21:05] <Jolo> so both ways he got punished
[21:05] <Min> which put unknown way at the top n'stuff
[21:05] <Jolo> but in the new scheme he dropped 8 ranks
[21:05] <Jolo> which is enough to make sure he's moving down
[21:05] <Jolo> without matching him up against newbies
[21:06] <Jolo> unfortunately this didn't take effect until too late
[21:06] <Jolo> so last night he got matched against newbies
[21:06] <Jolo> if i could turn back time and redo the math between week 1 and 2
[21:06] <Jolo> Dark would've probably ended up at the bottom of div1 instead of middle of div2
[21:06] <Jolo> and chances are div1 would've been a lot closer
[21:06] <Jolo> as well as div2
[21:07] <Sticks> I'm still troubled by the initial rank, but I'm glad the beginning is over with
[21:07] <Jolo> the spreading out is also intentional
[21:07] <Jolo> that way you avoid those uncertainties that come with ties
[21:07] <Jolo> it also accelerates the process a bit
[21:08] <Jolo> good players start higher
[21:08] <Jolo> unless they are grossly overrated
[21:08] <Jolo> in which case they can move hundreds of points at a time downward
[21:08] <Jolo> you can theoretically win K points every game
[21:08] <Jolo> if you are like #35 and #34 and beat #1 and #2
[21:08] <Jolo> so even if we screw up completely and put the best player at the bottom
[21:09] <Jolo> he can win 3K or 3*75 = 225 points per division
[21:09] <Jolo> if he plays twice a week, that's 450 pts
[21:09] <Sticks> I suppose that the actual rank matters more than the pts
[21:09] <Jolo> 2 wks later he'd be on top again
[21:09] <Jolo> riiiight
[21:09] <Jolo> now yer getting it
[21:09] <Jolo> it's the _order_
[21:09] <Sticks> if someone gets misplaced
[21:09] <Jolo> that's a pretty fine point
[21:09] <Sticks> I mean
[21:10] <Jolo> the ratings do matter in the math
[21:10] <Jolo> but the point is to sort people out for fair divisions
[21:10] <Jolo> and for that it don't matter what the point spread is
[21:12] <Sticks> I'm satisified
[21:14] <Min> heh
[21:14] <Min> good!
[21:15] <SuicidJky> i am also satisified
[21:15] <SuicidJky> as long as u promise no more big changes =)
[21:15] <Sticks> it not the rank, its the position
[21:16] <Sticks> you went up 3 the first week, then 0
[21:16] <Sticks> you got put in the right place
[21:16] <Sticks> I went up 4 the first week, then 1 more next week
[21:17] <Sticks> I'm almost to my "spot" as I slow down
[21:17] <Sticks> smegs, on the other hand, was overanked
[21:17] <Sticks> rather, ranked well
[21:17] * SuicidJky makes mental note
[21:17] <Sticks> he came out even for two weeks of play
[21:17] <SuicidJky> loading 16 players on one computer is not smart
[21:19] <FireIce> [23:13] <SuicidJky> as long as u promise no more big changes =)
[21:19] <FireIce> I think Jolo made it clear that he reserves the right to make any changes he wants later on
[21:21] <Jolo> shit
[21:21] <Jolo> spent hrs talking about fruit
[21:21] <Jolo> forgot to give my cat her insulin :(
[21:21] <FireIce> :)
[21:21] <FireIce> ouch
[21:21] <Jolo> SuicidJky: well never say never (re big changes)
[21:21] <Jolo> but obviously i would only do it if it was a total disaster
[21:21] <Jolo> like if i completely fuxed up somehow this time
[21:22] <Jolo> and if i did, you'd know about it
[21:22] <Jolo> b/c people would be bouncing all over and mismatched like crazy
[21:23] <Jolo> btw feel free to use the ibl forms
[21:23] <Jolo> you can play hypothetical games
[21:23] <Jolo> to get a better feel of how the ratings work
[21:23] <SuicidJky> hmmmm
[21:23] * SuicidJky goes playing
[21:23] <Jolo> oops
[21:24] <Jolo> i may have fuxed up
[21:24] <Jolo> this doesn't change the rankings
[21:24] <Jolo> but the verbose game results i pasted might have been based on the old K
[21:24] <Jolo> the rankings were calculated separately offline by a diff program that i did update
[21:24] <Jolo> lemme check
[21:25] <Jolo> yeah fux me
[21:25] <Jolo> i'll need to fix that a bit
[21:25] <Jolo> the ratings and rankings on the top half of http://www.lgm.com/bolo/ibl/ratings.html will not change
[21:26] <Jolo> it's just the bottom half, the stuff where "so and so beat so and so and lost X points"
[21:26] <Jolo> that's a bit off
[21:26] <FireIce> yeah jolo
[21:26] <FireIce> there's a slight error
[21:26] <Jolo> i think
[21:26] <FireIce> maybe not in ranking
[21:26] <Jolo> unless i fuxed both up, heh
[21:26] <Jolo> i know the bottom is fuxed
[21:26] <Jolo> nobody should've gained or lost more than K=75 pts
[21:26] <Jolo> and some did
[21:26] <FireIce> but in my change...it says I lost 15 points week 2 when i really gained points
[21:26] <Jolo> heh fux me
[21:27] <FireIce> but the overall score is still right
[21:27] <FireIce> it just says i lost points when i didn't
[21:27] <FireIce> no biggie
[21:27] <Jolo> err wtf?
[21:27] <FireIce> i mean it says i lost points when i show a point increase in the point total
[21:27] <Jolo> in week 2 you won +33 and +119 in the bottom half
[21:27] <Jolo> in the top half it added up to +115
[21:27] <Jolo> which i think is right
[21:28] <FireIce> FireIce 0 1685 -15 1a2+1b3;3-
[21:28] <Jolo> yup in fact 115 is 3/4 of the bigger sum
[21:28] <FireIce> io mean after week 1
[21:28] <Jolo> oh that's week 1
[21:28] <Jolo> lemme check that
[21:28] <FireIce> yeah
[21:28] <Jolo> o that's ok
[21:28] <FireIce> new
[21:28] <FireIce> 14 Galt 0 1615
[21:28] <Jolo> b/c i didn't use the web forms at all for week 1
[21:28] <Jolo> i had to do that painfully by hand
[21:28] <FireIce> then i have -15 points, but it increase to 1685
[21:29] <FireIce> oh
[21:29] <FireIce> oh ok
[21:29] <Jolo> week 2 i basically re-reported the results
[21:29] <Jolo> using the web form
[21:29] <FireIce> so you had moved up ranking to like 1700 ..
[21:29] <Jolo> which is what generates that report story
[21:29] <FireIce> ah ok
[21:30] <Jolo> i can prob change the offline program to make that story too
[21:30] <Jolo> it's all there, just need to add some printf's
[21:30] <Jolo> i might do that just for my sanity
[21:31] <Sticks> creating a spread sheet of who you won/lost to really helps understand where youa re
[21:35] <Jolo> i'll just put a note on the week 2 page
[21:35] <Jolo> i ain't gonna redo the bottom half
[21:36] <Jolo> i need to sleep more than i need that fixed :)
[21:36] <Jolo> later guys
[21:36] <Jolo> hope this clears things up
[21:36] <Jolo> if you have any Qs i'd be happy to explain more
[21:36] <FireIce> bye jolo
[21:36] <Jolo> just be real careful about bitching at Min
[21:36] <Jolo> he too is spending a lot of time and energy on this
[21:36] <Jolo> he's barely gotten to play at all
[21:37] <Jolo> so be real careful to draw the line between asking questions and bitching
[21:37] <Jolo> one is constructive, the other pisses people off :)
[21:37] <Jolo> ttfn
[21:37] *** Jolo (jyl@bishop.mc.duke.edu) has left #ibl