Life In 19x19

Since I so rarely seem to run into anyone on KGS, do any of you play on DGS or OGS? With a turn based server, timing becomes a lot less of an issue.

Hey skydyr and everyone

I've made an OGS account. IGN - Unusedname

So anyone who wants to send me a challenge is free to cause i have no idea how this thing works o_o

Could I make a suggestion here?

Instead of asking a group of people most of whom can be assumed to lack the background in mathematics to properly answer the question (except as a pure guess) how about changing the poll question into something that we can assume that they do understand and then the math folks among us can translate that into the proper number of games won/lost.

In other words, as "what is the acceptable level of error changing rank assignments?". Is it OK to be wrong one time in four? (changed rank but shouldn't have). One time in eight? Or am I going in the wrong direction? Would anybody be happy with changing the rank as soon as more likely than not justified? (in other words, wrong half the time).

Once that is decided upon, those of us competent in statistics can easily compute the number of wins necessary for that degree of certainty.

For those who don't understand my point, consider flipping an honest coin X times. The coin is assumed honest so no change of rank is correct (and a rank change wrong). Say that X is three. Do you think three heads in a row (promote) or three tails in a row demote)would be particularly unlikely outcomes? (for example, I'm seeing some suggest two in a row and that's as likely as not to be wrong --- and even the non-math folks should easily be able to see why as the pair of outcomes resulting in a rank change (HH and TT) and the pair that do not (HT and TH) have the same likelihood.

I am not too worried if the handicap gets out of whack a bit. It'll make for a challenging set of games and keep things dynamic.

Just to mention a bit different setup, used in our city league, works well there:

Initially everybody who joins gets a questimate of points based on current rank somewhere, say 1d= 70, a rank up or down is +-5.

For every win, you get +1 point, for every loss you lose -1 point. To get correct handi, you count the difference in points, divide by 5.

In effect, if two players play against each other, initially at even, and one keeps winning: 70/70 = even -> 71/69 = even -> 72/68 = 1 handi etc.

So the effect is quite like 2 wins -> change in handicap, but I think it is nicer and easier for those situations where people play multiple opponents, each of them quite seldom. YMMV.

The West Michigan Go Club uses this system:

First, players try to determine their starting handicap against each other. They play an even game and then estimate based on the outcome (usually a stone for every 10 points lost by - sometimes an additional game is played at an intermediate handicap to see if that is correct). After that, we play starting at that handicap.

The games are then best of 3 series. So, let's say John and Tim play where Tim has 3 stones. If Tim wins 2 games in a row, they move to 2 stones. If John then wins 2 they move back to 3. If each wins 1 game, then the next game is pretty exciting as it determines how the handicap moves. (This is actually how 2 of our players have been going for a couple months) I like this system because if feels motivating. I had to study hard this past week because I was playing a third match at the club.

As for the question of a 1 stone handicap, that's the weaker player taking black with no komi.

However, whatever system you guys agree to, I'd be interested in joining this League.

moyoaji wrote:The West Michigan Go Club uses this system:

First, players try to determine their starting handicap against each other. They play an even game and then estimate based on the outcome (usually a stone for every 10 points lost by - sometimes an additional game is played at an intermediate handicap to see if that is correct). After that, we play starting at that handicap.

The games are then best of 3 series. So, let's say John and Tim play where Tim has 3 stones. If Tim wins 2 games in a row, they move to 2 stones. If John then wins 2 they move back to 3. If each wins 1 game, then the next game is pretty exciting as it determines how the handicap moves. (This is actually how 2 of our players have been going for a couple months) I like this system because if feels motivating. I had to study hard this past week because I was playing a third match at the club.

As for the question of a 1 stone handicap, that's the weaker player taking black with no komi.

However, whatever system you guys agree to, I'd be interested in joining this League.

I think this is what we'll adopt a the Rochester Hills Go Club. We were actually hoping to try to arrange a tournament with you guys sometime this fall/winter, like an east-meets-west meetup to be hosted either in Lansing or Detroit.

moyoaji wrote:However, whatever system you guys agree to, I'd be interested in joining this League.

The system as discussed was no komi matches, every X wins ahead the handicap changes, and you keep track of handicap individually. How fine grained the handicap change is is up to you and the person you are playing. I think there was some consensus on half steps (BW > all B > B2 > all 2) but I may be mistaken.

I'd also be in.

When and where do we start ?

Sorry I haven't replied to this thread for a while; I keep seeing it when I haven't got much time. I'll update the first post with the new members/additional info later today.

Mike Novack wrote:Could I make a suggestion here?

Instead of asking a group of people most of whom can be assumed to lack the background in mathematics to properly answer the question (except as a pure guess) how about changing the poll question into something that we can assume that they do understand and then the math folks among us can translate that into the proper number of games won/lost.

In other words, as "what is the acceptable level of error changing rank assignments?". Is it OK to be wrong one time in four? (changed rank but shouldn't have). One time in eight? Or am I going in the wrong direction? Would anybody be happy with changing the rank as soon as more likely than not justified? (in other words, wrong half the time).

This doesn't work. Unless you really deeply understand the space of possible outcomes, and tradeoffs involved between various goals, this apparently precise method is not going to yield sensible results.

You can't even choose a good metric to optimize until you really know the statistics. So better to keep things wholly based on intuition/feel until you get to the kind of system (KGS, AGA, EGF) that demands a good statistically based system, and let that system be developed by the statisticians.

I've updated the first post. Let me know if I've missed anything.

Yeah, I would like to join.

Sorry Moyoaji, I've added you now.