Cut Ken some slack, his work is still in the early stages
Posted By: Barry Silliman In Response To: Hugely flawed to the point of being irrelevant (Rod)
Date: Saturday, 16 November 2013, at 1:48 a.m.
In Response To: Hugely flawed to the point of being irrelevant (Rod)
Ken knows there are some flaws in his data. He's a smart guy and he has some good ideas. I think what he's primarily interested in is coming up with an objective way to measure performance in the ABT-rated divisions. He knows that the numbers he started with contain jackpot events and should be filtered out. He did that manually and sure he made some errors. I have an automated process that basically screen-scrapes the publicly available HTML pages from backgames.org, categorizes each event into OPEN, ADVANCED and NOVICE. So side events such as Masters jackpots, Open jackpots, Amateur jackpots, etc. fall into OPEN or ADVANCED. Any person who plays in more than one category has a separate ELO for each category. That is why the ratings ELOs published bi-monthly in the USBGF magazine differ slightly from what is on backgames.org, because on backgames.org everybody is in one database. For example, Chen-Fu Yu, who has the highest rating on backgames.org at 1763.93, has an ADVANCED rating of 1776.55 in my system and an OPEN rating of 1495.44 in my system. I use the same ELO formula that backgames.org use. For example, in my system Chen's achievement of 1763.93 buys him nothing when he plays in OPEN, he started his OPEN rating at 1500.
These ratings aren't perfect. In fact, even though I'm listed as the 'USBGF Ratings Editor' I'm a believer that ELO ratings are WAY over-rated both due to the luck factor and small sample size. The only reason I am the USBGF Ratings Editor is due to a bizarre set of circumstances involving the United States Postal Service, a couple of bagel sandwiches, and a cross-country road trip. I am more than willing to offload that duty and return all of the compensation I have received for the post to boot.
I think Ken has some good ideas and he should be allowed to explore them before we dismiss it. Whereas my system simply calculates ELO and completely ignores ABT points, Ken has suggested ABT points per event as an indicator as to when somebody *should* be moved up. (Hold that word 'should' for a moment, I'll get to that later).
Let me give a summary of what Ken is proposing, these may not be his exact numbers, this is an example- there are 0.50 ABT points per person given in an ADVANCED ABT event. Ken is suggesting some higher number, perhaps 0.75 or 1.00 points per event, from Advanced, would indicate that person should move up to OPEN. He suggests dividing the number of points by a minimum number of events, such as 15. This would keep people from being moved up after just 1 lucky win- however many ABT points they got would be divided by the maximum of 15 or the number of events they entered.
Don't pick apart the particular numbers in the paragraph above, they may not be what Ken has proposed, I'm going by memory and I'm in the middle of one of these right now.
Now the reason I say *should* is because the effort that Ken is making is the kind of effort that will bring improvements to this game- organic efforts, bottom-up efforts driven by motivated volunteers, not some top-down diktats from some mythical national or international organization. If Ken puts together some worthwhile work efforts that individual tournament directors find useful, it will catch on and become more widespread. If it turns out to not be useful, then we'll thank Ken for the effort and maybe something else he tries will be useful.
In the US as it exists today, there is a lot of prestige and desire to get the #1 ABT spot and that trophy. If you agree with nothing else in this post Rod, I'm sure you'll agree with that. Now if Bill Davis looked at Ken's system and said "no tournament will be an ABT-sanctioned tournament without using Ken's method of weeding out the sandbaggers", then it would take effect tomorrow. But Bill isn't likely to do that, nor am I saying he should. But if Ken provides something helpful that some TD likes and decides to experiment with it and he likes it and it catches on by word of mouth, then that's great.
So bottom line- Ken knows there are some flaws in his data at present, there are some process improvements needed, and it won't be ready overnight. But I think Ken's a smart guy and he may be onto something that might be helpful down the road.
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.