| |
BGonline.org Forums
Let's Be Consistent -- An Error is an Error
Posted By: Marty Storer In Response To: Let's Be Consistent -- An Error is an Error (Bill Riles)
Date: Tuesday, 21 July 2009, at 5:48 p.m.
You can tag somebody all you want for errors, but you may miss a lot of what's going on.
Error rate is just one metric, and as people mention, there are plenty of others. I believe the ultimate, over the long run, is Elo rating (not ELO--no Electric Light Orchestra-type acronym, Arpad Elo was the man's name). But in the short run, ratings are subject to plenty of fluctuation. We have to use whatever tools we have if we want to edge closer to reality.
In the absence of good judgment, or in the absence of confidence in our judgment, it's better to stick to the objective metrics themselves. I think that's a large part of Bill's argument, but I don't think it has to be "rationalization" to say that a technical no-double, that gained a huge amount of equity by drawing a pass, was a good decision. It may have been fundamentally an error (doubler may have acted for all the wrong reasons and just been lucky), or it may have been laudable (doubler may have known the position was no-double yet applied good judgment and table presence and been rewarded). Without other information, we don't know.
Certainly, I wouldn't want a bot to adjust the error rate automatically if a "wrong double" led to a "wrong pass." I want to know my objective error rate, then I can adjust it in my own mind or notebook according to what I was actually doing during the match. If I rationalize my mistakes overly, "my bad," and I'll incur the consequences involved in having an inflated idea of my own skills. If I claim laudable opponent-handling decisions publicly, I'm subject to skepticism and I'll have to defend myself, which is just normal discourse, and all well and good.
I played a long match pre-bot and my rolled-out error rate was in the 8's. Horrible for me, maybe even for the time. But I know I handled my opponent with great skill.
In the prior match from that tournament, I made a huge "error" by redoubling a position that was a big no-double, but I did read my opponent to pass, and he did, and it was a monstrous swing in my favor. So I can justifiably pat myself on the back as well as remembering the nice experience of counting all the money.
Now what were my "true" error rates in those matches? I don't know, and I know my present self would rake my pre-bot self over the coals. But I don't feel too bad about my performance, considering.
Yet Bill is right that one has to be careful about rationalizing.
In trying to estimate the difference between my rating and an opponent's for whom I have no prior estimate, in the absence of any information I assume we're equal; if I have what I think is good information, I make an initial estimate and then close or widen the gap by 50 points (so that if I think I'm stronger the disparity is less in my favor, and if I think I'm weaker the disparity is more in my opponent's favor). That's my perhaps oversimple attempt to correct for overestimation of my own skill. Does it work? I believe it leads to better estimates on my part. Am I sure of that? Not at all. Work in progress....
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.