| |
BGonline.org Forums
Defend your cube action - elo diff.
Posted By: Henrik Bukkjaer In Response To: Defend your cube action - elo diff. (Stick)
Date: Monday, 8 October 2012, at 9:12 a.m.
This is a key point in improving real live win percentages, if you play a mixed group of players, no doubt.
And it's the major drawback of XG (and thus the area where there's plenty of room for improvement), that it cannot provide any sort of help in analyzing these fishy questions. Gnubg is ahead in this area, and I still use it occasionally for that single reason.
---
On to your questions:
I think it's difficult to adjust based on a rating alone. There are all sorts of players, and all sorts of ratings :-)
In this match, you have played 17 points so far, so your OtB experience should begin to affect your decisions as heavily as the initial rating input. Keep that in mind when you consider my guesses to proper cube-actions below.
The rating system I'm used to deal with for these situations is the DBgF one, where the population as a whole is probably stronger, and the ratings are a little bit different in respect to short vs. long matches, and finally the ratings are probably a bit more "true". Plus you "know" most of the players anyway, so you have an idea if they are temporarily under-/overrated.
Typically I use the rating as a "initial baseline" for my adjustments, and I use it a bit conservatively (if we have to put numbers on, I'd probably say that I'd choose a MET between half and two thirds of the stated rating difference). Then I'll work my way through the match from there, trying to pick up if the player is risk averse, aggressive, really bad or not so much, or if he/she seems to have any patterns to their error-rate. Keep in mind, that most of the games we play, is live 17 pointers, so we have time to get a feeling for this.
The result is - just as you say Stick - that I typically don't look too good PR-wise, but get ratings and results that are better than what should have been, for my PR.
If you want to look good and get hyped as a good player, this is NOT the way to go about it. Both my conception of this, and my data/analysis so far, shows that the PR effect of making relatively small sane adjustments, is much much higher than people would expect it to be, because you not only "pay" for the direct errors you produce, but both players "pay" for getting into a match with more difficult decisions, and fewer non-contact/bear-in/bear-off moves (your adjustments will lead to more passes at both sides).
Actually, my analysis so far shows, that making very small adjustments in the opposite direction (what we could call underdog adjustments), will help your error-rate go lower than it would for optimal play! This is true if you are a 3+ PR player only, and more true the higher your regular PR is. I can elaborate on that subject if anyone needs further explanations.
----
OK, once again, on to your questions (I'm trying not to derail myself here):
Pos 1. Adjustments or not, this is a big double. Even against very week players. First of all, this cannot be close to borderline, so it would require some skill difference to even consider not cubing, secondary it's a position that is not so dangerous when it goes wrong against a week player in particular. No-brainer double.
Pos 2. I'd really need to have a low esteem for my opponent to drop this. I wouldn't rule out a pass here, but we're talking donkey opposition if so. For some of these estimations you CANNOT look at just the rating difference, you also have to consider the absolute rating. Otherwise you don't have a feel for how bad the player is capable of misplaying a given position vs. misplaying the remainder of the match. I haven't played on FIBS since the nineties, so I'm poking around in the dark here, but I consider 1750 to be initial rating + 250, which means that the player must have some ideas about the game and about having a big lead in a long match, thus I would take this one... I don't think the position is too difficult to play for white, so the worse the opponent, the more I would consider passing here.
Pos 3, on the other hand, I wouldn't cube even if I was just +100. I think it's a slim one, and there's no need to offer up the match on 17 hitting numbers... We're not loosing that many gammons here, and I even think we're winning more gammons than we are loosing them. I don't know what numbers XG produce for gammons, but I suppose that skill-difference would also give us an even bigger gammon advantage than that. Roll - by a wide margin.
Pos 4. Now we're looking at a different picture. We're certainly not killing our own gammons by doubling anymore, and we're loosing more gammons now than before (and with fewer options for our opponent to mis-play). Plus we have 5 more hitting numbers (22). I'm cubing here vs. a 1750 player, if my feeling for that level is correct, but against a complete donkey, I wouldn't rule out playing on!
It's very clear when you look at it, that the sort of positions you need to adjust most, are the ones that swing the match a lot (eg. recubes, not initial ones), that brings the opponent to a win or crawford game (since that carries more weight for the underdog player to reach, not being able to err as much from there, AND the fact that the player will not misplay any cubes from that point on in the game either), and in particular in positions where your (bad) opponent is not able to screw up much equity for the rest of the game! The prime example being bear-offs and other non-contact positions. But stacked on the low points and threats being closed out, don't allow much room for misplay either. (checker-wise).
I'd love to see this topic get more thread-time at BGO, which could possibly make the subject more widely accepted as an important factor of the game, in contrast to the current PR focus. Maybe it could get Xavier to prioritize efforts in this direction for XG Rel. 3.
Best regards, Henrik
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.