|
BGonline.org Forums
From my "5pt vs. 3pt" folder...
Posted By: sebalotek In Response To: From my "5pt vs. 3pt" folder... (Timothy Chow)
Date: Tuesday, 17 January 2017, at 10:50 a.m.
TC>> The way to improve is to record the errors that I make, and see which kind of error I make more often -snip- Then, when in doubt, make the play that counteracts your known erroneous bias.
I agree with the improved effectiveness of your suggested method of learning and improving. But there is one big problem with it.
How many of us have a large sample size of such detailed historical '3 point or 5 point' data extracted, organised and available to analyse and come to a conclusion?
From the answers received so far, I would guess hardly anybody.
Perhaps not even you with your enviable categorised folders of interesting decisions as, so far, you have not been even able to offer a guesstimate.
TC>> But what constitutes a "tough" decision is going to vary from one person to another, and there's not going to be any way to arrive at any kind of objective figure.
In order to quantify difficulty my suggestion would be, in an ideal world, to be able to select and extract types of "tough" decisions from our historical gamedata not only by type of decision (3 vs 5) but also by a user definable 'closeness' of the decision in terms of equity e.g. say, between a range of -0.00 -> -0.06
So in absence of a personal library of say, 1000 analysed 3 vs 5pt positional decisions (which may take years to collate) it seems to me reasonable and generally stimulating/interesting to ask the community for an interim guesstimate of frequency.
TC>> How are you going to use this information? The only algorithm that makes sense is to choose the 5pt if X > 50 and to choose the 3pt if X < 50.
This is indeed the type of tie-break I had in mind.
I look forward to the next generations of backgammon software which might be able to offer us more sophisticated and automated analysis for counteracting our personal error-types.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.