Calculating luck rates
David Montgomery
>I've been trying to study ways of calculating luck in backgammon.
>I started by looking at the way Snowie 1.1 does it.
>Could someone just tell me if I'm understanding this correct.
I think you have it right.
>I have some questions regarding this method. What is the correct luck
>rate for the first move? I terminated a game after move 1 having
>played 31 8/5 6/5.
I did the same experiment.
See http://www.dejanews.com/getdoc.xp?AN=425677891
>Snowie says I had a luck rate of 7.784% I'm
>not quite sure where this figure comes from.
0.154 (after roll)
- 0.074 (before roll, erroneously, as you point out)
-------
0.080
x 100 (luck per 100 moves)
-------
8.000
Which is about 7.784. I would have to look more closely to figure
out why it's 7.8 rather than 8, but a couple potential reasons are
because of rounding, and because the luck rate might have used slightly
different evaluations (e.g., different ply).
>But what looks wrong is that it also
>includes doubles which obviously can't be thrown on the first turn.
Right.
>Also shouldn't the equity before the first roll be zero, as each
>player has an equal chance and so the difference after throwing 31
>will be 0.154 and the luck rate 15.4 %?
Yes, right again.
>What about if a player is on the bar? Should we exclude this move
>when averaging the error rate?
Snowie excludes forced moves. So if you dance or partially dance
or your entering move is completely forced, it gets skipped. But
if you have a move with choices from the roof, you can still make
a mistake, so it shouldn't be excluded. And isn't.
David Montgomery
mo...@cs.umd.edu
monty on FIBS and GG
Midas
I've been trying to study ways of calculating luck in backgammon.
I started by looking at the way Snowie 1.1 does it.
Could someone just tell me if I'm understanding this correct.
Before each dice roll the position has an equity, that equity is
the average of all the following 36 combinations. By using the dice
panel we can see the moves listed from best to worst and also their
differences from the average. A better move than average obviously
indicates good luck. We can then do this for each move and take an
average to get an overall luck rate.
I have some questions regarding this method. What is the correct luck
rate for the first move? I terminated a game after move 1 having
played 31 8/5 6/5. Snowie says I had a luck rate of 7.784% I'm
not quite sure where this figure comes from. The dice panel for move
1 shows the average position equity before rolling as 0.074 and the
equity for rolling 31 as 0.154. But what looks wrong is that it also
includes doubles which obviously can't be thrown on the first turn.
Also shouldn't the equity before the first roll be zero, as each
player has an equal chance and so the difference after throwing 31
will be 0.154 and the luck rate 15.4 %?
What about if a player is on the bar? Should we exclude this move
when averaging the error rate?
Are there any better ways to calculate true error rates ?
CYA
Midas