I committed a cube error according to XG but I think the "Cube Information" window actually says the opposite. I am confused. Anyone have any insight?

See image attached.

1. With a live cube, my double point should have been 70% according to "Cube Information" and so No Double would be correct at 63.29% winning chances. But according to the analysis, Double/Take was correct and I made a error/blunder.

2. XG "Cube Information" (shown on right) defaults the Gammon Ratio to different values than actually analyzes (shown on left).

The "live cube" number assumes all subsequent recubes are going to be perfectly efficient. So it's something close to an upper bound on how high your double point could possibly be in a particular match situation (taking into account the score and the gammon percentages, but not anything about the actual position). And the dead cube number is like a lower bound.

Or to say it another way -- whether you have a double is not solely determined by comparing your winning chances to these numbers, these numbers are just a rough guide as to what's going on at the match score.

The dead cube numbers, at least, are calculable over the board, so they can be a useful metric in a match.

Little more discussion here:
https://forumserver.twoplustwo.com/1...stion-1765866/

Thank you for your reply _Z_. I read through your reply here and in the linked thread.

My confusion is maybe based on a few assumptions, which you might be able to point out where I've gone wrong?

1. XG's "Cube Information" is the same thing as market window i.e. take/drop/too-good points, i.e. depicted here: https://bkgm.com/articles/mpd.html#gammon_potential

2. Market window is based on a simple formula from match equity tables (shown in XG's Analyze -> Match Equity Table from the menu bar: https://www.bkgm.com/faq/Matches.htm...my_take_point_

3. XG's cube error value probably takes into account "early/late cost ratio," more or less (although the way XG solves for these values is over my head.) For example, even though the 2a-2a market window linked above shows 30-70, it is known to be correct to double when you have 50.1% or greater winning chances typically.

Perhaps comparing "Cube Information" and "Cubeful Equities"/errors is comparing apples and oranges? So in other words, in another example shown below, even though I made the right move according to the market window (double), were I to roll out every possible ending to the game, it would have shown I somehow would end up with more wins with no-double decision.

I think that's right, but just for the "dead cube" numbers. The live cube numbers people don't calculate over the board, and I'm not even sure how XG calculates them.

You don't necessarily have a cube when you are in your doubling window (in other words, over XG's "Double Point"). Certainly that's true. But I don't know what XG is doing under the hood, nor do I know about early/late cost ratio. I suspect there's probably a lot going on and it's not something that can be modeled in some mathematically correct way.

I've certainly seen cases where some bot thinks something is a take because it assumes the taker has a lot of recube potential. But a rollout says pass because the position just doesn't lend itself to a lot of good recubes. Say in a volatile bear off position.

I wouldn't say you're comparing apples to oranges, but the cube information window just shows part of the picture, so it's like you're comparing an apple to a slice of an apple.