MHS: SP21 TC Rounding "nearest"? "floor"?

[MHS]

Some says "nearest" is recommended for SP21.
Are ther good reason to use?
But, I suppose most BJ players use "floor".

[Aruuba]

> Some says "nearest" is recommended for SP21.
> Are ther good reason to use?
> But, I suppose most BJ players use "floor".

In SP21, Kat says to round to nearest TC since her indexes were based on rounding. I assume her frequencies and advantages at various TCs were also based on rounding.
Basically, that's why you round.

She also seems to think rounding somehow gives a better view of things like advantage and index departures in neutral counts compared to flooring. But I got no idea how true that is. Or would be in BJ if BJ players chose to round TC's and had indexes based on rounding.

[David Spence]

> She also seems to think rounding somehow gives a
> better view of things like advantage and index
> departures in neutral counts compared to flooring. But
> I got no idea how true that is. Or would be in BJ if
> BJ players chose to round TC's and had indexes based
> on rounding.

Whether you round or floor hardly matters, as long as it's consistent with the manner of index generation. Truncating is noticeably worse (well, almost :-) ), since it creates too large a bin for 0 (all counts greater than -1 and less than 1 get truncated to 0).

[MHS]

Thanks for your response, Aruuba and David.

[Katarina Walker]

I used rounding in my index generation program. Since you
are using indices from that algorithm, you must also round.
Most Blackjack index generation programs use flooring, so
you should floor your BJ indices.
To be sure, check whether your BJ indices were calculated using flooring or rounding (or truncating).