samtools BAQ difference between 0.1.18 and 1.2 version

[mygump]

Hi, there
I compare the mpileup outputs between samtools 1.2 and 0.1.18. And I find
some discordance of the base quality between the old and new version.
I extract the pileup file with -B option to disable BAQ of samtools firstly. And I get a position like this:
gi|57116681|ref|NC_000962.2| 104830 C 23 AAAAAAAAAAAAAAaAAAAAAA^IA BBBB?@@ABBB@=@?C>ABBCB;

Then I perform the same command except the -B option, but for 0.1.18, I get this:
gi|57116681|ref|NC_000962.2| 104830 C 23 AAAAAAAAAAAAAAaAAAAAAA^IA <<<<>>:=;<<5=@;:>==;::'
for 1.2:
gi|57116681|ref|NC_000962.2| 104830 C 23 AAAAAAAAAAAAAAaAAAAAAA^IA BBBB?@@ABBB@=@?C>ABBCB'

There is a huge difference of the base quality between these two version. The old version seems change base quality a lot.
I failed to find such information through google. I am confused about this. Did I miss something?
Thanks

[Brian Bushnell]

Different algorithms will give different results; I don't really see what's surprising. Specifically, probability calculations are highly variable because they are subjective; so, it's not unlikely that different versions of a program that does probabilistic calculations will give vastly different results.

For the record, my company still uses samtools 0.1.19 due to problems with more recent versions.

[mygump]

Hi, Brian
Do you mean different algorithms have been used by these two version?
The different base quality will influence the downstream SNPs analysis. If I perform base quality filter >20 and strand filter with Varscan2. I will get the SNP at this position from old version but not 1.2 version. Is it normal?

Thanks!