[600MRG] WSJT-X 1.9 v 1.8 second night

[N1BUG]

Last night 1.9 continued to have a clear advantage but by a much
narrower margin that the first night.

In 12 hours ending 1100z:

LF 178 decodes with 1.9, 174 with 1.8
MF 890 decodes wtih 1.9, 838 with 1.8

Maybe propagation is the reason? Joe Taylor offered this information 
regarding the new WSPR decoder:

...

You may be interested to know a bit more about how this enhancement
works -- and why it's so effective at LF and MF but provides little or
no advantage at HF.

All of the JT modes use continuous-phase frequency-shift keying (CPFSK),
but with the exception of MSK144 the software demoduators estimate
soft-symbol values independently for each symbol interval. In other
words, we generally do not take advantage of the phase continuity that
should exist across symbol boundaries. The new WSPR demodulator treats
received data coherently in blocks up to 3 symbols long. If the radio
channel is stable enough, we find the resulting WSPR sensitivity
threshold is approximately -31.5 dB.

The scheme provides no benefit at HF because even a block size of two
symbols (about 1.4 s) generally exceeds the coherence time of the HF
channel.

...

My parallel 1.9 vs 1.8 test will continue on both LF and MF.

73,
Paul N1BUG