Most information on storms that long ago comes from times when they passed over some landmass or in the vicinity of a ship at sea. You'll notice that there are very few storms way out at sea in the historical database of storms in those times (good example -- 1933, with 21 named storms in total -- http://weather.unisys.com/hurricane/atlantic/1933/index.html), while there are also quite a few storms with only a couple of points near land -- as no one knew that a storm was even out there at the time. Thus, the majority of observations potentially underestimate a storm's true peak intensity, and it is likely that there were more storms out there that were never tracked in the past.

The intensity of these storms generally is determined from recorded historical accounts, land- or sea-based measurement systems (wind speeds in general, occasionally barometric pressure), and pictoral recordings of damage for those storms that made landfall. Some years in the historical database are more complete than others, with efforts underway (and complete for some time periods) to update and revise the historical database -- see the Hurricane Reanalysis Project for one such example. I think they are into the early 1900s now, as evidenced by a decent number of Cape Verde storms in the historical record for the late 1800s/early 1900s.