So you're saying the Columbia County Board of Elections got it wrong? Because that's what I used to make the map.
The last change in ward boundaries was 130 years ago, and that was to split the 5th ward off from the 4th ward. The previous change was to split two wards into four by drawing the east-west boundary down Warren Street. The original division into two wards using 3rd Street was 200 years old. The city charter was enacted and amended by the legislature. In New York, city charters are somewhat like corporate charters, a grant by the state to exercise specific municipal functions. The last (or at least latest) city charters were granted in the 1940s.
I speculate the original division into two wards along 3rd Street, a boundary that lasted for 200 years was in a bill carried by then-state senator Martin Van Buren from nearby Kinderhook.
Despite this lack of change which was reflected in delightfully hand colored atlases from the late 19th Century, the CCBOE had stopped following the actual ward boundaries. They compounded the problem by publishing maps on their web site that were wrong. If a voter relied on the web site map, they could go to the wrong polling location, or find an unexpected set of candidates.
Over time, the populations of the wards diverged greatly. After the OMOV decisions of the 1960s, rather than drawing equal population wards, Hudson implemented weighted voting where the voting weight of aldermen in Common Council decisions was (roughly) proportional to the population of their ward.
They did not bother to update the weights until after the 2000 Census.
At one time, censuses were compiled and reported by ward (this was true throughout the US) but by 2000 this was no longer true. To "assist" cities and states in redistricting, the census bureau had devised the concept of voting tabulation districts (VTD's). These were delineated by state authorities, but conform to census standards. It is a cooperative effort. The state might suggest VTD's, but the Census Bureau might reject them or adjust them. Some states don't have VTD's. In New York, VTD's correspond to voting precincts (election districts).
The Census Bureau requires VTD's to match census blocks (since 2010 they have permitted VTD's to cause the division of census blocks, just as city and county boundaries do). VTD's are simply aggregates of census blocks (the same is true for all other entities that the census bureau compiles data for).
Because of the old rule, Wards 1, 2, and 4 were combined into one VTD; The two precincts in Ward 3, 3-1 and 3-2 were combined into a single VTD, and the two precincts in Ward 5, 5-1 and 5-2 were combined into a single VTD. That is, seven election districts were combined into three VTD's. In addition, because parts of the boundary between wards 4 and 5 did not follow streets, the VTD boundary does not match the ward boundary. In addition, the VTD boundary does not match the ward boundary between wards 3 and 5.
And finally, the CCBOE did not follow the ward boundaries. When the Common Council asked the CCBOE to use the actual legal boundaries for the city's wards, they refused. The council passed a resolution to sue the CCBOE, but that was vetoed by the mayor.
So the CCBOE had a map on their web site that did not follow the legal boundaries; and which they themselves they did not use.
Ward 5 was divided into two precincts. New York law limits election districts to 1000 registered voters. Since Ward 5 exceeded that limit it had been divided into Ward 5-1 and 5-2. It appears that the division was done to balance the voting rolls, rather than a more logical split. But it didn't matter two much since both precincts voted at the same polling location. But the CCBOE map showed the division in more of figurative sense, and the boundary line cut through areas that were actually in 5-3.
After the 2000 Census, the city calculated the ward populations. They used the actual wards in the city charter, and correctly estimated the populations of split census blocks. The ward populations were used to calculate new voting weights. It appears they did not like the results, because some wards share of population had declined well below 1/5 of the population. The Common Council proposed redistricting into five equal population wards and eliminating weighted voting. After the voters rejected the idea, the city finally updated voting weights after 30 years.
After the 2010 Census, the city again attempted to calculate ward populations. They went to the CCBOE, which gave them the wrong map. The city compounded the error by bollixing the division of a census block, putting the larger population in Ward 1, which had a smaller portion of the block, and the smaller population in Ward 2, which had the larger population. This error was racially discriminatory, since Ward 2 is majority minority district, and its population was underestimated, while the wealthy, mostly white, Ward 1 had its population exaggerated.
Because of all the errors, the voting weights were miscalculated. Once this was discovered, Hudson did ... nothing.
In 2016, there was an initiative to make the wards equal population and eliminate weighted voting. This was approved at the November 2016 (presidential) election. The current map on the CCBOE web site, and which your map is based on is from the campaign for this proposal.
The 2016 presidential election used the old wards (and election districts). This can be seen from the relative share of votes cast.
2016: Ward 1: 10%; Ward 2: 18%; Ward 3: 21%; Ward 4: 14%; Ward 5: 37% (ED 5-1 19%; ED 5-2 18%).
Votes cast were roughly proportional to the population of the wards. There is some variation due to socioeconomic factors. There are fewer children per household in some areas, and differences in citizen population (Hudson has a significant Bangladeshi population).
There are also some areas with persons who have 2 homes (or at least 2 residences). They might have an apartment in NYC, and own a house in Hudson. There has been a deliberate effort to get such people to register to vote in Hudson where their (Democratic) vote will have more effect. Voters may be able to choose their residence for election purposes, while the Census Bureau uses a usual residence rule - where your head is planted most nights. These area may produce more votes in even-numbered years for state and federal elections. In the odd-numbered years, these dual-homed voters may not be aware of the local elections, they're not resident there.
The new wards took effect for the 2017 elections, with the new aldermen and supervisors taking office in January 2018. These wards and precincts are being used for this years elections as well as presumably the 2020 elections.
For the 2017 local election one can see that the wards were much more balanced.
2017: Ward 1: 21%; Ward 2: 18%; Ward 3: 18%; Ward 4: 23%; Ward 5: 21%.
Wards 4 and 5 had competitive races for alderman or supervisor. which may have attracted a few more voters. Otherwise voters were mostly voting out of habit. The mayor-elect was an alderman from Ward 1. He was unopposed for mayor, but voters in Ward 1 were much more likely to actually vote for him, rather than skip the race. He may have attracted a few more votes.
Your map has two errors:
(a) The boundary between Ward 2 and Ward 4 is on Rope Alley rather than Robinson Street (between 2nd and 3rd streets).
(b) The boundary between Ward 4 and Ward 5 is an extension of 5th Street between Clinton Street and Harry Howard.