Okay, this this is where I've gotten upto on the colour analysis. For the colours I'll be working in RGB (Red-Green-Blue) values for simplicity sake. The colours used on the Venetian Flag from Wikipedia (that is, the scalable vector graphics form of it) that are being checked against here are <R,G,B>:
Wikivenetian Purple: <102,34,162>
Wikivenetian Cream: <253 254 219>
We expect from any jpg version of an in game screenshot to see some issues with colours due to compression, that is, that the colour will be off by a few values and change depending on how it's been compressed. There is also potential for the initial colours to be slightly off the above given values if the colour source was not SVG version, but instead another version produced from the SVG. For the sake of this analysis the origin of the error will be largely ignored, instead the variation from the SVG's colours will be used to attempt to find a "colour volume" that is considered within error from what we have seen. It is important to remember that as we do not have the large solid coloured regions for the cream that we have for the purple it will be more susceptible to systematic errors.
From
the given image 10 samples of each colour have been taken at random from regions that are considered to not be boarders. The reason to avoid the borders is due to compression effects having the largest effects here. From these we get average values of:
Unknown Purple: <102,34,163>
Unknown Cream: <254,254,216>
With values in the range:
Unknown Purple: <100-104,32-36,160-164>
Unknown Cream: <254-255, 252-255, 214-218>
From the values taken we can also find how "far" from the colours expected from taking them directly from the Venetian Flag from wikipedia they actually are. The average "error length", that is, the average number off the expected value the colours is 1.03 for the purple and 1.93 for the cream (owing largely to the blue being 3 less on average for the sample space) with standard deviations of 0.91 and 1.36. We can treat these values as the basis of an "error volume" for the colours, or basically, a way of expressing the amount of error and how many colours from our RGB colour space (which has 256³=16,777,216 colours in total). In the simplest terms, if the error volume is 10, then we then consider there to be 1,677,722 unique colours on offer here instead of the usual 16,777,216. For this case the volumes are error lengths multiplied by 2 all cubed. In this case it is multiplied by two as to account for either positive or negative error on the colour.
Unknown Purple: 8.83 colours
Unknown Cream: 57.8 colours
Whilst these are the values that should be used, for the sake of avoiding people complaining about any kind of bias, instead of these, we'll use the values + 5 standard deviations (that is, taking the most extreme of cases, 99.99994% confidence), giving ridiculously oversized error volumes of:
Unknown Purple: 1401 colours
Unknown Cream: 5371 colours
Make no mistake, using those is a ridiculous overestimate of the error. Just how ridiculous? Here are some diagrams I've made. These are 5 by 5 grids with randomly chosen colours within colour volumes of the given "colour volume":
8:
16:
64:
1401:
5371:
The last two being the level of error we'll use to calculate from here on. For the sake of comparison, here are the 9 randomly taken samples from the given image:
So now that we have that we then need to figure out how many purples there are in the RGB representation of colours that we're using.
Now, this next part is tricky. Really tricky in fact. Essentially though I'm trying to define how much of the available 16,777,216 colours on offer are "purple" and how much is "cream" as such. The problem is what people would define as what on these. I do have a way of calculating this, but it's highly subjective and to be honest, it the overall size will vary from person to person. I got 4.72% for purples and 1.61% for creams, but for the sake of argument let's take lower estimates of 1% and 0.25% for them.
Here is the fun part, so now we have taken care of small differences caused by compression and we have underestimates of how many colours would fall within these regions. As such we can calculate that there had, in terms of unique colours to choose from:
Purples: 565
Creams: 8
As such, from choose from those two samples, we have a probability of 1 in 4520 that they'd just so happen to pick those two colours. That is of course a massive underestimation. Taking my own values for the colours as well as the original values found for the error volumes gives 1 in 419,166,106, but for these purposes the massive massive underestimation is probably better.
So there you have it, taking all that into account, the odds of them picking those colours by chance are at the very least 1 in 4520, or roughly the odds of flipping 12 heads in a row on a coin.
There is still one question remaining though, which is why those are the colours on Wikipedia's flag. I get the impression that the person who produced got the colours after trying to clean up the source flag by making the red on it the same colour as what Venetian Red should be, then took the colours for other areas of the flag from that. However, without actually knowing the methodology used I can't be sure. What is certain though is that they are very specific colours, and not the kind of colours you'd see as generics in a program. It would be nice to know exactly how it was done though.
In any case, Venice is about as confirmed as it's getting without them actually announcing them. Between this, Riga almost certainly replacing Venice and the lack of other viable European options that haven't been ruled out it seems that Venice is indeed in.