Well, another trakt.tv statistics “smorgasm”, if you check the stats for season 1 episodes you will see how trakt stats are becoming garbage and with them most of it not all of its rankings and charts. H
ow likely do you think is for any tv show to suffer extreme volatility in views between back to back episodes and experience a total inconsistent relationship between downloads and views? The probability of occurring, as it's not a random event, is so low that is equivalent to 0, or impossible in the real world.

Let see what we have in this today's Top show in the top watched TV Shows chart:

**Episode. Dowloaded. Watched**
1x06 23.4k 76.3k
1x07 10.6k 9.3k
1x08 11.1k 76.9k
1x09 11.0k 8.3k
1x10 9.3k 72.9k

Somebody could conclude that as the show loses interest (the number of downloads halves from > 20k to barely 11k) it's audience is constant (by times watched) except for odd numbered episodes, people really don't like odd numbers because the audience it's just a little over 10% of an even episode. That could be the conclusion from a nut head.

if you think about it, only the odd episodes data seems to be coherent, if for episodie 1x07 there were 10.6k downloads of the episode (had the intention of viewing it) but only 7k watchers viewed it a total of 9.3k times, a conversion rate of 66% and 32% viewed it twice, these numbers pretty good specially taking into account that hoarding content doesn't carry a direct cost and that is more common for soft players to get stuck or crash at the end of a video reporting a complete watch and when the user tries to replay the last minutes reports another complete watch.

On the other hand for even episodes the data doesn't make sense, if for episode 1x08 there were just 11.1k downloads of the episode (had the intention of viewing it) but a massive 55.4k watchers viewed it a total of 73.9k times, a conversion rate of 486% and 33% viewed it twice). There are five times more people that watched than those who got the episodes, each copy dowloded was shared/watched by 5 persons at different times (more than twice than the people in the average western household - which would mean that every one in your family viewed independently and then you invited the neighbors one by one to watch it independently). Now, more realistically if our conversion rate from 1x07 was just 66% (as the series tanks this falls as you can check with 1x09 where conversion rate had lost ten points reaching 56%), which tells me, the real watchers should have been around 7.3k and not 55.4k and viewed just 9.7k and not 73.9k.

I think someone from the trakt team should start tracking where they are doing the math wrong because although it's not like we are asking for the precision calculations required to launch rockets into orbit and bring back their main parts intact back to earth, but to have some more or less coherent and reliable statistics about the movies and Tv Shows we watch, an EECS from MIT is not required as long as you know basic algebra, basic statistics and some client/server programming and have a little common sense.

It is time you start reviewing your code base, something has been stinking too long and as ex-vip everything I don't understand how people still pay (probably don't know) if the most important one and basic flaw, the soundness of the data, hasn't been addressed.

As the CS saying goes "Garbage in, Garbage out". Please, get your act together and do something to fix it or break it completely, but do something other than contemplating.