Completely Unreliable
Those sites are completely unreliable for unbiased numbers for operating systems. NetMarketShare, for example, has had Linux numbers that vary relatively wildly up and down recently.
Wikimedia stats numbers are much more unbiased and believable. The only issue with them is that Linux growth on the desktop appears relatively flat percentagewise because of the growth of mobile over recent years. If you consider only the percentage of non-mobile Linux use compared to total non-mobile figures you will see that Linux desktop use has grown pretty steadily since the numbers started in 2009, with over a 30 percent increase during that time period. Whether this will continue or not I couldn't tell you, but it appears that Linux desktop use has been rising.
