It depends. If you weren't planning on voting, then a vote for a 3rd party is not a vote for trump. But if you were going to vote for Clinton and now are voting for a 3rd party, then you are reducing the threshold of votes Trump needs to win by 1, which is effectively a vote for Trump, mathematically. Switching from Clinton to Trump gives Trump a net of 2 votes since you are depriving Clinton of 1 and giving 1 to Trump.
But it all depends on your original intentions.
i mean if you want to get formal theoretic about it, you can create a utility function (utility functions are basically scores of how happy we are with a choice and how we compare choices we face):
https://en.wikipedia.org/wiki/Calculus_of_voting
Benefit of voting = (Difference in utility between a Clinton presidency and a Trump presidency) * (probability that you cast the tie-breaking vote that elects Clinton) - Costs of voting + Consumption value of voting
Costs of voting are the time and effort associated with voting, which are very small but non-zero. Consumption value is the good feeling you get from voting, telling people you voted, etc even if your candidate loses.
A vote for Clinton is:
V = pB - C + D
A vote for a third party is:
V' = 0 - C + D + D'
The 0 here is because there is a 0 probability a third party will win, so we needn't compare the relative magnitude of the probability or our preference for the third party to Clinton. D' here is the added consumption value of voting your conscience, helping third parties get funding, etc. Maybe you have a preference for the third party but don't feel any better voting for them, in which case D' is 0--but I think most people feel better about voting for candidates they prefer.
Someone who supporters a third party should vote third party when:
V > V', which happens if and only if D' > pB
They should vote Clinton when:
V' > V, which happens if and only if pB > D'
We can't really parameterize utility units, so B and D' are left unspecified to be specified by the voter. We
can parameterize p. The likelihood that you're the deciding vote is either 0 in a non swing-state (in which case this is a moot discussion--if you're mad that a Californian votes for Johnson or Stein, you're just wasting breath) or 1/n in a swing state, where n is the number of voters in that state. Let us take Florida as the decisive state but allow that in the unlikely scenario that Trump loses Florida but runs the gamut on other swing states, he'd win. In Florida in 2012 there were 8.3 million voters. So p = 1/8,300,000 or 1 * 10^-7, which rounds to 0. The smallest swing-ish state is maybe Ohio, which had 5.4 million voters, so p = 2 * 10^-7, which rounds to 0.
Let us grant that B is unusually large in this election, because Trump would make an unusually bad president. But are the units of p and D' so large that they are able to make up for the tiny magnitude of p? That seems a little dramatic, even if you are a member of a group targeted for animus by Trump. We can probably make inferences about utility magnitude by considering how people respond to other very unlikely events. Should you not swim in a swimming pool during a lightning storm? You should not. But should you wake up every day scared that you might die in a lightning strike? Do you build your life around lightning strikes? Around Zika? Around dying in a car crash? Probably not, right. So that implies something about our utility calculus; the magnitude of our global maximum B, the difference between life and death, is fairly small compared with the magnitude of p.
So I would argue that a formal model of voting would suggest that someone who has even a small preference of voting for a third party should probably vote for a third party... if you're thinking "what the hell is this rabbit hole he's going down with this fancy math and Wikipedia links", then maybe it was a mistake to make this a numbers game instead of just discussing it conceptually.
Overall, I think it is the case that people worry a great deal too much about third party voters. First, most third party voters are basically ideological schizophrenics; analysis of Nader voter downballot activity in Florida suggests that Nader drew votes from both presidential candidates (perhaps Gore a little more, but a lot less than most people thought)--setting aside the fact that it's not clear Gore actually "lost" Florida (i.e. if a slightly different count was used, we wouldn't be blaming Nader voters for costing Gore the presidency, another forum would be blaming Nader voters for costing Bush the presidency); analysis of Perot voters using public opinion data was I believe similarly inconclusive. Most of the people threatening to vote third party won't vote at all. Most of them who do vote will come back to one of the major two parties. That's without getting into the argument that no one owes anyone a duty of any particular vote, and if a third party somehow bleeds off 10 or 20% it's probably due to a big structural failure to court voters--which does deserve some consideration even if you also resent the electoral result. Also, as we saw from the morning after "I had no idea what I was voting for" Brexit people, if you want to blame someone, probably start with the people who literally have no idea where or who they are rather than the people who just have preferences you disagree with.
I agree that it is worth asking third party voters to consider a strategic vote, and in a world where just this forum thread were voting it'd make a lot of sense to expect compromise. But given the number of voters in any state of a US election I think it's foolish to worry too much about this problem in any large scale.