I don't think there is any structural advantage. This means, the tie-break is in no way designed such that it benefits big servers. As I said, on a case by case basis, it may be that playing more tie-breaks, which big servers tend to do, can give one more practice in that situation and a greater % of tie-breaks won than tie breaks 'expected'. I am interested Sophocles, in what intuition it is that causes you to think TB's could favour big servers. As in regular sets, the tie-break is a 50-50 combination of serving and returning. I don't even understand what the intuition is that makes some people think there is a structural advantage to the server. Sure, it's nice to have a big serve in a tie-break, but to no greater extent than it's nice to have a big serve when one is playing a tennis match.
Duong, we both know how often observation and intuition can lead to faulty beliefs. As for trying to find a more representative statistic, I have an idea, I'm not sure whether or not Jeff (who writes the HeavyTopspin blog) has measured it. In order to check reliability of serve in TB's vs. in non TB sets, what about measuring first serve % on its own? In this case, the disinterest of the returner on some points doesn't factor into it at all. I take it there aren't really any circumstances in which a server purposely misses first serves in a set. This could at least give an accurate test as to whether serving is + or - effected by the tie-break. The problem is, there isn't a way to test the reliability of other shots and make a comparison.
For what it's worth, here's my view. I'm pretty confident there is no structural advantage. Rather, any advantage, will be from having a game style conducive to high performance under pressure. So with the big serve, the debate is, is this conducive to high performance under pressure? The answer may even be, a highly qualified, 'yes, sometimes'. However, as you pointed out Duong, you can say for any game style whether or not it is typically easy (relative to the complete set of game styles) to perform under pressure. In this case, the lower risk game styles are usually supposed to be 'easier' (so look at Murray, Nadal, Simon but according to your 'intuition', also big servers?). This doesn't mean we should expect all players of style X to excel in TB's while all players of style Y should fail, only that there is some sense in which it is 'easier' to excel as a player of style X. Whether or not 'big server' is a style that could function as style X in this example is essentially what we have been discussing.
IMO, the best way to try and test (although this could easily be inconclusive), would be to gather large data in which players were sorted into categories of playing styles (obviously subjective but could probably be done to a reasonably effective degree). Then, using Jeff's stat 'TBOR', see if there are any generalisations to be made. TBOR = Tie Break Out-Performance Rate, which is, for any player, 'tie-breaks over expectation' divided by 'total tie-breaks played' ('tie-breaks over expectation' is derived from % of serve and return points won in the match in question). My hunch is that, should this task be completed, there would be no meaningful patterns at all. Yes, some styles are easier to play under pressure, but by such small amounts that individual differences far overshadow the influence it has. Although the sample is too small to be conclusive, this is the top 20, in order, in TBOR this season up to Shanghai, to me, this as about as random a list (in terms of game style) as can be:
Steve Darcis Jurgen Melzer Andy Murray John Isner Tommy Haas Kevin Anderson Janko Tipsarevic David Ferrer Pablo Andujar Julien Benneteau Radek Stepanek Sam Querrey Andy Roddick Jarkko Nieminen Paul Henri Mathieu Andreas Seppi Jeremy Chardy Philipp Kohlschreiber Denis Istomin