Tournament Performance Rating

Can anyone corroborate Giri's TPR at Reggio Emilia or explain the differences?

2820 at http://www.chessbase.com/newsdetail.asp?newsid=7819

2821 at http://paxmans.net/performance_calc.php?

2830 at http://www.kosteniuk.com/EloCalc/elo.php

I calculate 2822

Average Opponent Rating;  (2762 + 2758 + 2727 + 2775 + 2729) / 5 = 2750.2

Performance Ratio: 6/10 = .6

Performance Change: +72 (source: http://www.fide.com/fide/handbook.html?id=58&view=article)

TPR: 2750 + 72 = 2822

So what do you say is the TPR?

Kosteniuk is apparently using the algorithm of 400, which is a simplification of the FIDE system that you're using. However, the FIDE system is a simplification of the underlying statistical distributions. Direct calculation from the underlying distributions is what I expect paxmans.net is using from their description, but it's not clear (they don't give details and I can't make sense of their javascript). Chessbase I have no clue, because they give no indication how they arrived at their number.

Great. The algorithm of 400 is easier for humans but what are computers for?

Performance rating = [(Total of opponents' ratings + 400 * (Wins - Losses)) / Games].

(2762 + 2758 + 2727 + 2775 + 2729) * 2 games against each opponent = 27502

27502 + 400 (4 - 2) = 28302

28302/10 = 2830