The method you're trying to apply does not seem to account for ties.
You see, in tournaments, ties are a distinct outcome.
For example, 3 players have a 1000 stack.
Player 1 goes all in for 1000, player 2 folds and player 3 calls.
There's now 3 distinct outcomes:
- player 1 wins (chip distribution: 2000-1000-0)
- player 2 wins (chip distribution: 0-1000-2000)
- there's a tie (chip distribution: 1000-1000-1000)
Now in cash games the tie outcome does not affect the math in any way.
You can just use the equity as the chance for a player winning.
However, in tournament calculations you can't just use equity as the chance for a player winning. The 2~ish % of the cases that there's a tie will slightly screw up the results by a few percent.
I'm pretty sure that this is the cause of the discrepancy.