Solution:
We first prove the following lemma.
LEMMA. If in the (t;a,b)-game some of the two players has a winning strategy, then in the (t+a+b,a,b)-game the same player has a winning strategy.
Proof of the lemma. Denote the players by A and B and let B have a winning strategy for the (t;a,b)-game. In the (t+a+b;a,b)-game after the first move of A we obtain either the (t+a;a,b)-game or the (t+b;a,b)-game with B to go first. In both cases B can get the (t;a,b)-game with A as first player in which case B has a winning strategy.
Let us now assume that A has a winning strategy for the (t;a,b)-game. Then after the first move of A we obtain either the (t−a;a,b)-game or the (t−b;a,b)-game with B to go first. Therefore the second player has a winning strategy for some of these games.
We consider (without loss of generality) the case of the (t−a;a,b)-game. It follows from the above that the second player has a winning strategy for the (t−a+a+b=t+b;a,b)-game. Since A can obtain the (t+b;a,b)-game with B to go first from the (t+a+b;a,b)-game, it follows that A has a winning strategy for the (t+a+b;a,b)-game. This completes the proof of the lemma.
We now prove that for t=2004 and a+b=2005 the first player A has a winning strategy. We may assume that a≤b. Since a>0, we have b≤2004. Then A subtracts b from t=2004. The resulting number 2004−b is less than a since a+b=2005. This means that any move of B leads to a negative number. Now the lemma implies that A has a winning strategy for the (t,a,b)-game for every t≡2004(mod2005) and a+b=2005.