I read that one of the ways to make AB more efficiant is to first search one level, then search two levels but first checking the node that had the highest score on the fist check, then check three... untill you check the amount you want (10). My question is, does the ID help you on the first level (by checking the bet node first) or should I store the best nodes of the other levels too? Thanks.

Normally if you implement Iterative Deepening, you keep track of the best line of play on each iteration. When you re-search, you search the best line first.

So, short answer: No, you don't limit the search ordering to the first ply only.

If you have transposition tables implemented, you can also use that to see if you have a 'best' move for ALL nodes, and search those first too.

Will