Hi, I wanted to know what is the best way to model a player for a game. Should the player be high poly or low poly ( one player mini game ) and how much detail should i put in the character i model?
if it is fps - you do not even need a player 
In general,
- a model should be as low detailed as possible to reduce performance impacts
- a model should be as high detailed as possible to look good
This to statements do not necessarily exclude each other. It is a balance of both influenced by multiple factors:
Hardware, Game design, Screen size, Model size, Importance to the story/atmosphere, …
I personally would never consider this outdated Possibility, but you also can make the Player a flat prerendered Animation like in MDK.

This would only work with a rather stiff Camera always sticking to the Back of the Player.
The Advantage would be that your Character can be as highpoly as possible and as detailed as possible, 'cause in the End he would be a flat Plane either Way.
Whatever looks the best for the least amount of resorces you have. Focus more detail on the objects you see the most, less so on other objects.
It really depends on your computer’s processing power (if you want to go for high detail), and on your personal preferences.
Some criteria you may want to consider when modelling a player would be the following:
-Do you have a Large scene? (the larger, the fewer polygons a player should have - small influence)
-Is it a “close-up” game or one with a camera positioned further away? (The closer the camera, the more the polygons a player should have - high influence)
-Does your game have many objects/characters? (If yes, less polygons per player - medium to high influence)
-and many other…
If you like the look of the Fifa characters, you may want to limit yourself to a maximum of a few thousand polygons
If you want to add a lot of characters, perhaps a maximum of 1-2 thousand per player would be enough
And finally, if you want a state-of-the-art looking character, a decent (and EXTREMELY high) limit would be 20K (20 Thousand) polygons per player. A maximum of about 10K would be recommended though’…
P.S. Textures (and the way you use them (normal maps, parralax mapping)) along with other effects (like shading) may have a much larger impact on the feel and look of the game than the simple(and not-so-very-important) “number of polygons.”
P.S.2: MDK - one hell of a well made game (in terms of creativity, approaches of using graphics along with the story itself). It’s a pity there aren’t enough games that could even attempt to match its quality(very few can actually do that)…
use Makehuman for high poly models. http://sites.google.com/site/makehumandocs/download
By combining normal maps and baking AO, you can make a low poly model look more like a high poly model. Just make a high poly version for details and a low poly version for the actual engine.
I was wondering about the in game impacts of tangential normal maps?
When doing standard rendering I use them all the time, my character can be as low as 500 faces (usually 2500 with subsurf), but the high res sculpt is usually around 2M faces before I bake them off. But it generally saves 10’s of minutes per character render time…