If they could have done a bowling game or soccer game with the PSEye, I'm sure SCE would have done it.

HV Bowling would have been an obvious choice, but since they opted for Six Axis control, one could only guess that either they didn't want players to have to use the camera, or that it just wasn't feasible. If it was the former, they could have made PSEye control optional or even supplemental, so I'm guessing it was the latter.

Never bought the EyeToy, but I thought they had at least one game that involved smacking a ball with your hands, even if it wasn't sensing Z axis motion from the player. 

"Kick ball" games in general are probably not the best way to showcase Kinect's capabilities anyway, even if it is one of the more obvious examples.

I think a decent boxing simulator would tell how well Kinect really handles 1:1 depth motion since distance and reach are key even for basic, but realistic reflex mitt drills, meaning the game has to at least differentiate between a close range uppercut, a medium range hook or a long range jab or straight punch. Without working depth perception, the game is simplified to "touch the pixels when they appear" with no regard to distance in the Z axis.

It's actually 48 points or 20 joints. There seems to be some confusion among some people as to the difference. There are more calculations involved in tracking articulating joints than simple points. Generally, you're tracking 3 vertices on the X,Y and Z axis to accurately track one joint.

If the software does a good job of approximating user positioning and changes over time, even 20 joints is enough to represent a fair approximation of full body movement. Just not full body movement accurate to individual digits and precision facial recognition (approximating expressions by tracking points on the face).

Of course one really can't tell how accurate Kinect really feels without doing a full objective break down of the initial games, which may not even be the most fair assessment of what the system is capable of doing.

Thats what I meant by precision.

sorry say again?

so one joint has 3 axis for 3 points per joint?  i'm confused as 20 x 3 = 60 points?  can we track some joints with only 2 axis?  i tried googling a few sources but everything seems to point back to the 20 joints thing.

This is pointless until they both come out. It also is pointless until we can all see a wide diveristy in game selections for both units.

A functional joint in a 3D model rig (the underlying system that controls the mesh, or polygon geometry) is essentially 3 vertices or points structured in a hierarchal sequence (ex. hip joint, knee joint, ankle joint).

A joint in this particular sequence, say the knee, maintains a relation with the hip joint above and the ankle joint below (distance, which would represent the length of the femur or the length of the tibia/fibia, which stays constant). If you can envision the knee being the central pivot point, then the hip joint and the ankle joint would represent either ends of this basic joint chain.

The basic concept is that a joint isn't just a simple point floating in X,Y,Z space; it has to maintain its relation to the joint above and below in the chain. Additionally, that joint will have pre-determined restrictions (a knee joint for example can only pivot, and only on one plane, and not past full extension).

A simple joint layout could go something like: balls of feet (2) heels (2) ankles (2) knees (2) hips (2) spine (5) neck (1) clavicles (2) shoulders (2) elbows (2) wrists (2).

That's 24 joints right there with no hand/finger joints.

Even with an extremely limited spine (robot spine with 1 or 2 joints), one can see how there aren't enough joints for finger sensing as each hand will generally have about 15 joints.

The weight of the bowling ball is key as well.  So is the movement of the body towards the lane.  Wrist movement generates spin in real bowling.  Kinect Sports uses arm positioning to generate spin.  Still a virtual game of bowling.

They are most definitely not using the depth map for facial recognition.

And no, there is no finger recognition from Kinect.

kinect is completely directed toward a family audience) and move trys to do both family and core gamers. Kinect wins the casual market with its family orirented games while I think core gamers and some families will be quite satisfied with the Move with its vast categories of games. This could also change in the next few weeks or months, but as far as known titles that will be released, i think move wins with quality, even at the 60 dollar higher price mark