Rapsodo Pitch Tracking - Focus on location accuracy

We have been beta testing the Rapsodo unit for several months. Following is our evaluation of the location accuracy - using the most recent hardware, firmware and software releases.  (same version tested at Driveline).​

Interested in pitch tracking and the issues experienced
​in our beta testing? Check out Kyle Boddy's (Driveline) post HERE>
​where he compares Rapsodo with their Trackman unit.

How accurate is it?
We used our Hack Attack Pitching Machine and HitTrax system to provide the input and comparison measurements for Rapsodo. HitTrax has 3 hi-speed cameras to track the flight of the baseball on the way in (pitch) and way out (hit) using the triangulation method to determine where the baseball is.

Hi-Speed Video analysis has proven HitTrax to be accurate in reporting the location of the pitch. We will use this knowledge in our testing here and not re-confirm with video. Ten 77mph fastballs from the Hack Attack pitching machine were thrown and measured by each system. We have measured several hundred pitches of various types and spot checked the location on many which indicated the unit was in the ballpark but the focus was on determining if the other aspects of the unit were viable (spin rate, spin axis, true spin and movement). This small sample was used to put concrete numbers on location accuracy.

We have a couple of issues with the data that need to be discussed before looking at the results.
First:   the strike zone itself and dimensions/location. We have no way to align both systems to have the exact same center point location of the strike zone for vertical and horizontal directions or size of the zone. Therefore, pictorial representations may be off.
Second:  each system reports the location as above/below and left/right of a reference point. This reference point is not the same for each system. For HitTrax it’s from the middle/middle location of the strike zone. For Rapsodo, it’s the middle of the strike zone horizontally and appears to be from the bottom of the strike zone vertically (vertical reference point couldn’t be confirmed as of publication).

To address these issues, we used the first pitch as our primary reference point and then measured how far away each pitch was from that reference point. By doing this we are measuring how much each pitch is different from the original pitch for location as measured by each system. We can then compare the two systems to each other as we are looking at the relative relationship of one pitch to the next.

The Results
The first 4 columns of data show the difference in location from the first pitch for both systems in the vertical and horizontal directions.
The last 2 columns show the systems compared to each other.  ​

The Rapsodo differed from HitTrax on average 1.3” horizontally and 1.8” vertically. The diameter of a baseball is just under 3” so the difference in location is roughly half of a baseball. However, a couple of the pitches differed by more than a baseball vertically.

Only using one camera as compared to three, we didn’t expect the Rapsodo to be as accurate as HitTrax. It turns out to be more accurate
​than we expected. Having the catcher’s view with the camera certainly proves to be helpful in this regard.

Here are a couple of pictures from the two systems showing the location of the ten pitches. As you can see, visually they are very similar.
The differences of exact location with respect to the strike zone can be attributed to calibration (user dependent) and the  actual dimensions of the different strike zones. The HitTrax strike zone dimensions in this case are 18” wide by 16.4” tall (varies by hitter). I don’t have the Rapsodo dimensions as of this posting.

HitTrax (shows hit speed, since we didn’t swing it all values are zero) ​​

Rapsodo

Conclusion
Is half a baseball close enough? Well, that’s really up to you. An error of 1.5" in this environment is reasonable in both a training or game situation; most umpires are not this good. 

I would just like to reiterate some of the statements that Kyle made in his post. Rapsodo has been extremely responsive during all of our testing and very appreciative of the feedback. They are a pleasure to work with are very much interested in providing a product that is accurate and reliable. It’s not without it’s issues but I’m confident that not only will they be resolved but new enhancements and features will be added. 

Teams and facilities should strongly consider adding this to their programs if they want to provide objective feedback to their players
to aid in their development.

Disclosure: we were provided a unit free of charge in exchange for beta testing and providing feedback to help them bring their product to market.