Cursor alignment with data points on scatter plot chart is offset

[Guest]

I have noticed that the selection cursor, when hovering over a scatter plot
in the Windows version of Excel (2003 and 2007 versions), will display the
value of the data point the cursor is supposedly pointing at. This is a
handy feature, however, there appears to be a bug that directly impacts
accuracy of measurements taken from charts when using this feature.

I wrote "supposedly," because the data values displayed when hovering the
mouse over a data set in the chart are for a datapoint about 5-8 pixels to
the right, not the one at the tip of the arrow! This is an offset effect
that is peculiar to Excel charts.

I have tried this on several machines with different graphics cards, all
running WIndiows XP SP2 and two versions of Excel (2003 and 2007); all with
the same results. I have also tried it using a freeware utility called MB
Ruler, which provides a loupe to see the cursor hotspot magnified, so you can
see the effect very clearly.

The offset is not there for selecting cells in the worksheet or using other
features of the windows desktop, such as task bar buttons; it only happens
with the Excel charts. It appears that the charts, as displayed on a Windows
machine, are actually the thing that is offset slightly from where the
program thinks they are.

This problem is reportedly not present on the Excel 2004 for the Macintosh.

This problem has been verified to exist with other users and there does not
apper to be a solution for it. Perhaps MS needs to fix it.

 

[Jon Peltier]

I just tried to replicate this problem in 2003, and I couldn't. I made an XY
chart with around 50 points, with X and Y both determined by =RAND(). I
shrunk the chart so 0 to 1 on the Y axis was 1/2 inch, to get a lot of
points aligned or nearly aligned horizontally. Whenever I hovered over any
point, I got the chart tip for that point, without any offset that I could
discern. I don't think it happens in 2007, either, because I'm sure I would
have noticed. Maybe in one of the earlier betas, but not in the later stages
or in the commercial version.

- Jon

 

[Guest]

The reason that does not replicate the problem is probably because the RND
function points tend to be too far apart. This does happen on smoothly
varying functions. Try it by placing your Xs from 0 to 10 in 0.01 increments
and set the y's to SIN(PI()*X). Viewing the resultant chart at 400% zoom
near one of the SIN wave peaks is the best way to see the chart display
offset problem happen.

My versions are all the enterprise versions - fully patched.
--
Bas V

I just tried to replicate this problem in 2003, and I couldn't. I made an XY
chart with around 50 points, with X and Y both determined by =RAND(). I
shrunk the chart so 0 to 1 on the Y axis was 1/2 inch, to get a lot of
points aligned or nearly aligned horizontally. Whenever I hovered over any
point, I got the chart tip for that point, without any offset that I could
discern. I don't think it happens in 2007, either, because I'm sure I would
have noticed. Maybe in one of the earlier betas, but not in the later stages
or in the commercial version.

- Jon

 

[Jon Peltier]

Okay, I see a bit of an offset at 400% zoom, maybe four or five pixels. I
used a line without markers, because each marker overlapped with six or
eight other markers.

Excel isn't a graphic design program, and in general it behaves best if the
zoom remains at 100%. I recreated my chart, and duplicated the resolution in
two ways, first my tweaking the axis scales, then by resetting the axes and
making the chart 14 inches tall and twenty-something inches wide. In both
tests, the chart tips showed information for the point the cursor was
directly over.

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
Peltier Technical Services, Inc. - http://PeltierTech.com
_______


Excel works best graphically when the zoom is kept at 100%

The reason that does not replicate the problem is probably because the RND
function points tend to be too far apart. This does happen on smoothly
varying functions. Try it by placing your Xs from 0 to 10 in 0.01
increments
and set the y's to SIN(PI()*X). Viewing the resultant chart at 400% zoom
near one of the SIN wave peaks is the best way to see the chart display
offset problem happen.

My versions are all the enterprise versions - fully patched.