Bsp is the most common error, but compasses are also a good source of error. The office Maretron SSC200 changes by 2.5-3 degrees when rolled to 20 degrees.

Also SOG depends a lot on the GPS chipset used. Decent ones will estimate cog & sog from the same filter as the position, but about an order of magnitude precision higher.

Some cheap, generally older, GPS chipsets just use the difference between sucessive position.

A fresh start on Magnus' questions and the first answers that he received, and some additional thoughts.

1. I am not sure that Instrument systems actually apply a Leeway correction to the Tide Set and Rate calculations. Instead, they tend to bundle the two... which creates strange Current readings in certain situations (low current - or no current as in Magnus' case). Please let me know what the facts are on this for various systems.
2. When it comes to Wind calculations, my understanding is that the Tide component is ignored, except for angles. On the H5000 system, applying Leeway amounts to taking Course as angle reference instead of Heading. But there is no correction to the magnitude of True Wind or Ground Wind to reflect the Leeway component.

I am pretty sure of these answers for the B&G H5000 system (but always ready for surprises - I haven't seen the code!!). Why would that be? Why not "do things right" and include Leeway properly in calculations?

One issue in using Leeway to correct for Current or Wind quantitatively is that the way we generally get to Leeway is rather imprecise for the accuracy we would need in these corrections. Using an empirical formula (usually k x Heel / BS^2) with a fixed factor is fine for navigation to correct courses under certain limitations, but it does not provide the level of detail required to perform the instantaneous corrections on Current and Wind, even with damped data.

One way to check this is to get a Leeway table from the naval architects of your boat, together with the optimum Heel and of course the polar curves. This allows you to back-calculated k factors for each pair of TWA and TWS...
In the only example I was able to run (only complete data set I was able to get), k factors vary between 5.2 and 17.7 - and I am only limiting the calculation to TWA less than 75 deg to make the point. Now, assume you had applied a reasonable value of, say 10 for the k factor and see where this gets you. The differences between "True" and "Calculated" Leeway varies in this case between -2.4 and 2 deg., which put differently means -90% to +45%. Clearly not something that can be exploited for our quantitative purpose.

On the bright side, the average k I got over the range of TWS and TWA (to 75 deg.) is... 9.46!

So where do we go from here?

Best would be to MEASURE Leeway. The new 2D speed sensors may provide a solution, I have no experience with them but would love to test a few...
Until then, we could use tables of Heel and Leeway to complement the polar tables and maybe calculate our Boat velocities and Wind speeds more accurately... Of course, this would need to be implemented in the Instrument system, or in a navigation software such as Expedition. And the architects should be forthcoming in providing the data, which I have found pretty impossible.

In the absence of these two, I stopped analysing Tide data to refine Boat Speed and Leeway calibration. I have put all my effort in getting very accurate Boat Speed and Heading, and I check Tide and Leeway in other ways (high resolution current forecasts, cross-track error etc.) for consistency. By the way, working this out helped me understand why B&G recommends turning the Leeway correction off (which is just the angular part) when using the Autopilot.

And then of course, you could be on Magnus' waters with no current and do the perfect Leeway calculation - for each pair of TWA and TWS!

One obvious thing to check is of course that leeway is calculated in the right direction. If signs from heel or wind angles does not follow the normal convention, leeway could easily become inverted to what would be expected.

The easiest way to check this is to time plot heading and course when sailing upwind and ensure that the difference between the headings are smaller than between the two courses. If headings and courses are on either side or North, one tack has to be unwrapped obviously.

There is an option to use leeway from the sensor in the Exp calculations options.

Yes - definitely check the signs of heel/roll and leeway are correct. If you mount the heel sensor backwards, it may be wrong ....

Current set & drift in Exp is calculated from cog, sog, bsp and course so of "course" it includes leeway.