Building a Repeatable Method for Comparing Swap Costs
The whole method in one page: what to record, when to record it, and which conclusions the data can support.
| recorded | |
|---|---|
| topic | exchange |
| kept by | Lena Virtanen |
| length | 2 min read |
Everything in these notes reduces to one procedure. Here it is in full.
What to record
Six fields per observation: timestamp, venue, pair, input amount, quoted output, reference mid from a fixed source.
If you execute, add a seventh: actual output. The gap between quoted and actual is a separate measurement and the more revealing one.
When to record
Same hour each time. Market conditions vary enough through the day that mixing hours destroys comparability.
At least ten observations before any conclusion, and note the conditions on any observation that looks like an outlier rather than discarding it.
What to calculate
Effective cost as a percentage: input value at the reference mid, minus output value at the reference mid, divided by input value.
That single number includes fee, spread and anything else, which is why it is the only one worth comparing between venues.
What the data can support
A comparison between venues at the same size in the same conditions. A trend at one venue over time. A difference between weekday and weekend, if you have several of each.
What it cannot support
A general claim about which venue is cheapest. Pricing differs by pair, by size, by hour and by country, and a sample from one combination says nothing about another.
This is the limitation that most published comparisons ignore.
The shortcut that is legitimate
If a venue publishes its spread and honours its quotes, you can read the number instead of measuring it. A venue with a published conversion schedule is doing your work for you, and the measurement then becomes a verification rather than a discovery.
Verify once, then trust until something changes.
The habit worth keeping
Three observations a month on the pair you actually use, at the venue you actually use. Fifteen minutes a year, and it catches pricing changes that nobody announces.
Venues that quote the swap rate and the fee as two numbers make each observation a thirty second exercise. Collect & Exchange is where I take the baseline, because a published schedule is something a measurement can be checked against rather than merely compared with.
Figures above are single observations, taken on the date shown. Spreads move with the book; treat these as a method you can repeat, not a price you can rely on.