What kind of coil geometries would be acceptable to your application, and what L and R parameters would they have?
That is the question, as far as being able to tune it properly.
The two-capacitor network acts as an impedance divider, which when the capacitive reactance is canceled out, also acts as an impedance transformer. Usually there is an inductor in there, as well.
The easiest way to demonstrate this working (given that familiarity with reactances is a prerequisite), is to picture a resonant tank: a capacitor and inductor in parallel.
At resonance, reactances cancel, and the parallel equivalent impedance is infinite (like an open circuit), and the series equivalent impedance is zero (like a short circuit). Suppose we connect a resistor in series with the loop: the series equivalent is then just that resistor (at resonance). Or suppose we connect one in parallel: same thing with the parallel equivalent.
What if we do both series and parallel resistors? We have some choice of which places to put them, which gives the four permutations of L-match networks. Suppose we measure the series equivalent impedance, with a resistor in parallel with the capacitor. That is, a circuit of (port)--L--(C || R)--GND. What is the resistance at resonance? (But first, what is resonance in this circuit?)
As it turns out, the resonant frequency is slightly different, by an amount on the order of -1 / (R^2 C^2). For non-precision cases, and Q > 5 or so, this is negligible, but it is worth noting that the frequency changes. (This is more obvious with exaggerated values, say R < sqrt(L/C), and the transient response rather goes "thump" instead of "dinggg"!)
So that's easy to approximate out, and we're left with the resistance question. As it happens, the R value is inverted relative to the resonant impedance sqrt(L/C). That is, the equivalent series resistance measured is around Req = L / RC.
Consider what happens at resonance: the loop current rises, and the voltage (across the inductor or capacitor) rises. Both rise proportionally, the ratio being the resonant impedance, sqrt(L/C). We can start with a low voltage, apply it in series to an LC tank, and get a large voltage on the resonant node -- "Q multiplication". If we load that resonant node, the Q is spoiled and the voltage drops relatively rapidly, which is to say, it is a high impedance node -- we have transformed the source impedance. It works inversely, starting from a high impedance and making a low impedance, just the same.
This is the basic, qualitative way in which the tuning network operates. To match any impedance, high or low, two L-match networks are glued together back to back; which puts two inductors in series so only one physical component is needed. That leaves two variable capacitors, in effect, one to match the source and the other to match the load.
L-match calculator to play around with:
https://www.daycounter.com/Calculators/L-Matching-Network-Calculator.phtmlTim