Yep, as you gain experience you will become better at knowing implicitly which figures to use.
Ask yourself what a particular ratio is trying to tell you about a company. A "return on capital invested" figure can tell you a couple of different things. You need to understand a company's balance sheet, and how they got it to where it is to understand this figure properly.
For an example, in Australia we have a financial institution called Suncorp (ASX:SUN) that purchased a large general insurance business just before the GFC. It paid a peak multiple on peak cyclical earnings, issuing a significant amount of new equity at the time to do it. The value of this acquisition sits at its historical cost in Suncorp's balance sheet (unless they ever write it down), and given earnings have fallen, detracts from its "return on capital invested".
At the other end of the spectrum, you may have a conservative IT company like Technology One (ASX:TNE) that has essentially been run as a private company (ie - minimise tax by expensing as many costs as possible rather than capitalising them). This keeps the balance sheet small, and "returns on capital invested" high. If TNE ever goes and acquires a peer, your calculation for ROCI would likely fall.
To further complicate things (!), you need to understand what returns a company can be expected to generate on a marginal dollar of capital invested (this figure is arguably of more value than the vanilla stat, but difficult to quantify). Sometimes companies don't 'need' as much capital as they have, so they would be best served returning it to shareholders.
The point I am trying to get across is you shouldn't get too hung up on ratios like these, rather you should understand what they tell you, and their inherent limitations. There is usually limited value in making your ratios 'more accurate', ie - it is unlikely to help you make more money.
May I suggest a good one to start with is EV/EBIT.
The numerator EV is Enterprise Value, and is calculated as market capitalisation (number of shares on issue x current share price) + net debt (long term debt + short term debt - cash). This reflects the price you as a minority shareholder will have to pay for the "capital employed" and also gets rid of the nuances surrounding a firm's equity figure as reported on its balance sheet.
The denominator EBIT is usually quite easy to calculate (pretax profit plus net interest expense if it is not reported explicitly).
The value in an EV/EBIT multiple is it normalises out the impact of a firm's capital structure. That is, it is total earnings before any money is distributed to debt or equity holders. I would argue that the inverse of your EV/EBIT multiple is probably a more useful proxy for return on capital invested than the others you mentioned.
As always, there is always a few ifs and buts. In the case of an EV/EBIT multiple, banks do not have them (a lot of their earnings come from net interest received, so you are better served using a PE ratio). Also, do you want to know average EV for a financial period, the EV at the balance date, etc, etc. For some businesses an EBITDA figure is of more use, especially those that have big depreciation bills and little capital expenditure requirements. You may see a theme emerging?
The biggest limitation is that by looking at these figures, you are investing by looking in the rear view mirror, that is, watching stuff that has already happened. The 'top investors' you refer to are probably more interested in all the factors that will affect the EV/EBIT multiple investors will be willing to pay 3 years from now, and what that implies for the price being paid for shares today!