I quipped that it was 'worth every penny', but actually I don't know, since I never tried it.
But it might be worth looking into.
What would really be great would be if Tiger had a contact for 'Lorenzo'. (of yesteryear) He created a 'Jai-Alai Supercomputer' which is better than a sim, in that it calculates the EXACT probabilities for each game. Including the supers.
Despite being an obviously very bright engineer type, he blundered big-time by thinking there would be a market for his product. So I think it was just quietly shelved. If you could buy a copy of that, it would be more accurate than any sim.
However, I have my own 'Kruncher' prog which does the same thing for Spec7 games. So I can tell you for certain that if you have a well-designed sim, and run it at about 1 million cycles, the top 20 tri list will be fairly consistent. And the exactas and Q's even more so.
Therefore, a sim like that it is good enough for any practical research.
You could argue that it is also necessary to have a lot of supporting 'infrastructure', for handling entries, changes, results, doing ratings calculations etc. But not totally true.
IMO, if you used your own judgement - high quality, let's assume, and comparable to 'Sensei' style - to provide sim inputs, you would still get big benefits from letting the sim calculate how those ratings interact. Seriously, even ratings at a chimp level (ooh-ooh, 3 good, 6 sucks) might be worthwhile. Especially if they didn't match the obfavs in the program pages.