I'm not going to say I'm that elusive guy who makes money on the trifectas but I have some interesting results in that direction and an opinion rather than an explanation as to why that may be.
I have a computerised rating system that makes it's own ratings from past results and returns the field of each race in order of merit. It's certainly not backfitted as it's been constructed on a limited period of results from one state only and works pretty consistently across all states results. It's not (intentionally) slanted towards exotics either as no consideration of them was made when the ratings system was created.
If (as it runs through 10 years of past results) it makes a bet on every predicted winner of every race it comes out about even, so it just overcomes the rake on the NSW win pool. If it makes a bet on every exacta it comes out about 20% ahead and if it makes a bet on every trifecta it comes out about 30% ahead. This indicates to me that with the same level of skill in picking a horse's merits it is possible to do better on the exotics than the win pool. Having said that let me say this, the runs of outs and drawdowns are truely scarey and although it always seems to recover any lost ground (eventually) I don't actually bet them purely because I haven't got the nerve.
Now why this is so I'm not entirely sure, and you jfc would probably have more of a mathmatical idea than I but I reason along these lines of logic.
The results on the win bets indicate that the ratings are better than the average punter (including mug money, wise money, bots, droids, whatever.) by an amount roughly equal to the win rake (what's that? 14%?). When you are betting trifectas does this mean you are better than the average by (1.14 x 1.14 x 1.14) because you are making three selections and using your skill level three times? If this is so that would make you almost 1.5 times better than the masses. The maths would doubtless be more complex but this simplistic view will do to get the idea across. Now if this is true and the rake is 21% this would come close to explaining why the figures give me a 30% lead over the TAB on trifectas.
KV