I have repeatedly suggested incorporating DISTANCE from the price feed as a factor in the reward.
If something is worth $4 what value is there for bids of $3? Granted the ask of $3.1 has value, but if the bid is for 100000 STEEM at $3 and ask $3.1 with another 100000 STEEM, then regardless of what the theoretical value is, the price will be $3/$3.1

The old points system did not take into account value provided properly and it awarded points regardless of the risk taken. What risk is there to selftrade 100000 STEEM over a short period? Adding duration is alway gameable with enough capital. Just put "corks" as the inside spread and the aging ones just outside. Wait for time to pass and wash all the trades, make tens of millions of points.

DURATION is risk, so the more DURATION the more points makes sense to some degree. An order on the books for 3 seconds is not much risk, especially as a washtrade can be queued up right behind it, so clearly for a 3 second duration, the reward should be low. But a 3 day duration that is far from the market price is also not taking risk. So DURATION by itself is not enough.

I am sure @theoretical could make some fancy calculable function of DISTANCE and DURATION that is much, much better than the old method. Simplistically it would be DURATION/DISTANCE integrated over time. But I am sure having it purely linear like that is too simple and would have some gameable aspects.

So we add a third factor ORDERSIZE, I mean what benefit is a 100000 STEEM bid at any price? There is barely enough SBD in existence to fill that amount! With the vast disparity between the SBD and STEEM that is available extremes are being used.

So one solution is to limit the size of bids and asks that qualify and also rate limit the number of orders that can be done (otherwise we will get rapid fire 100,000 orders per minute just so people get the points).

And that gets me back to my idea of having both an automated rewards system and a human moderated one. It is pretty hard to be making a market without being noticed, so like http://steemprices.com/market/ does, if most of the volumes can be properly tagged as to who is controlling the order, then we can actually do a good job of determining washtrades!

Like making a spam filter purely algorithmic is a very hard problem, but people can tell what is and isnt spam. Oh, like curating, get the people involved!