04/07/11 Edit: For whatever reason, I didn't see fit to discuss how terrifyingly broken Marathon has been for the past 10 years. Sensible fixed-goal Marathons (e.g. the simple 200-line one in Tetris DS) are rare. We're usually stuck with the abortion that is variable goal Marathon. For the sake of completion, let's first talk about...
I couldn't tell you how many times I've had to explain to some poor casual player looking to improve their Marathon score that nothing makes sense anymore. If you aren't acquainted with the way things work these days, you'd think they were doing everything right: going for those back-to-back tetrises and t-spins to earn more points per line. Much to the player's dismay, -- no matter how many high scoring maneuvers they pull off -- they can't get much more than 700k points by the end of the game. Meanwhile, replays on the leaderboards show scores on the better side of 1 million points composed of almost nothing but singles. So, what on Earth is going on?
The place where things start breaking down is that Marathon doesn't advance to the next level after fixed number of lines have been cleared as in most other marathon-style variants. Instead, it uses a mysterious metric called "Goal." As it turns out, the Goal-based progression is the source of all our problems: floor(base_value_of_move/100) Goal points are deducted whenever lines are cleared.
As a result, the number of points earned by a player unaware of the abuse necessary is very nearly a linear function of their progress. Other than two methods of abusing the Goal formula, the only variance in score comes from points earned for dropping pieces. Unlike a game that ends after a set number of lines, going for moves that award a higher number of points per line cleared is no better than just skimming low value clears. (In fact, -- ignoring the abuse for a moment -- you'd be worse off with tetrises and t-spins for having reached the end of the game in fewer lines and skipping out on points for dropping more pieces in the process.)
So what can one do to earn "extra" points? The first of the two tricks available abuses the round down in the Goal deduction. (For example: doing a tetris with 1 Goal left brings the counter down to -7, which advances the player to the next level and resets the Goal counter to Level*5 without considering the "extra" Goal used just prior.) The combo bonus in current guideline games is worth (combo_step*50) points. This makes a Combo 1 clear conveniently round down to 0 additional Goal deducted. The best way to capitalize on this is to fill levels with single-single chains -- 250*Level points for 2 Goal, 125% the standard exchange rate. As previously mentioned, scumming singles will stretch out the game duration and net you additional drop points as well. The second trick is to abuse the lack of carry-over between levels by setting up for a high value clear with 1 Goal remaining. Typically, this means whittling the Goal counter down to 9, performing a tetris to set up the back-to-back bonus, then doing a t-spin triple (or double if kick t-spins are not recognized in this particular game). With a back-to-back t-spin triple crossing the end-of-level border, that 1 Goal earns you 24 times as many points.
Explaining to casual players why they must cast aside their reasonable intuition -- "Do things that earn a lot of points at once! Yeah!" -- in favor of almost the exact opposite -- "No! Avoid getting a lot of points at once except for at these specific times!" -- seems hopelessly difficult at times. People sometimes joke that the game shouldn't be called Tetris anymore because of T-Spins, but spin bonuses aren't nearly as disparate from the tradition of the game as the broken mess that is modern Marathon. I'd say the only ways to go about fixing it are to either value a fast completion over heavy systemic abuse or (preferably) to scrap the Goal system entirely and go back to directly rewarding valuable moves. (What a thought!)
...and now back to our regularly scheduled re-runs...
(First posted June 15th, 2009 on Tetris Concept)
First off, I would like to second all of jujube's suggestions. Combos are exceedingly worthless, and the tips on TetrisFriends.Com like to suggest otherwise. The rewards for back-to-back along with the drop points you get for keeping your stack short in the meantime are definitely more valuable. Things need to be rebalanced in some way or another in that regard; either back-to-back needs to be weaker or combos need to be stronger. The only instance where this is not the case is versus. If anything combos might be a mite too strong there. They're very interesting in TOJ though, so those should either stay how they are or be weakened slightly. (04/07/11 Edit: In light of the development of high power combo openers, it's safe to say they should be reworked... but that's a write-up for another day.)
I also agree that t-spins should be brought back to be in line with or weaker than tetrises. They're interesting, but they've turned out to be a dominating strategy when they should probably be a supplement. Also, their dominance means that the soft drop speed has to be balanced in order to prevent them from becoming even _more_ powerful. If you limit their power, you no longer have to unduly limit the speed of the soft drop. I think it would be for the best to tweak the soft drop speed up a little bit to make it worth using for fixing mistakes, so t-spins probably need to get a little weaker to balance everything out.
I've also felt that t-spins are too exclusive, as well as occasionally inconsistent. Why just the T? There are many interesting twists that can be performed with the other pieces as well. And why just 3-corner? Why sometimes only those without wall kicks? My suggestion is that things be a little bit more open when it comes to rewarding twists. With the twist bonus weakened a bit, it would be safe to open things up to twists done with _all_ of the pieces. For detection, use a simple immobile algorithm -- check if the piece would collide when shifted left, right, or up one cell and award a twist bonus if all collision checks return true. Twists would become interesting supplements for recovery or defense, while still maintaining the scoring power of the powerful back-to-back tetris. I imagine it could lead to a lot of dynamic recoveries -- I think it'd be a mechanic worth checking out.
As for back-to-back, I'm not entirely sure how I feel about it. With the Bag randomizer, you can certainly maintain it. However, it doesn't seem entirely right to penalize the player for making a good stacking decision by skimming or taking away such a significant portion of their score for one single mistake in the game. The useless soft drop makes things hurt even more; you might be faced with the choice of eating a delay while you wait for the piece to drop or continuing to play at top speed and lose your back-to-back. Should back-to-back be as strong as it is? Should it exist at all? I'm not entirely sure myself.