Ages ago, I posted a thread asking questions about how to run a successful and dynamic martial arts game. I doubt anyone remembers it, but I found many of the suggestions useful. I've finally finished my playtesting and actually run the damn game, and I thought I would share my experience.
1. GURPS actually runs a really great martial arts game.
When I posted my questions, I was concerned about how well GURPS would actually handle the cut and thrust of a martial arts fight. It's true that GURPS has lots of options, but in my experience (up to that point), most of them boil down to configuring your attack and defense into optimal values and then using those again and again. A few maneuvers have lasting consequences, like feints or AoAs, but those tend to last no longer than a single turn, and then everything resets to its original point.
What I actually found is that those one-second consequences tend to last, and once you understand how they work, you end up with a game of chess. For example, if I attack you with Karate and you successfully parry with Judo, then you might lock me into an arm lock or judo throw me, and thereafter start kicking me in the head. So, I need to make sure that if I attack you, I do so in a way that you can't defend properly against, which means that I should feint against you, but knowing that I'll feint, you might evaluate and, if so, I might be better off evaluating first as well. And so on.
GURPS also has other tricks, as it turns out. Targeted Attacks and Combos, when used too often, become predictable and your opponent gains a bonus to defense, which means you should use them sparingly, when circumstances warrant it. It's a subtle mechanic, but I found it made quite a difference.
2. Use Techniques Sparingly
I honestly find 4e's treatment of techniques slightly frustrating. For one thing, I'm not sure why they need to have an Average/Hard split. If you want to charge 1 point more for a trick, make it -1 more difficult. What you currently have is that Hard techniques are easier to pull off by default than Average techniques (That is, both a Hard technique at -2 and an Average technique at -3 cost 3 points to max out, but the Hard Technique is paradoxically easier to pull off by default, and less worth your time to study).
Moreover, in 3e, you were required to take every technique in a martial art to "know" that martial art. Now, this caused problems, but it meant that two different styles that both had Judo and Karate were distinct, because one had X and Y, and the other had A and B. Moreover, having all those techiques listed on one's sheet encouraged players to try those moves out. If you know you're good at spin kicks, jump kicks, feints and hammer-fists, you're not going to bother with exotic hand-strikes or head-butts, and you have a distinct fighter that encourages the player to do more than just "Attack/Attack/Attack!"
In 4e, it makes little sense to have more than a couple of techniques, and so you have to pick and focus on what you want your character to be good at. If you have a Judo of 18 and that's "enough" for Arm Lock, you don't take the Arm Lock technique, while another fighter might focus almost exclusively on Arm Lock. You get distinct fighters within a given style, which means you can get a lot of mileage out of a single style, and that's great. It's just that techniques aren't where you go to write out how your character fights...
3. Use Signature Moves
... you use Signature Moves instead. I took Toadkiller Dog's advice about writing down the more complex moves that a fighter might use and giving them swell names. Made all the difference in the world. Take my example above about the Judo 18 guy who doesn't bother with Arm Lock because his Judo is more than high enough. He can and should still use Arm Lock, and if you want to remind a player to do that, you simply note down a special armlock "signature move."
You can even construct very complex moves that take advantage of several unique mechanics to make a killer move. For example, one samurai had a signature move that involved shifting to a defensive grip for a parry, while side-slipping (this being Chambara, was worth +2), and then spending 1 fatigue for a total of +5 to parry, which he turned directly into a riposte, followed up with a Counter Attack (for a total of -7 to his opponent's defense) as a thrust (since defensive grips are terrible at swings) to the vitals. Nobody's going to think of that move in the middle of a fight, and there are numerous things that you need to calculate (can you take a -5 to your defense with a riposte? What's your effective skill between Counter Attack and going after the Vitals?). By working it out in advance, you only have to glance at a piece of paper to pull of this off. The extreme potential complexity of GURPS martial arts becomes a feature, not a bug, and people begin fighting in a very complex way.
(And this isn't entirely unrealistic. The whole point of things like kata is that they teach you to use very complex tactics that you'd never think of in the middle of a harried fight.)
By giving the player's signature moves and including signature moves with the NPCs, I found that my fights exploded with details and rich tactical depth. Nobody made straight attacks, because they had customized their fighting styles to support far better tactics, and clever opponents would exploit the weaknesses of a given move with their own tactics, which would lead to rapid exchanges of constantly flowing and evolving tactics.
4. Perks are where it's at!
As I said before, Techniques aren't the best way to add lots of character to your fighter, because you can only really take so many and eventually you're better off simply making him a better, all-around fighter. That's where Perks come in. Perks are almost always worth it (unlike Techniques), and they tend to encourage specialized tactics. Secret Styles, Trademark Moves and Finishing Moves all encourage and expand upon the Signature Moves above. Tricks like Iron (Body Part), Teamwork, Resistant to (Chi Power), Schticks, Drunken Fighting, Sexy Feints, and so on, all shift and change the nature of the character in neat little ways or provide him powerful advantages that, if he's clever, he can really exploit. A good martial arts game is about turning every character into a "fighter," and still seeing plenty of diversity. Perks really provide that. I'd never try to run GURPS Martial Arts without perks.
5. GURPS, as always, exceeds expectations...
I mentioned already that GURPS ran a great martial arts game. I didn't mention how often I thought a rule needed to be changed, when it didn't. For example, many people argue that Evaluate is a poor tactical choice, so few people use it. However, I found that there are so many consequences to attacking a capable foe (Counterattacks, ripostes, judo throws, arm locks, exposing the trick to your secret Combo awesomeness) that people often fell into Evaluate when it became obvious that their opponent was more powerful than expected.
Likewise, I worried that GURPS would overwhelm some of my players (we had one player who generally hates system and fears any game with any level of complexity much beyond free-form), and yet, I found that between the Signature Moves and how GURPS normally handles, the game was surprisingly intuitive, people could do what they wanted, the fights played out swiftly, and even the system-phobe had a great experience.
6. ...but change is good.
Even so, I did make some changes. I used Icelander's Beat rules (that is, beats apply their penalty to both the user's attack and defense and at 5+, there's a chance that it'll unready the weapon entirely), though I allowed people to defend against Beats with DX-based weapon skill, in addition to ST-based weapon skill (I found that forcing players to always defend with ST-based weapon skill unfairly punished weak fighters, especially since ST-heavy fighters tend to spend lots of points in their skills to compensate for their low DX, and so when they beat, they REALLY beat). I highly recommend them, as they make Beats useful for everyone, and it means that you're taking a calculating risk every time you parry or let someone parry your blade, especially if they're stronger than you are.
I did eventually modify Evaluate some. I find it odd that one can make a Defensive Attack (inflicting damage) and gain a +1 defense bonus, but Evaluating gives you no defensive bonus, and so I made a little "cautious fighter trinity" of moves: In my version of the rules, Evaluate grants a +1 to your defenses, you can choose to forgo the +1 defense from a Defensive attack to instead gain a +1 evaluate bonus, and you gain an +1 evaluate bonus if you Wait and nothing happens. I also allowed players who were evaluating to make rolls against certain skills or IQ to understand something about their opponent. I found this made all three moves more interesting, but didn't unbalance anything in particular.