While Compendium I and II are good books,
they brought in a number of rules that some found to be either unrealistic
or unbalanced. Part of the problem was that many of these rules were
play-tested only in specific genres as part of a world book. Some
of these genres include a view of what is realistic that is a more liberal
and/or reflects (not surprisingly) a generally favorable view of how effective
the "cool" techniques of those settings are. Also, Martial Arts introduced
a set of mechanics that handled combat at a more detailed level than in Basic.
While that was certainly appropriate to the genre, it created a system that
had different levels of detail for different combat techniques.
The compendiums are being, rightly so, mentioned as books that people should buy to get rules that were previously scattered among numerous books. However, I also usually warn people about the concerns mentioned above (if only so they can judge for themselves wether to use them). It occurred to me that giving advice to look out is incomplete unless you also give advice on what to look out for. Thus I decided to write one person's take on where the books need to be changed to keep them realistic and balanced. These comments are based, in part, on discussions on the GURPS mailing list, but are certainly not complete and can't possibly include all views on the issue.
Extra Fatigue - Some find 3 points/level too cheap. Primarily due to it use by mages. You could note that it as more useful to a mage than advantages that give a +1 reaction, a +1 to senses, a sense of fashion, or a level of Hard to Kill (all 5 points/level), but one can make up his or her own mind. In the end, however, the main thing a GM should be aware of is that fatigue is a major limiting factor for magic. When the spells in Magic were created, mages needed to raise ST to get more fatigue. While an Extra Fatigue advantage was needed, to reduce this cost all the way down to 3 points/level (or even 2 points/level if one applies the "magic only limitation") makes fatigue a lot cheaper and magic more powerful. (Also, one advantage before was that you didn't have mages who were _always_ weaklings or always strong. With 3 points/level, there will be a significant incentive to always play a weak mage.) I would recommend 5 points/level and not allow any significant limitation for magic use only (magic use is already the primary point cost driver). I would allow a -80% limitation, "doesn't apply to magical fatigue" for campaigns that use fatigue but don't have magic.
Gadgeteer - While not unbalanced, this is a bit cinematic, particularly when allowing the production of high tech devices, unless the GM is strict. In reality, it is pretty hard to whip up a nifty gadget (particularly higher tech ones) like that. New devices, even prototypes, don't come out of a few weeks in the lab.
Manual Dexterity - This should definitely not be applied to combat
skills (which is what I think was intended, but this is worth reinforcing).
Boxing - This is a bit unrealistic with respect to fighting against armed men. See the comment regarding Judo (below, under "Maneuvers").
Katana - When used one handed, the Katana does a point more damage
and you don't have to ready the sword every time you swing. When you
use it two handed, you still do one more point of damage and you also, when
you are lightly encumbered or less, get a 2/3 parry. Additionally, a
Bastard Sword requires you to learn two different skills to be used with one
or two hands while the Katana one requires one. On top of all this,
the katana costs $100 less. (This is covered in excruciating detail,
since was the subject of a number of heated discussions, in the file katana.objections)
One alternative would be just to use the stats for a Bastard Sword. However, some feel it should be more balanced than a Bastard sword. An alternative is the katana uses the Broadsword skill one handed and the Bastard sword skill two handed (you could make up two new separate skills, but I don't really see any reason). The katana is said to be bit better balanced than a Bastard sword, so it does one point less damage two handed and does the same damage one handed but, unlike a Bastard Sword, it doesn't need to be readied. This is still a little generous (it beats the bastard sword in the role it was designed for since loosing the ready is more important than a +1 to damage) but not objectionably so.
Main Gauche - The 2/3 parry presumably comes from this skill being a fencing style. This skill should only be used (or at least give the 2/3 bonus) when fencing (ie using a fencing weapon) and I would make if P/H like other fencing skills. Also, it is not clear why you lose the -1 to parry that a knife has. Getting better at knife fencing by training is already covered by higher skill level.
Parry Missile - This is cinematic skill. Parrying a thrown rock is tricky (harder than baseball since it is coming right at you, jamming you up, and you can get it while it goes by you). Parrying an arrow is nigh impossible.
Monowire - These are cinematic. Besides the fact that a monowire is impossible to make by today's standards and is clearly significantly higher than TL 8 or 9, it is not clear that they would even be effective. It can be argued that a wire that thin can pass trough a material without causing any significant disruptions (since bonds are broken at the molecular level and simply reforming behind it and the channel is so thin as to only kill off a few cells). This should be reserved for campaigns that are cinematic or in which monowires are a part of the genre convention.
Short Staff - The 2/3 parry seems excessive, especially when compared to a shortsword. The apparent justification is that you hold it in the middle (so it is balanced and light), in which case this needs to be made clearer in the rules. One can see this giving some advantage to parry (as good as a staff is debatable, but not clear enough to constitute an objection). Then, however, we need to drop the damage down from sw+1 and sw to maybe sw-1 to reflect that one will do less damage when you hold it in the middle. (You might leave the Jo stick at sw+1 but drop the 2/3 parry, but either it is light enough to be easily moved one handed or it is heavy enough to do broadsword damage, but not both).
Throwing Stick - The base damage is as much as broadsword and seems excessive.
Tonfa - This is another case where a 2/3 parry seems to have been given out much to casually. Holding against your arm will lose as much in reach as it will gain and it is not clear why the parry should be any better. The main advantage would utility in close combat.
Military skills - Unless you are running a very military oriented
campaign, you are not going to need skills that narrowly defined. I
personally would keep just Forward Observer, Naval Tactics, and Naval Strategy.
Close Combat - One problem with balancing unarmed combat is the immunity to attacks in close combat. It is not that hard to hit legs with sword or to conduct a wrap around shot with many weapons. I would reduce the penalty to a -2 to -4 (though I might also drop damage to 1/2 damage).
Disarming - see Arm or Wrist lock.
Dual Weapon Attack - This is should be labeled cinematic. GURPS already allows you to attack once a second which, for most types of melee, is already at the upper end (mostly because PC's don't take a moment to collect their thoughts the way a real combatant would). There are exceptions, but they are just that, exceptions. If you watch more skilled fighters, they don't attack more often, they attack more effectively. When you attack more than once a second you simply don't have time to consider what you are doing tatically and tend to flail away.
Hit Location - You really can't train yourself to hit one location more accurately without being able to hit other things more accurately, which is almost the definition of high skill (there might be an exception in martial arts where set combinations seem more important, but I wouldn't use it outside of fights between martial artists). Hitting one location in battle is different enough each time that you can't learn to hit it by rote. What you do is just get better at hitting any location better, no matter what it is. This maneuver has, in fact, been the basis for critism of GURPS as allowing some to make a character being a "specialist in lopping hands off".
Horse Archery - Shooting an arrow off a galloping horse just seems too difficult to justify being raised at 2 points/level (let alone at the 1 point/level for the first level). I would either just have the player raise skill level or make it a maneuver that goes up at 4/level (for even the first level).
Jump Kick - It's mentioned that this maneuver is discouraged in real world training. However, the disadvantage that causes this to be true isn't clear. For an attacker with decent skills, the odds of falling are slight, especially compared to the utility. The roll to fall might be made harder. Also, I would give the attacker at least a -2 to -4 to subsequent parries.
Spin Kick - This a lot to do in a round (spin around, feint, and
attack) and still be able to defend normally. I would have the person
be at penalties to defend or have to make a skill roll to recover before
the next round.
Basket Hilt - This is a lot smaller than a Buckler (which is usually thought to be about 12" across), which only gives PD 1. I would restrict it to providing PD and DR for the hand.
Dau - Given the weight and damage, it should need readying.
Cloak - The main problem is how do you block swung attacks with a cloak? The only way, esp for high swing, is contact with the cloak around your arm. This is consistent with the skill being used in fencing which concentrates on thrusting techniques. Thus, damage to penetrate should be applied on almost all swung attacks (esp against non-fencing weapons).
Katar - It not clear why it does so much damage (holding the blade perpendicular to the knuckles won't give you more leverage) or how you parry better with it (the parrying surface isn't any bigger and, if anything, the angle is poorer). The only real difference is that it would protect the hand better (say PD 2 to the hand only).
Knife Wheels - Again, compared with a Buckler, why do they get PD 1 (see basket hilt). Also, like the Katar, why do they do more damage?
Slashing Wheels - Again, why PD 1 when compared to a buckler?
Flight - This appears to be listed under cinematic rules but it is not clear if they are intended to be cinematic. Actually, these rules seem too harsh on a species that was born to fly. They make fighting harder for a creature born to fly than for someone on a galloping horse. A mode of location you were born to should be easier than one you had to train at (and in which you are also controlling another creature at the same time). I would not give species that are born being able to fly, and don't have the Cannot Hover limitation, penalties to attack or parry (whether they are using a weapon or not). I'm not sure I would give any flyer that can hover a penalty.