In GUMSHOE what can be done to make the combats more dynamic, more exciting?
First, here are some suggestions that don't involve hacking the rules.
None of this, however, fixes a particular problem with Gumshoe combat: you can always hit if you spend enough points. To fix that, you need to start hacking. Here is a small hack and a big one.
This last hack, although it's a big one, has the pleasing effect that spending points makes you more likely, but never certain, to hit. Also, you get to roll handfuls of dice. You'll find the probability of hitting, when spending different numbers of points, is more or less the same.
Finally, here's a fun hack, stolen from Tunnels and Trolls.
"Dynamic" and "exciting" are two different things.
Dynamic fights are in constant flux, with the upper hand vacillating back and forth between the various antagonists. "Dynamic" is fun because the players are constantly swinging between the euphoria that they're going to win and the fear that they're going to lose.
For dynamic fights, first, balance the enemies, statistically, so that they're close in general power level of the Investigators. Second, work out in advance ways that the antagonists can amp up or fall down in the course of combat, in order to swing the pendulum in whichever direction it needs to be swung at any given time during play. If the antagonists are getting stomped, perhaps there are reinforcements waiting in the wings. If the antagonists are wiping the floor with the Investigators, perhaps some environmental advantage for the Investigators becomes apparent.
Exciting fights involve interesting stakes. To be interesting to RPG players who've probably already fought every imaginable variation of "fighter vs. ork" in the breadth of tabletop roleplaying, stage fights that involve stakes other than the wholesale slaughter of the other side. Consider fights where one side must escape, where one side must gain possession of some macguffin, where one side must accomplish some feat within a time limit, and so on.
Statistical balance, on-the-fly opportunities to swing a fight's momentum, and interesting stakes.
One thing I've done in a few different systems is to use three-round combat. Instead of going back-and-forth until everything is nicely resolved you only use three rounds.
Round 1: Beginning Stakes.
Players state their stakes/intentions for the beginning of the combat. Example: we surround the thugs, we keep the Deep One at bay, etc.
The GM, as the NPCs, does the same. Example: O'Bannion's thugs use suppressive fire to keep the Investigators froma advancing, the Deep One charges the Investigators
Then you roll and resolve that round. Keep track of who "won."
Round 2: Intermediate Stakes / Reactions
After reviewing the results of the last dice roll(s) and updating the situation, the players and GM both declare new stakes based on their changed positions.
Player example: After failing to surround O'Bannion's thugs, the players' new stakes might be "Take out as many of them as possible!" Or, after successfully keeping the Deep One from advancing, their new stakes might be "Advance and try to subdue the beast"
Round 3: Final Results
Who won round 1? Who won round 2? You'll either have a 1-1 tie, or a 2-0 winner. The 2-0 winner gets Round 3 to narrate how they "win." For example: The Investigators managed to hold the Deep One at bay, then subdue it, so in this round they get to narrate how they capture the beast. Feel free to require point spends or rolls, but let the Investigators win. Failed rolls might just complicate things later, but not eliminate success right now.
In the case of a 1-1 tie, use Round 3 for a final combat moment. O'Bannion's thugs held the Investigators from getting the advantage, but in round 2 the Investigators killed or wounded enough thugs to stay in control. Each group now states their final stakes. The winner gets to narrate how they win. The loser has to yield.
I find that this three-round model tends to make people think strategically about combat and keeps it short and "punchy." There's a lot less flailing around. However, you will need to balance how the point-spends work, since combat might not last as long as usual.