To recap, Hearts of Iron 3 (HOI3, developed and published by Paradox Interactive) is an amazingly detailed 1936-1948ish-era simulation. Generally speaking, you're supposed to fight World War 2, although it is rumored that careful diplomacy can avert it.
You take on the role of a "guiding spirit" of sorts, a veritable Illuminati of your country. You can take actions as diverse as appointing a new cabinet minister to upgrading your aircraft carriers' engine to setting up trade agreements. This is not a simple war game, but an attempt to simulate the economic, geographical, technological and historical factors that surrounded the war. One could imagine learning a lot by watching a replay of the actual war played out in the Clausewitz Engine used in HOI3.
The problem this introduced was that running a war is a complex business. While some obsessive-compulsives would enjoy countless hours poring over minute decisions, most players would be primarily interested in a few areas (perhaps sea battles and logistics) while loathing the time spent on the areas they considered less interesting (maybe land battles and trade relationships).
The solution to this problem was brilliant. Gameplay was divided up into several areas: Diplomacy, Production, Technology, Politics, Intelligence, and Military. The player could set each of these areas (and sometimes sub-areas) to AI control. The AI, in general, had to be able to make these sorts of decisions in order to run its own countries, so why not allow the user to let the AI make those decisions for his?
This solution is both the pride and the downfall of HOI3. The pride because this is obviously the way all future games of this scale will be written. It is both such a good idea and so obvious it's hard to believe no one had thought of this before. It is the downfall because the AI plays like the "after" picture of "Flowers for Algernon". You know, after the super-smart serum wears off.
Most of the additional features in the Semper Fi expansion can be viewed as attempts to work around the AI problems.
Probably the most important change is the addition of an "Arcade" mode. The point of this mode is that supply is drawn directly from the national stockpile. This differs from the original HOI3 model in that the original model included a full-blown logistics simulation. The engine would keep track of the path a supply would need to take to get to a particular unit. If it went overland, the engine would keep track of how many supplies were going through that territory and limit the flow of the infrastructure in that territory could not handle the traffic. If the supply had to go overseas, the engine made sure there was a transport to take it, and both ports were required to have enough handling capacity. It was a practical plan for Germany to allow the U. S. to land at Normandy then sink enough supply boats to make them withdraw. The only problem with the idea was that the engine could not track supply paths in any meaningful way. Examples abounded of supply paths that were laughably suboptimal, often crossing the same territory multiple time. Arcade mode is an admission of defeat by the developers.Another change is the ability to allow AI HQs to auto-reorganize. One of the first things HOI3 players would do was reorganize all the units into command hierarchies that made sense. It is now possible to tell the AI to reorganize itself. This saves a lot of time and aggravation. On the other hand, it looked like some of my units were reorganizing every other day. It didn't effect play that I could tell, but did not raise my confidence level in this feature. Even with this change the new Command Hierarchy Browser and Visible Command Hierarchy features indicate that Paradox is not entirely comfortable with the way command hierarchies work.
Another change allows to player to manually define a theater. HOI3 defined theaters for the player. For example, the U. S. might have a European and a Pacific theater. Again, the AI fell down on the job, often defining theaters that had no connection with what was actually going on. Semper Fi allows the user to define their own theaters. Again, this is a feature to correct an AI failure.
There are a few new things that add to play, rather than fix underlying problems. The player can now set air and naval stances separately from land units. For example, Britain could set land units to be defensive, while air units could be attacking. This is bolstered by the addition of Air and Naval map modes. These modes show, on the main map, things like AA batteries and air strips along with more abstract ideas such as areas covered by air superiority orders.
Other new features add color. One might build enough ships to have the "Grand Fleet" Strategic Event fire. This would give your nation a bonus to ship-related activities. Strategic Events provide nice ways to reflect the consequences of actions (yay! I own the Strait of Taiwan) that otherwise wouldn't be reflected in the engine (now I can move supplies more efficiently).
Also nice are the Historical Battle Effects. Even though your war will be rather different than the real war, certain events are likely to occur in similar manners in both. The Germans may end up besieging Stalingrad in your game, for example. It is a rather natural thing for a German player invading Russia to do. When this happens some flavor text will appear on the screen describing the real battle. It can be fun to see these pop up and then do better than the historical analogue.
HOI3 was built for hard-core grand strategy war gamers. It stands alone at the top of the heap for breadth, depth and detail. Semper Fi does not change this. Instead, it attempts to fix some of the more obvious flaws by modifying how a troublesome AI controls things. This expansion is worth it for fans of the original - some of the problems it fixes really needed fixing and were beyond the reach of a mod. Those who were not interested in the original game will find nothing here to change their minds.
This reviewer must admit to feeling bad for being so tough on Semper Fi - he likes the game. However, it wouldn't be fair to (both my) readers to allow personal fondness for a game to overrule objective judgment. Paradox is onto something here. It just isn't ready yet.