-
Notifications
You must be signed in to change notification settings - Fork 7
Roadmap
Plans are scattered in many places: the file 'notes', the TODO block comment at the top of harris.c, items in the game manual tagged as [not implemented], and a file 'changes' which isn't in the repo, but details things I want to change but which would break saves (it lives with my savegames, which seemed to make sense at the time).
Also, specific tasks go in the issue tracker.
The first big aim is to get all the key gameplay features through to 1945 at least partially implemented.
- DONE: Start points. You can now start at 03-09-1939, 15-05-1940, 14-02-1942 or 04-03-1943.
- DONE: Tame Boar fighter control. In particular, guessing a target and devising an interception plan; this is going to take better AI than I'm used to designing.
- There are also some other AI issues (like fighter production and bomber routeing) whose current implementations are rather dumb, but just about good enough for a first version.
- Window. Basic behaviour is implemented (as far as Tame Boar is concerned), but a UI to control its usage is still needed (ie. diversionary raids Windowing more heavily).
- Gee-H. In principle, this could be simple - if I'm happy with "unlimited concurrent users". However, accuracy demands "limited but more than Oboe gets" and "unequipped planes can follow a Gee-H bomber's lead".
- Game-end conditions. Low morale/confid should lose, and reaching VE day should win (though, long-term, the intention is to have some kind of score that's counted up at VE day, to measure how well you won).
- 1944+ changes to targets. Liberation of cities; invasion-related targets; V-weapon targets; Operation Manna.
While balancing the game is an ongoing requirement every time a new mechanic or entity is introduced, this is easier if the existing game is already balanced. This means not only getting the overall difficulty right, but also ensuring that each mechanic or entity contributes an appropriate amount to the challenge or the toolset with which the challenge is met. For example, it shouldn't be the case that getting H₂S is necessary and sufficient to hit any and all targets, but nor should it be the case that H₂S is so weak to not be worth having. Similarly, the advent of German nightfighter radar shouldn't suddenly make losses unsustainable, but nor should it be something you can entirely ignore.
One strategy for achieving balance is to get lots of playtesting, and find out what things players rely on, what mechanics they think are 'broken', etc. However, that's not enough on its own; true balance will require careful judgement to trade off gameplay issues against realism issues. For instance, realistically the Manchester should be so bad that there's no reason to build any (instead sticking with Wellingtons at least until the Stirling comes along). Yet, gameplay demands that Manchesters do something to make the game more interesting, because otherwise the 'struggling with obsolete bombers' phase drags on altogether too long. Similarly, there's no realistic reason for the player to want to build Halifaxes once Lancasters are available in sufficient number, and thus no reason for the Halifax Mk III to be in the game, yet historically the RAF did end up with Halibags. To what extent should the player's optimal strategy coincide with what historically happened? What about role-playing - if there were reasons to do X in the real War, but those reasons aren't modelled (or even modelable) in-game, should other mechanics be tweaked to make X more desirable? These are difficult questions, shared by all historical games, and they can't be solved by throwing coding-hours at the project. Can they be solved by throwing player-hours at it?
The game would be a lot more fun if the player's performance over the years were tracked, so that success or failure could be judged by more than just "Do I have more planes than before?" and "How much blue is there in the target list?"
For instance, the player might want statistics that allow the following questions to be answered:
- What tonnage do I get per aircraft dispatched (or lost)?
- What proportion of aircrews finish their tour of 30 ops? How does this vary by aircraft type?
- How do the various causes of losses compare? Do fighters get a bigger share when I raid more distant targets? Is radar-predicted flak a major danger?
Most of these should be computable from the event logs now embedded in every save file. Several graphs are already provided by Python scripts in the stats/ directory.
Two key things should depend on the time of year:
- The weather. This includes fog-at-base, which should limit operations until FIDO is developed.
- DONE: The length of the night (making long raids on summer nights risky).
In general, the weather model needs improving. It's quite computationally expensive, yet gives pretty poor realism; it's only barely good enough for gameplay. The actual algorithm (specifically the generation of macroscopic perturbances) is frankly embarrassing. However, it was hangups about the weather model that delayed starting this project for years, and I don't think I can tackle it alone.